Datasets:
Languages:
English
Size:
100K<n<1M
Tags:
information-retrieval
dense-retrieval
contrastive-learning
synthetic-data
query-generation
hard-negatives
License:
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: TypeError
Message: Couldn't cast array of type
struct<file: string, hf_dataset: string, qwen_model: string>
to
{'id': Value('string'), 'categories': List(Value('string')), 'update_date': Value('null'), 'hf_dataset': Value('string'), 'qwen_model': Value('string')}
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1816, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2303, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1852, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2149, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
TypeError: Couldn't cast array of type
struct<file: string, hf_dataset: string, qwen_model: string>
to
{'id': Value('string'), 'categories': List(Value('string')), 'update_date': Value('null'), 'hf_dataset': Value('string'), 'qwen_model': Value('string')}
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1683, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1869, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
queries list | queries_top1 list | document string | source string | quality_score int64 | meta dict | hard_negative_idxs list |
|---|---|---|---|---|---|---|
[
"accelerated proximal stochastic variance reduced gradient method",
"how to implement momentum acceleration in svrg with one variable"
] | [
true,
false
] | This paper proposes an accelerated proximal stochastic variance reduced gradient (ASVRG) method, which incorporates a simple and effective momentum acceleration trick. Unlike most existing accelerated stochastic variance reduction methods such as Katyusha, ASVRG utilizes only one additional variable and one momentum pa... | arxiv_abstracts | 4 | {
"id": "1810.03105",
"categories": [
"cs.AI",
"cs.CV",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
12906,
43729,
55343,
34374,
55282,
51371,
15789,
58457,
49810
],
[
41009,
29304,
38414,
38461,
20007,
34374,
20561,
41448,
31703
]
] |
[
"scientific relation extraction using concept embeddings",
"how to improve relation extraction with character level attention"
] | [
true,
false
] | This paper presents a submission for the SemEval 2018 Task 7 shared task on semantic relation extraction and classification in scientific papers. The approach extends the end-to-end relation extraction model of Miwa and Bansal by incorporating enhancements, specifically a character-level encoding attention mechanism us... | arxiv_abstracts | 4 | {
"id": "1808.08643",
"categories": [
"cs.CL",
"cs.IR"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
27432,
13760,
26632,
55880,
59052,
12535,
7190,
2761,
34434
],
[
45540,
37364,
29660,
2201,
36151,
13760,
739,
14855,
14610
]
] |
[
"sentence coherence classifier gpt-2 reinforcement learning",
"fine-tuning language model final layer coherence objective"
] | [
true,
false
] | Conditional story generation and contextual text continuation have become increasingly popular topics in the NLP community. Existing models are often prone to output paragraphs of text that gradually diverge from the given prompt. Although the generated text may have a reasonable perplexity and diversity, it could easi... | arxiv_abstracts | 4 | {
"id": "2009.06358",
"categories": [
"cs.CL",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
30088,
27571,
18808,
10964,
54859,
48449,
2583,
1424,
22214
],
[
30088,
35483,
31483,
24828,
599,
7207,
18808,
46001,
50661
]
] |
[
"variance reduction quantized stochastic gradient descent distributed learning"
] | [
true
] | Due to its efficiency and ease of implementation, stochastic gradient descent (SGD) has been widely used in machine learning, particularly as a popular optimization method for distributed learning. Recently, quantized SGD (QSGD), which adopts quantization to reduce communication costs in SGD-based distributed learning,... | arxiv_abstracts | 4 | {
"id": "1901.03040",
"categories": [
"cs.LG",
"math.OC",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
18156,
112,
12964,
15247,
9751,
10296,
6706,
37968,
12754
]
] |
[
"predicate paraphrase scoring using event coreference annotations",
"how to improve event coreference models with paraphrase features",
"distant supervision for predicate paraphrase ranking"
] | [
true,
true,
false
] | We study the potential synergy between two different NLP tasks, both confronting predicate lexical variability: identifying predicate paraphrases, and event coreference resolution. First, we used annotations from an event coreference dataset as distant supervision to re-score heuristically-extracted predicate paraphras... | arxiv_abstracts | 4 | {
"id": "2004.14979",
"categories": [
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
6757,
42779,
46619,
13719,
53492,
47643,
13636,
25679,
48967
],
[
47643,
35601,
4779,
35500,
14612,
3160,
6757,
5108,
49168
],
[
23219,
51950,
17222,
46619,
10499,
20893,
6757,
18540,
9106
]
] |
[
"visual drag and drop interface deep learning no code paradigm",
"how to make deep learning accessible to non experts without programming",
"system usability scale and task load index comparison visual programming vs traditional coding"
] | [
true,
false,
true
] | Deep learning is one of the fastest growing technologies in computer science with a plethora of applications. However, this unprecedented growth has so far been limited to the consumption of deep learning experts. The primary challenge is a steep learning curve for learning the programming libraries and the lack of int... | arxiv_abstracts | 4 | {
"id": "1905.02486",
"categories": [
"cs.HC",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
34210,
45822,
32379,
40821,
49563,
33242,
42496,
25579,
54367
],
[
25959,
8241,
50890,
28823,
46457,
7038,
31615,
55319,
58439
],
[
21447,
24376,
16560,
7529,
54697,
45410,
57291,
30302,
15833
]
] |
[
"topology design distributed consensus random link failures",
"how to optimize sensor network convergence with communication budget",
"algebraic connectivity condition mean square convergence consensus"
] | [
true,
true,
false
] | In practice, communication among sensors in a network is subject to errors or failures at random times, costs, and constraints due to scarce resources like power, data rate, or bandwidth. The signal-to-noise ratio (SNR) is a main factor in determining the probability of error or communication failure in a link, which s... | arxiv_abstracts | 5 | {
"id": "0704.0954",
"categories": [
"cs.IT",
"cs.LG",
"math.IT"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
10192,
23486,
58260,
51160,
8190,
14890,
45227,
32452,
50397
],
[
45227,
2398,
1918,
22941,
17331,
5319,
1635,
47882,
22190
],
[
23483,
19520,
20134,
50397,
58260,
53568,
54615,
58385,
18556
]
] |
[
"graph neural networks oversmoothing random walk mixture",
"how to handle oversmoothing in graph neural networks with random walks",
"GESM graph entities step mixture inductive transductive learning"
] | [
true,
true,
true
] | Existing approaches for graph neural networks commonly suffer from the oversmoothing issue, regardless of how neighborhoods are aggregated. Most methods also focus on transductive scenarios for fixed graphs, leading to poor generalization for unseen graphs. To address these issues, we propose a new graph neural network... | arxiv_abstracts | 4 | {
"id": "2005.08485",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
31867,
12686,
3192,
12401,
5544,
33078,
24449,
37796,
42613
],
[
34874,
42613,
1013,
12401,
3192,
24449,
5544,
37796,
3767
],
[
14654,
53400,
3938,
10471,
31867,
44713,
44671,
33866,
50654
]
] |
[
"blind source separation independent component analysis acoustic emission",
"how to locate two simultaneous acoustic emission sources using ICA",
"ICA application for non-destructive testing of aircraft frame structures"
] | [
true,
true,
true
] | Part I describes an intelligent acoustic emission locator, while Part II discusses blind source separation, time delay estimation, and location of two continuous acoustic emission sources. Acoustic emission (AE) analysis is used for characterization and location of developing defects in materials. AE sources often gene... | arxiv_abstracts | 4 | {
"id": "0704.0050",
"categories": [
"cs.AI",
"cs.NE"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
19310,
13549,
12359,
24677,
12950,
41794,
33760,
15412,
44115
],
[
27480,
50448,
45792,
11164,
46008,
40820,
878,
9824,
6380
],
[
23549,
13538,
11319,
25358,
8060,
18604,
58305,
10231,
35077
]
] |
[
"unsupervised voicing classifier zero crossing rate",
"how to track pitch in noisy speech without training data",
"forward backward Kalman filter smoothing pitch contour"
] | [
true,
true,
false
] | The detection of voiced speech, the estimation of the fundamental frequency, and the tracking of pitch values over time are crucial subtasks for a variety of speech processing techniques. Many different algorithms have been developed for each of the three subtasks. We present a new algorithm that integrates the three s... | arxiv_abstracts | 4 | {
"id": "2103.01173",
"categories": [
"cs.LG",
"cs.SD",
"eess.AS"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
53102,
40356,
42668,
1509,
20278,
26294,
30025,
58519,
25316
],
[
41607,
1592,
50133,
46404,
3570,
13331,
27353,
34446,
57855
],
[
7809,
26123,
39533,
8700,
9719,
2242,
49342,
57065,
24569
]
] |
[
"active learning sequence labeling gradient embedding diversity uncertainty",
"how to reduce labeled data in batch active learning for NLP"
] | [
true,
true
] | Recently, several studies have investigated active learning (AL) for natural language processing tasks to alleviate data dependency. However, for query selection, most of these studies mainly rely on uncertainty-based sampling, which generally does not exploit the structural information of the unlabeled data. This lead... | arxiv_abstracts | 4 | {
"id": "2011.13570",
"categories": [
"cs.AI",
"cs.CL",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
9720,
58490,
17057,
45971,
6909,
50783,
24004,
51684,
1044
],
[
45984,
58490,
15648,
21926,
7350,
55028,
45911,
12599,
55605
]
] |
[
"barometric tactile sensors slip detection robotics",
"how to detect slip with low resolution barometric sensors",
"91 percent slip detection accuracy barometric tactile sensors"
] | [
true,
true,
true
] | Despite the utility of tactile information, tactile sensors have yet to be widely deployed in industrial robotics settings. Part of the challenge lies in identifying slip and other key events from the tactile data stream. In this paper, we present a learning-based method to detect slip using barometric tactile sensors.... | arxiv_abstracts | 4 | {
"id": "2103.13460",
"categories": [
"cs.AI",
"cs.RO"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
1699,
35324,
24218,
1014,
52952,
51029,
38092,
3519,
33207
],
[
1699,
41416,
40576,
8060,
2633,
17448,
8985,
30772,
13206
],
[
1699,
24218,
52952,
35324,
1014,
51029,
55812,
41416,
21328
]
] |
[
"unicast multicast qos routing soft constraint logic programming",
"how to model multidimensional qos metrics with c-semirings",
"solving qos routing constraints using soft constraint logic programming"
] | [
true,
true,
true
] | We present a formal model to represent and solve the unicast/multicast routing problem in networks with Quality of Service (QoS) requirements. To attain this, first we translate the network adapting it to a weighted graph (unicast) or and-or graph (multicast), where the weight on a connector corresponds to the multidim... | arxiv_abstracts | 4 | {
"id": "0704.1783",
"categories": [
"cs.AI",
"cs.LO",
"cs.NI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
31762,
16481,
42922,
30348,
18130,
4469,
26364,
11412,
2479
],
[
12775,
47735,
42800,
12526,
36122,
29143,
1325,
40047,
41557
],
[
30348,
6523,
37141,
31762,
7926,
11412,
47046,
7583,
35362
]
] |
[
"deep learning chest x ray covid 19 diagnosis pneumonia classification",
"can you diagnose covid 19 from chest x rays using neural networks",
"accuracy of deep convolutional neural network for covid 19 x ray detection"
] | [
false,
false,
false
] | Testing for COVID-19 has struggled to keep up with demand, with false negative rates projected as high as 30% and significant delays in obtaining results. X-ray machines are widely available and provide images for diagnosis quickly. This paper explores the utility of chest X-ray images in diagnosing COVID-19. The study... | arxiv_abstracts | 3 | {
"id": "2004.02060",
"categories": [
"cs.CV",
"cs.LG",
"eess.IV"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
35551,
47479,
50007,
24788,
3375,
22926,
36264,
26073,
16716
],
[
13047,
39628,
24788,
311,
17087,
51511,
38363,
48738,
38470
],
[
22926,
37694,
38363,
24788,
48738,
33705,
10403,
31977,
4298
]
] |
[
"BP-SPARQL language for querying process graphs",
"how to summarize process graphs and discover concept hierarchies",
"scalable architecture for analyzing business process execution data"
] | [
true,
false,
true
] | In modern enterprises, Business Processes (BPs) are realized over a mix of workflows, IT systems, Web services, and direct collaborations of people. Accordingly, process data (i.e., BP execution data such as logs containing events, interaction messages, and other process artifacts) is scattered across several systems a... | arxiv_abstracts | 4 | {
"id": "2105.10911",
"categories": [
"cs.DB",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
53672,
10934,
36784,
21091,
10783,
13763,
43101,
39046,
16555
],
[
43812,
16555,
54161,
47289,
48296,
57011,
38195,
14736,
3205
],
[
11489,
50916,
55272,
20291,
38195,
41286,
1004,
13234,
49219
]
... |
[
"belief propagation interacting multiple models SLAM",
"how to handle strongly changing agent state dynamics in SLAM"
] | [
true,
true
] | In this paper, we present a Bayesian multipath-based simultaneous localization and mapping (SLAM) algorithm that continuously adapts interacting multiple models (IMM) parameters to describe the mobile agent state dynamics. The time-evolution of the IMM parameters is described by a Markov chain and the parameters are in... | arxiv_abstracts | 4 | {
"id": "2103.12809",
"categories": [
"cs.LG",
"cs.SY",
"eess.SP",
"eess.SY"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
26631,
43330,
19786,
18362,
36131,
24295,
55222,
59314,
36093
],
[
43330,
57426,
19786,
40677,
44801,
36131,
23333,
6371,
5011
]
] |
[
"impedance control vs torque action spaces reinforcement learning manipulation",
"does task-space impedance controller reduce sample complexity in RL"
] | [
true,
true
] | Designing reinforcement learning (RL) problems that can produce delicate and precise manipulation policies requires careful choice of the reward function, state, and action spaces. Much prior work on applying RL to manipulation tasks has defined the action space in terms of direct joint torques or reference positions f... | arxiv_abstracts | 4 | {
"id": "1908.08659",
"categories": [
"cs.LG",
"cs.RO"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
5958,
55993,
23225,
792,
58184,
35504,
26848,
46003,
39972
],
[
55764,
38163,
38051,
9102,
56682,
29060,
40081,
9596,
24706
]
] |
[
"graph cross network multiscale feature learning architecture",
"vertex infomax pooling versus traditional graph pooling methods"
] | [
true,
false
] | We propose a novel graph cross network (GXN) to achieve comprehensive feature learning from multiple scales of a graph. Based on trainable hierarchical representations of a graph, GXN enables the interchange of intermediate features across scales to promote information flow. Two key ingredients of GXN include a novel v... | arxiv_abstracts | 4 | {
"id": "2010.01804",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
15072,
21252,
35022,
20615,
28038,
58458,
2855,
33314,
55609
],
[
14137,
47025,
32474,
43232,
53917,
27202,
23715,
32404,
20513
]
] |
[
"unified benchmark suite for data-driven physical system simulation",
"extensible benchmark problems for machine learning scientific computing time integrators"
] | [
true,
true
] | Simulating physical systems is a core component of scientific computing, encompassing a wide range of physical domains and applications. Recently, there has been a surge in data-driven methods to complement traditional numerical simulations, motivated by the opportunity to reduce computational costs and/or learn new ph... | arxiv_abstracts | 4 | {
"id": "2108.07799",
"categories": [
"cs.LG",
"physics.comp-ph"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
9955,
52593,
10175,
40092,
8140,
8884,
41746,
41848,
17921
],
[
35016,
25776,
9722,
35451,
17970,
31152,
4213,
48699,
26896
]
] |
[
"kernel mahalanobis distance multiclass uncertainty quantification",
"does kernel mahalanobis distance handle out-of-distribution audio samples"
] | [
true,
true
] | In this manuscript, we propose a multiclass data description model based on kernel Mahalanobis distance (MDD-KM) with self-adapting hyperparameter setting. MDD-KM provides uncertainty quantification and can be deployed to build classification systems for the realistic scenario where out-of-distribution (OOD) samples ar... | arxiv_abstracts | 4 | {
"id": "2108.12857",
"categories": [
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
38531,
42388,
30781,
57802,
52677,
34906,
2325,
9896,
57444
],
[
13305,
4893,
33760,
37730,
52917,
42494,
25133,
3963,
38531
]
] |
[
"pointnets 2d object detection radar data amodal bounding box",
"how to combine classification and bounding box estimation with radar pointnets",
"automated driving 2d object detection sparse radar pointnet regression"
] | [
true,
true,
true
] | For many automated driving functions, a highly accurate perception of the vehicle environment is a crucial prerequisite. Modern high-resolution radar sensors generate multiple radar targets per object, which makes these sensors particularly suitable for the 2D object detection task. This work presents an approach to de... | arxiv_abstracts | 4 | {
"id": "1904.08414",
"categories": [
"cs.CV",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
48016,
17848,
21957,
17814,
15901,
46813,
58017,
32349,
23314
],
[
48016,
36812,
31264,
58771,
58295,
42095,
35588,
6882,
44644
],
[
15901,
38231,
48016,
13486,
20338,
46813,
57120,
57781,
19276
]... |
[
"LoAdaBoost federated learning intensive care data"
] | [
true
] | Intensive care data are valuable for improving health care, policy making, and other purposes. Vast amounts of such data are stored in different locations, on many different devices, and in different data silos. Sharing data among different sources is a significant challenge due to regulatory, operational, and security... | arxiv_abstracts | 4 | {
"id": "1811.12629",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
37902,
56886,
51757,
21877,
22485,
58642,
557,
36854,
21879
]
] |
[
"DLV DB recursive query performance comparison commercial databases",
"how to handle massive distributed data reasoning with external DBMSs",
"recursive query efficiency logic programming versus relational databases"
] | [
true,
true,
true
] | This paper addresses the challenge of reasoning over massive amounts of data, which may be distributed. Current approaches face three main limitations: (i) limited data capacity due to reliance on main-memory reasoning; (ii) non-trivial or impossible interaction with external, independent database management systems (D... | arxiv_abstracts | 4 | {
"id": "0704.3157",
"categories": [
"cs.AI",
"cs.DB"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
393,
47372,
59240,
36645,
16098,
13435,
26289,
12260,
27286
],
[
20952,
18894,
15291,
16902,
46514,
37213,
43541,
41734,
47372
],
[
36187,
3147,
19111,
24323,
13435,
56466,
40987,
6128,
50559
]
] |
[
"intelligent input methods software engineering hci nlp",
"how to use input methods for non east asian languages",
"design philosophy text service platform intelligent input methods"
] | [
true,
true,
true
] | Intelligent Input Methods (IM) are essential for making text entries in many East Asian scripts, but their application to other languages has not been fully explored. This paper discusses how such tools can contribute to the development of computer processing of other oriental languages. We propose a design philosophy ... | arxiv_abstracts | 3 | {
"id": "0704.3665",
"categories": [
"cs.CL",
"cs.HC"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
43750,
51352,
12569,
37824,
27188,
12751,
48265,
9209,
5560
],
[
46393,
13465,
38185,
17869,
8528,
30081,
29812,
54052,
27522
],
[
35584,
53740,
19563,
3923,
4674,
43753,
48265,
371,
13465
]
] |
[
"bayesian rough set mcmc hiv risk prediction",
"how to train rough set models with fewer rules using mcmc",
"metropolis algorithm acceptance criteria rough set granule space"
] | [
true,
true,
false
] | This paper proposes an approach to training rough set models using a Bayesian framework trained via the Markov Chain Monte Carlo (MCMC) method. Prior probabilities are constructed from the prior knowledge that good rough set models have fewer rules. Markov Chain Monte Carlo sampling is conducted through sampling in the... | arxiv_abstracts | 3 | {
"id": "0704.3433",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
34681,
49019,
15596,
20633,
27811,
53508,
37783,
46906,
21421
],
[
14681,
2369,
4897,
47712,
16613,
20662,
56231,
46414,
33635
],
[
10062,
37703,
1479,
2620,
55464,
39214,
18052,
1280,
48639
]
] |
[
"counterfactual explanation brain activity classifiers generative adversarial network",
"how to generate synthetic fMRI data to explain neural network decisions",
"image-to-image transfer counterfactual activation generator behavioral tasks"
] | [
true,
true,
true
] | Deep neural networks (DNNs) can accurately decode task-related information from brain activations. However, because of the nonlinearity of the DNN, the decisions made by DNNs are hardly interpretable. One of the promising approaches for explaining such a black-box system is counterfactual explanation. In this framework... | arxiv_abstracts | 4 | {
"id": "2110.14927",
"categories": [
"cs.CV",
"cs.LG",
"q-bio.NC"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
24787,
11145,
7078,
6519,
46427,
38476,
41408,
14770,
18261
],
[
37981,
21295,
262,
23174,
53086,
36683,
36777,
13284,
56570
],
[
18543,
43357,
59111,
27069,
40167,
24787,
6537,
2075,
37224
]
] |
[
"short term traffic prediction deep neural networks survey",
"how to choose datasets for short term traffic prediction benchmarks"
] | [
false,
false
] | In modern transportation systems, an enormous amount of traffic data is generated every day. This has led to rapid progress in short-term traffic prediction (STTP), in which deep learning methods have recently been applied. In traffic networks with complex spatiotemporal relationships, deep neural networks (DNNs) often... | arxiv_abstracts | 4 | {
"id": "2009.00712",
"categories": [
"cs.AI",
"cs.LG",
"eess.SP"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
8924,
36035,
40323,
32324,
24731,
31639,
38168,
34358,
15699
],
[
52692,
20639,
40323,
18858,
14493,
36035,
24731,
19448,
8495
]
] |
[
"efficient algorithm minimum information partition integrated information theory",
"does greedy optimization work for non submodular integrated information measures"
] | [
true,
false
] | The ability to integrate information in the brain is considered an essential property for cognition and consciousness. Integrated Information Theory (IIT) hypothesizes that the amount of integrated information ($\Phi$) in the brain is related to the level of consciousness. IIT proposes that to quantify information inte... | arxiv_abstracts | 4 | {
"id": "1712.06745",
"categories": [
"cs.IT",
"math.IT",
"q-bio.NC",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
23082,
22383,
5635,
42079,
6788,
624,
36549,
736,
7985
],
[
12797,
56968,
30141,
44472,
9372,
19171,
1279,
52876,
33568
]
] |
[
"conditional generative adversarial network adversarial example generation"
] | [
false
] | Recently, deep neural networks have made significant progress and found successful application in various fields, but they are found vulnerable to attack instances, e.g., adversarial examples. State-of-the-art attack methods can generate attack images by adding small perturbations to the source image. These attack imag... | arxiv_abstracts | 4 | {
"id": "1903.07282",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
53413,
44775,
22853,
55423,
48165,
48621,
29504,
56579,
23772
]
] |
[
"non-autonomous neural ode time varying weights",
"how to control smoothness of weight trajectories in neural odes",
"neural ode performance vs resnet and cnn on video prediction"
] | [
true,
false,
true
] | Neural Ordinary Differential Equations (ODEs) are elegant reinterpretations of deep networks where continuous time can replace the discrete notion of depth, ODE solvers perform forward propagation, and the adjoint method enables efficient, constant memory backpropagation. Neural ODEs are universal approximators only wh... | arxiv_abstracts | 4 | {
"id": "2005.01906",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
35153,
41030,
15806,
54950,
6513,
15222,
18825,
36742,
47570
],
[
35153,
41030,
36742,
11253,
6513,
15222,
15806,
31472,
36225
],
[
28593,
53691,
48668,
31377,
30460,
7173,
1131,
33150,
4246
]
] |
[
"document level neural machine translation survey methods evaluation",
"what are the current architectures and decoding strategies for document level translation"
] | [
true,
false
] | Machine translation (MT) is an important task in natural language processing (NLP) as it automates the translation process and reduces the reliance on human translators. With the resurgence of neural networks, the translation quality surpasses that of the translations obtained using statistical techniques for most lang... | arxiv_abstracts | 4 | {
"id": "1912.08494",
"categories": [
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
44479,
17382,
29163,
14324,
25489,
50277,
15862,
8222,
12483
],
[
33673,
17382,
15862,
29163,
8222,
48422,
29018,
52892,
32921
]
] |
[
"wasserstein distance guided policy optimization algorithms",
"how to steer reinforcement learning agents with behavior scores",
"behavior guided policy gradient and evolution strategies performance"
] | [
true,
false,
false
] | We introduce a new approach for comparing reinforcement learning policies using Wasserstein distances in a newly defined latent behavioral space. By utilizing the dual formulation of the Wasserstein distance, we can learn score functions over policy behaviors that guide policy optimization towards or away from desired ... | arxiv_abstracts | 5 | {
"id": "1906.04349",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
13894,
21304,
10502,
31420,
21496,
21564,
54525,
51447,
448
],
[
1906,
46547,
49435,
38055,
49560,
56569,
42148,
17044,
2514
],
[
10317,
58275,
47039,
6392,
41266,
35192,
4982,
46696,
56133
]
] |
[
"american sign language identification using hand track points",
"best pre-processing scheme for sign language gestures using neural networks"
] | [
true,
true
] | Sign Language helps people with speaking and hearing disabilities communicate with others efficiently. Sign Language identification is a challenging area in the field of computer vision, and recent developments have achieved near-perfect results for the task, though some challenges remain. In this paper, we propose a n... | arxiv_abstracts | 4 | {
"id": "2010.10590",
"categories": [
"cs.AI",
"cs.CV",
"cs.HC",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
32588,
51890,
47367,
27407,
29910,
33142,
41053,
18870,
54231
],
[
27407,
33142,
8173,
41053,
260,
47367,
55444,
51890,
58962
]
] |
[
"contrastive learning acronym disambiguation scientific documents",
"how to improve masked language model representations for ambiguous acronyms",
"phrase-level contrastive pre-training for acronym meaning disambiguation"
] | [
true,
true,
true
] | Acronym disambiguation involves identifying the correct meaning of an ambiguous acronym from a dictionary within a given sentence, a critical task for scientific document understanding. Recent approaches have attempted to solve this by fine-tuning pre-trained masked language models (MLMs) to improve acronym representat... | arxiv_abstracts | 4 | {
"id": "2111.14306",
"categories": [
"cs.AI",
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
14291,
2513,
17091,
4140,
34284,
1903,
41528,
20730,
34339
],
[
41528,
14291,
4140,
17091,
1903,
2513,
43725,
16493,
34284
],
[
14291,
2513,
17091,
4140,
8028,
41528,
15768,
38031,
22420
]
] |
[
"categorical model of resources in game semantics",
"how to extend bracketing to multi-bracketing in arena games"
] | [
true,
false
] | The description of resources in game semantics has never achieved the simplicity and precision of linear logic, due to a misleading conception: the belief that linear logic is more primitive than game semantics. We advocate instead the contrary: that game semantics is conceptually more primitive than linear logic. Star... | arxiv_abstracts | 4 | {
"id": "0705.0462",
"categories": [
"cs.CL",
"math.CT"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
12266,
49333,
37489,
13327,
11035,
45829,
3330,
32660,
39534
],
[
55732,
8868,
58364,
53380,
40753,
47451,
9014,
19824,
3063
]
] |
[
"clustering frequent subgraph patterns using lattice information",
"Lattice2SAR framework for analyzing molecular fragment patterns"
] | [
true,
true
] | Mining frequent subgraphs involves searching for connected subgraphs contained in many graphs within a given set, where each graph can be viewed as a transaction. This work discusses techniques used in the Lattice2SAR framework for mining and analyzing frequent subgraph data and their corresponding lattice information.... | arxiv_abstracts | 4 | {
"id": "0705.0593",
"categories": [
"cs.AI",
"cs.DS"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
22576,
40313,
22504,
4199,
10676,
7816,
22391,
45604,
37888
],
[
13910,
54259,
20826,
41235,
7816,
21387,
48409,
16683,
53331
]
] |
[
"continuous time multi agent pathfinding optimal solutions"
] | [
false
] | Multi-Agent Pathfinding (MAPF) is the problem of finding paths for multiple agents such that every agent reaches its goal and the agents do not collide. Most prior work on MAPF was on grids, assumed agents' actions have uniform duration, and that time is discretized into timesteps. We propose a MAPF algorithm that does... | arxiv_abstracts | 4 | {
"id": "1901.05506",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
39966,
19020,
10586,
39415,
11001,
22171,
55989,
40447,
39567
]
] |
[
"how to extract semantic relations between named entities in claims"
] | [
false
] | The enormous amount of discourse taking place online poses challenges to the functioning of a civil and informed public sphere. Efforts to standardize online discourse data, such as ClaimReview, are making available a wealth of new data about potentially inaccurate claims, reviewed by third-party fact-checkers. These d... | arxiv_abstracts | 4 | {
"id": "2102.11105",
"categories": [
"cs.CL",
"cs.SI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
14559,
13760,
23889,
30765,
56424,
27333,
17494,
44890,
50759
]
] |
[
"optimal solutions preservation order reflecting semiring homomorphism",
"how to solve concrete soft constraint problems via abstract homomorphism",
"relationship between semiring homomorphism and optimal solution values"
] | [
true,
true,
true
] | Semiring-based constraint satisfaction problems (semiring CSPs), proposed by Bistarelli, Montanari, and Rossi, provide a general framework for soft constraints. This paper proposes an abstraction scheme for soft constraints utilizing semiring homomorphism. The approach involves working in an abstract problem to find op... | arxiv_abstracts | 4 | {
"id": "0705.0734",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
36111,
43616,
34245,
15498,
23344,
11024,
18554,
35323,
20411
],
[
16481,
3686,
37141,
32019,
26364,
7733,
7583,
43616,
10569
],
[
36111,
43263,
16421,
29715,
10996,
37180,
23344,
8355,
6615
]
] |
[
"artificial neural networks vs support vector machines water demand forecasting",
"which model predicts water demand better ann or svm",
"generalization ability comparison neural networks support vector machines water"
] | [
true,
true,
true
] | Water plays a pivotal role in sustaining human, animal, and plant life. Consequently, water supply entities must provide clean and safe water at the rate required by consumers. Implementing mechanisms to predict both short-term and long-term water demands is therefore necessary. Computational intelligence techniques ha... | arxiv_abstracts | 3 | {
"id": "0705.0969",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
21025,
30286,
3256,
22582,
43557,
54031,
26021,
20602,
52934
],
[
30286,
21025,
43557,
27182,
15778,
22582,
54031,
13519,
54649
],
[
13559,
20230,
380,
52934,
44608,
26021,
41846,
44923,
18854
]
] |
[
"semi-supervised learning spanish poetry psychological categories",
"how to extract affective features from spanish sonnets using transformers",
"performance gain lexico-semantic features transformer sentence embeddings poetry"
] | [
true,
true,
true
] | Text classification tasks have improved substantially in recent years through the use of transformers. However, the majority of research focuses on prose texts, with poetry receiving less attention, especially for the Spanish language. In this paper, we propose a semi-supervised learning approach for inferring 21 psych... | arxiv_abstracts | 4 | {
"id": "2109.04152",
"categories": [
"cs.AI",
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
26172,
53100,
52209,
25161,
41919,
11540,
57499,
48983,
14260
],
[
41919,
47794,
29708,
30483,
12110,
41429,
49682,
24249,
27301
],
[
53100,
10419,
14277,
54717,
21339,
43260,
849,
49793,
58340
]
... |
[
"detect frequent itemsets with balanced intervals between occurrences",
"how to prune patterns using standard deviation and average for balanced intervals",
"mining patterns that repeat every three transactions with balanced gaps"
] | [
true,
true,
false
] | In many applications, it is useful to identify patterns that occur with a balanced interval, such as a specific combination of phone numbers being called almost every Friday or a group of products selling frequently on Tuesday and Thursday. In previous work, we proposed a new measure of support where we count the numbe... | arxiv_abstracts | 4 | {
"id": "0705.1110",
"categories": [
"cs.AI",
"cs.DB"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
34565,
4901,
5315,
1096,
45604,
35422,
20603,
4199,
47750
],
[
28793,
46663,
34943,
14670,
20071,
24043,
34465,
39412,
31170
],
[
1096,
19697,
20603,
12788,
4199,
28847,
4901,
18356,
5315
]
] |
[
"meta-embeddings graph autoencoders medical concept representations",
"joint reconstruction heterogeneous data types medical informatics"
] | [
true,
false
] | Distributed representations of medical concepts have been used to support downstream clinical tasks recently. Electronic Health Records (EHR) capture different aspects of patients' hospital encounters and serve as a rich source for augmenting clinical decision making by learning robust medical concept embeddings. Howev... | arxiv_abstracts | 4 | {
"id": "1912.03366",
"categories": [
"cs.CL",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
7643,
9579,
21791,
41594,
2333,
29074,
33436,
6463,
24663
],
[
26662,
13950,
18840,
40170,
52081,
30609,
37898,
30521,
6082
]
] |
[
"how to learn optimal hypothesis privately when fat shattering dimension diverges",
"novel filtering procedure for stable nonparametric function classes"
] | [
true,
false
] | Given a real-valued hypothesis class H, we investigate under what conditions there is a differentially private algorithm which learns an optimal hypothesis from H given i.i.d. data. Inspired by recent results for the related setting of binary classification, where it was shown that online learnability of a binary class... | arxiv_abstracts | 5 | {
"id": "2111.12786",
"categories": [
"cs.CR",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
2653,
7107,
8513,
972,
12182,
12180,
60,
10842,
412
],
[
16809,
47826,
46326,
16559,
47859,
24974,
18847,
4977,
8700
]
] |
[
"derivation of inverse document frequency using Robertson Sparck Jones model",
"why is IDF estimation considered intractable in probabilistic models",
"simple assumption for inverse document frequency effectiveness"
] | [
true,
false,
true
] | There have been a number of prior attempts to theoretically justify the effectiveness of the inverse document frequency (IDF). Those that take as their starting point Robertson and Sparck Jones's probabilistic model are based on strong or complex assumptions. We show that a more intuitively plausible assumption suffice... | arxiv_abstracts | 4 | {
"id": "0705.1161",
"categories": [
"cs.CL",
"cs.IR"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
24578,
36757,
40672,
32164,
54111,
1322,
50092,
23714,
24675
],
[
12667,
1735,
43537,
52733,
59081,
34864,
46589,
17731,
54934
],
[
24578,
1322,
23714,
40672,
50092,
36757,
32164,
24675,
25287
]
] |
[
"compact embedding for facial expression similarity human preferences",
"how to learn continuous facial expression embeddings from pairwise similarity data",
"16 dimensional embedding for facial expression retrieval and summarization"
] | [
true,
true,
true
] | Most existing work on automatic facial expression analysis focuses on discrete emotion recognition or facial action unit detection. However, facial expressions do not always fall neatly into pre-defined semantic categories, and similarity measured in the action unit space need not correspond to human perception. Differ... | arxiv_abstracts | 4 | {
"id": "1811.11283",
"categories": [
"cs.AI",
"cs.CV"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
8829,
57064,
27098,
21032,
3402,
19689,
18169,
13547,
34346
],
[
34346,
12999,
54823,
35407,
8829,
31109,
27098,
50147,
47730
],
[
34346,
8829,
56097,
57064,
18169,
47504,
54186,
50147,
27098
]
] |
[
"symbolic controllers evolutionary robotics elementary behaviors",
"sensitivity and robustness of symbolic controller generalization"
] | [
true,
true
] | The idea of symbolic controllers tries to bridge the gap between the top-down manual design of the controller architecture, as advocated in Brooks' subsumption architecture, and the bottom-up designer-free approach that is now standard within the Evolutionary Robotics community. The designer provides a set of elementar... | arxiv_abstracts | 3 | {
"id": "0705.1244",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
28752,
40814,
29552,
1785,
17426,
50893,
18907,
26987,
8850
],
[
40918,
24152,
9635,
50940,
34399,
9234,
2321,
17426,
3114
]
] |
[
"neural network developmental design grid based growth model",
"how to evolve self healing multi cellular organisms with neat",
"emergent stabilization criteria for continuous developmental systems"
] | [
true,
true,
false
] | This paper introduces a continuous model for Multi-cellular Developmental Design. Cells are fixed on a 2D grid and exchange chemicals with neighbors during growth. The quantity of chemicals produced by a cell and its differentiation value in the phenotype are controlled by a Neural Network (the genotype), which takes a... | arxiv_abstracts | 4 | {
"id": "0705.1309",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
45276,
43220,
28415,
7396,
43587,
3351,
1627,
15340,
2239
],
[
36081,
46991,
9775,
52738,
10763,
28752,
20366,
46424,
40449
],
[
58934,
679,
27603,
31484,
1058,
22646,
26677,
10810,
40918
]
] |
[
"sub-band speaker identification linear merging performance",
"how to improve live speaker identification accuracy with noise",
"support vector machines gaussian mixture models sub-band merging"
] | [
true,
true,
true
] | Speaker identification is a powerful, non-invasive, and inexpensive biometric technique. However, recognition accuracy deteriorates when noise levels affect a specific band of frequency. This paper presents a sub-band based speaker identification method intended to improve live testing performance. Each frequency sub-b... | arxiv_abstracts | 4 | {
"id": "0705.1585",
"categories": [
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
12128,
40159,
28909,
20185,
39397,
26083,
24710,
6077,
31817
],
[
18799,
40686,
42500,
43319,
40570,
26083,
12128,
43104,
56085
],
[
22700,
3864,
22239,
45594,
16538,
8232,
36059,
48568,
49968
]
] |
[
"determine concentration graph perfect Markov distribution conditioning",
"how many variables needed to identify perfect Markov graph",
"relationship minimal separators concentration graph conditioning"
] | [
true,
true,
true
] | A concentration graph associated with a random vector is an undirected graph where each vertex corresponds to one random variable in the vector. The absence of an edge between any pair of vertices (or variables) is equivalent to full conditional independence between these two variables given all the other variables. In... | arxiv_abstracts | 5 | {
"id": "0705.1613",
"categories": [
"math.ST",
"stat.ML",
"stat.TH"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
18788,
33421,
8651,
15142,
8789,
26189,
21437,
40352,
5896
],
[
6024,
9087,
41953,
46617,
58690,
49064,
40272,
9529,
21437
],
[
40444,
26252,
31098,
45627,
7953,
40077,
46162,
50827,
58684
]
] |
[
"data aware weighted sampling covariance estimator",
"unbiased covariance estimation using weighted sampling"
] | [
true,
true
] | Estimating covariance matrices from massive, high-dimensional, and distributed data is critical for various real-world applications. This paper proposes a data-aware weighted sampling-based covariance matrix estimator, named DACE, which provides unbiased estimation and achieves higher accuracy under the same compressio... | arxiv_abstracts | 4 | {
"id": "2010.04966",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
19080,
51708,
30870,
4055,
16015,
1907,
59183,
8989,
56747
],
[
51708,
30870,
16015,
59183,
56747,
19080,
1907,
20091,
52760
]
] |
[
"multi-agent reinforcement learning informational propagation model",
"how does learning slow down in multi-agent systems over time",
"mathematical model for information evolution across multiple agents"
] | [
true,
true,
true
] | We introduce a new mathematical model of multi-agent reinforcement learning, the Multi-Agent Informational Learning Processor (MAILP) model. The model is based on the notion that agents have policies for a certain amount of information, modeling how this information iteratively evolves and propagates through many agent... | arxiv_abstracts | 4 | {
"id": "2006.06870",
"categories": [
"cs.AI",
"cs.LG",
"cs.MA"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
32007,
38465,
5147,
10752,
26461,
27894,
41962,
12591,
20049
],
[
47228,
7664,
47937,
55694,
44796,
16136,
44356,
38480,
45301
],
[
12032,
9331,
51697,
38465,
49807,
44952,
37157,
59125,
34853
]
] |
[
"perturbational complexity distribution mismatch reinforcement learning",
"how does perturbational complexity measure rl problem difficulty"
] | [
true,
true
] | Most existing theoretical analysis of reinforcement learning (RL) is limited to the tabular setting or linear models due to the difficulty in dealing with function approximation in high dimensional space with an uncertain environment. This work offers a fresh perspective into this challenge by analyzing RL in a general... | arxiv_abstracts | 4 | {
"id": "2111.03469",
"categories": [
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
58987,
18874,
48886,
13282,
55764,
37547,
36253,
9488,
43581
],
[
3916,
46298,
58431,
11212,
26076,
20810,
55764,
55220,
21169
]
] |
[
"random fourier features asymptotics large n p N",
"why does random fourier feature regression deviate from gaussian kernel limits",
"double descent phase transition random fourier features test error"
] | [
true,
false,
true
] | This article characterizes the exact asymptotics of random Fourier feature (RFF) regression in the realistic setting where the number of data samples $n$, their dimension $p$, and the dimension of feature space $N$ are all large and comparable. In this regime, the random RFF Gram matrix no longer converges to the well-... | arxiv_abstracts | 5 | {
"id": "2006.05013",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
48394,
23795,
3515,
28037,
31687,
51776,
15504,
18955,
13796
],
[
48394,
28037,
3515,
23795,
15504,
31687,
47826,
18955,
5139
],
[
36960,
26222,
12205,
13516,
48137,
6925,
42458,
15604,
16293
]
] |
[
"neuro fuzzy model fault classification cylindrical shells",
"how to classify faults in cylindrical shells better than mlp",
"pseudomodal energies feature extraction vibration signals accuracy"
] | [
true,
true,
true
] | This paper presents a fault classification method utilizing a Takagi-Sugeno neuro-fuzzy model and pseudomodal energies calculated from vibration signals of cylindrical shells. The calculation of pseudomodal energies has previously been established as an accurate method for extracting features from vibration signals for... | arxiv_abstracts | 4 | {
"id": "0705.2236",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
3327,
23549,
1573,
40160,
23615,
18140,
42019,
16851,
12269
],
[
23549,
18140,
43428,
19526,
568,
56533,
21428,
49828,
11121
],
[
33550,
55258,
4812,
54387,
7073,
10468,
50295,
29235,
35038
]
] |
[
"self teaching paradigm machine reading comprehension weakly labeled data",
"how to use examqa dataset for machine reading comprehension tasks",
"improving extractive and multiple choice mrc with web search snippets"
] | [
true,
true,
false
] | Despite recent research, the utility of subject-area question-answering data for machine reading comprehension (MRC) tasks remains unclear. This paper investigates this question by collecting a large-scale multi-subject multiple-choice dataset, ExamQA. Incomplete and noisy snippets from web search engines are used as r... | arxiv_abstracts | 5 | {
"id": "2102.01226",
"categories": [
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
34337,
20565,
46875,
43041,
24989,
9544,
55395,
21037,
30606
],
[
3350,
56283,
16360,
20565,
3647,
19168,
9914,
5555,
24989
],
[
28353,
43242,
18521,
53170,
42072,
6626,
20385,
12260,
2737
]
] |
[
"visual context improves next word prediction perplexity",
"does training with images help language models generalize"
] | [
true,
false
] | We examine the benefits of visual context in training neural language models to perform next-word prediction. A multi-modal neural architecture is introduced that outperforms its equivalent trained on language alone with a 2% decrease in perplexity, even when no visual context is available at test. Fine-tuning the embe... | arxiv_abstracts | 4 | {
"id": "1805.11546",
"categories": [
"cs.AI",
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
2436,
5177,
40258,
30621,
55900,
21313,
51169,
36651,
54200
],
[
28026,
1236,
6937,
25443,
8040,
57681,
36343,
55603,
50945
]
] |
[
"adaptive bin splitting method nonconvex bandit regret",
"how to achieve locally minimax optimal regret in nonconvex bandits",
"upper bound cumulative regret simple bin splitting nonconvex"
] | [
true,
true,
true
] | We analyze continuous armed bandit problems for nonconvex cost functions under certain smoothness and sublevel set assumptions. First, we derive an upper bound on the expected cumulative regret of a simple bin splitting method. Then, we propose an adaptive bin splitting method that significantly improves performance. F... | arxiv_abstracts | 4 | {
"id": "2103.16082",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
51809,
57009,
51358,
26376,
16402,
18297,
8453,
31729,
54278
],
[
57009,
45685,
22298,
42688,
58142,
31729,
18297,
12491,
30932
],
[
22298,
58142,
39871,
55046,
35637,
58544,
44186,
39395,
54278
]... |
[
"differentially private sparse learning knowledge transfer",
"how to improve utility in private sparse regression",
"linear convergence rate differentially private student estimator"
] | [
true,
true,
false
] | We study the problem of estimating high-dimensional models with underlying sparse structures while preserving the privacy of each training example. We develop a differentially private high-dimensional sparse learning framework using the idea of knowledge transfer. More specifically, we propose to distill the knowledge ... | arxiv_abstracts | 4 | {
"id": "1909.06322",
"categories": [
"cs.CR",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
50096,
4142,
41197,
7710,
51530,
14164,
7137,
42443,
54567
],
[
4142,
77,
13194,
32975,
17474,
5820,
33685,
57393,
4974
],
[
378,
15738,
10390,
28527,
23270,
20057,
5955,
1797,
11775
]
] |
[
"bayesian stochastic game model football tactics optimization",
"how to predict match outcomes and optimize team formation in soccer",
"increase team winning probability using game theory based tactics"
] | [
true,
true,
true
] | In this paper, we present a novel approach to optimise tactical and strategic decision making in football (soccer). We model the game of football as a multi-stage game composed of a Bayesian game to model pre-match decisions and a stochastic game to model in-match state transitions and decisions. Using this formulation... | arxiv_abstracts | 4 | {
"id": "2003.10294",
"categories": [
"cs.AI",
"cs.GT",
"cs.MA"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
20885,
25077,
20972,
11986,
49020,
52042,
51686,
42037,
175
],
[
20885,
37175,
12349,
11986,
40247,
175,
29225,
40844,
20686
],
[
32342,
16916,
43269,
42639,
46469,
31031,
16356,
10146,
1436
]
] |
[
"Bayesian model averaging nuclear mass models drip lines",
"how to predict proton neutron separation energies far from stability",
"probability of existence for nuclei up to Z equals 119"
] | [
true,
true,
true
] | The chart of the nuclides is limited by particle drip lines beyond which nuclear stability to proton or neutron emission is lost. Predicting the range of particle-bound isotopes poses an appreciable challenge for nuclear theory as it involves extreme extrapolations of nuclear masses beyond the regions where experimenta... | arxiv_abstracts | 5 | {
"id": "2001.05924",
"categories": [
"nucl-ex",
"nucl-th",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
18060,
400,
511,
19885,
45780,
8941,
47501,
25754,
15596
],
[
400,
40867,
6821,
58051,
19857,
34783,
49199,
17369,
53488
],
[
18060,
400,
5201,
32816,
12612,
12486,
23021,
23940,
40867
]
] |
[
"symbolic interval propagation adversarial attacks neural networks",
"improving gradient-based attacks using symbolic interval propagation"
] | [
true,
true
] | Recent breakthroughs in defenses against adversarial examples, such as adversarial training, have made neural networks robust against various classes of attackers, including first-order gradient-based attacks. However, it remains an open question whether adversarially trained networks are truly robust under unknown att... | arxiv_abstracts | 5 | {
"id": "1906.02282",
"categories": [
"cs.CR",
"cs.LG",
"cs.LO",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
56579,
51116,
34160,
57245,
9919,
16119,
54775,
44884,
15369
],
[
44994,
32967,
16119,
3751,
48882,
9965,
15956,
34149,
37097
]
] |
[
"decision tree soft labels knowledge distillation",
"how to compress decision trees without backpropagation",
"rectified decision trees interpretability vs compression"
] | [
false,
false,
true
] | Obtaining a model with both good interpretability and performance is a longstanding research challenge. This paper proposes Rectified Decision Trees (ReDT), a knowledge distillation-based approach that rectifies decision trees to achieve high interpretability, small model size, and empirical soundness. We extend the im... | arxiv_abstracts | 4 | {
"id": "1903.05965",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
52320,
21945,
17665,
53864,
23796,
2906,
8537,
34986,
28453
],
[
50465,
36999,
25275,
58957,
35005,
19279,
19599,
11462,
27365
],
[
52320,
19599,
19048,
52115,
39165,
58957,
46061,
55920,
50465
]
... |
[
"multi-agent football game reinforcement learning ai system",
"how to train multi-agent football agents from single agent data",
"offline algorithms for distributed learning in google research football"
] | [
true,
true,
false
] | Deep reinforcement learning (DRL) has achieved super-human performance on complex video games like StarCraft II and Dota II. However, current DRL systems still struggle with multi-agent coordination, sparse rewards, and stochastic environments. To address these challenges, we use the Google Research Football (GRF) vide... | arxiv_abstracts | 4 | {
"id": "2110.04507",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
18221,
41397,
1673,
46075,
36625,
83,
52133,
41308,
58437
],
[
18221,
4305,
2364,
13523,
46075,
59055,
44796,
52133,
35025
],
[
9267,
8955,
25548,
886,
23036,
39784,
2371,
50521,
46164
]
] |
[
"Q learning neural network convergence ordinary differential equation",
"does Q learning converge to Bellman equation optimal control",
"limiting behavior stochastic gradient descent Xavier initialization single layer"
] | [
true,
false,
false
] | We prove that a single-layer neural network trained with the Q-learning algorithm converges in distribution to a random ordinary differential equation as the size of the model and the number of training steps become large. Analysis of the limit differential equation shows that it has a unique stationary solution which ... | arxiv_abstracts | 4 | {
"id": "1911.07304",
"categories": [
"cs.LG",
"math.PR",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
46554,
36742,
16959,
41030,
40184,
33162,
46843,
55984,
30
],
[
48322,
34402,
30930,
8687,
49128,
33917,
39205,
44093,
31203
],
[
6485,
6794,
35084,
28271,
53874,
29999,
55517,
30398,
31642
]
] |
[
"optimized gene transmitting crossover complexity boolean linear programming",
"is optimized gene transmitting crossover np hard for set packing",
"complexity of best offspring crossover 2-colorable hypergraph independent set"
] | [
true,
true,
true
] | We consider the computational complexity of producing the best possible offspring in a crossover, given two solutions of the parents. The crossover operators are studied on the class of Boolean linear programming problems, where the Boolean vector of variables is used as the solution representation. By means of efficie... | arxiv_abstracts | 4 | {
"id": "0705.3766",
"categories": [
"cs.AI",
"cs.NE"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
51344,
9053,
29271,
52723,
26109,
33399,
57247,
37966,
3900
],
[
51344,
58755,
26109,
29271,
27094,
9053,
52723,
28375,
43943
],
[
44179,
19848,
1565,
17491,
11497,
32000,
36187,
27195,
35487
]
] |
[
"cross-view training vs bert sequence tagging",
"does cross-view training match bert performance on tagging"
] | [
true,
true
] | Leveraging large amounts of unlabeled data using Transformer-like architectures, like BERT, has gained popularity in recent times owing to their effectiveness in learning general representations that can then be further fine-tuned for downstream tasks to much success. However, training these models can be costly both f... | arxiv_abstracts | 4 | {
"id": "2010.14042",
"categories": [
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
5109,
20868,
21664,
47196,
22995,
26996,
17263,
5784,
25762
],
[
36525,
15382,
24457,
51558,
54094,
22995,
46558,
41894,
33791
]
] |
[
"state-space model auditory attention decoding EEG",
"reduce processing delay auditory attention decoding small window"
] | [
true,
true
] | Identifying the target speaker in hearing aid applications is crucial to improve speech understanding. Recent advances in electroencephalography (EEG) have shown that it is possible to identify the target speaker from single-trial EEG recordings using auditory attention decoding (AAD) methods. AAD methods reconstruct t... | arxiv_abstracts | 4 | {
"id": "2004.00910",
"categories": [
"cs.LG",
"cs.SD",
"eess.AS",
"eess.SP"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
24664,
32979,
7519,
25019,
18972,
6275,
59207,
24653,
51036
],
[
32979,
18972,
35235,
32756,
16291,
59207,
18696,
7519,
9306
]
] |
[
"loop corrected belief propagation continuous variable gaussian model",
"how to get covariances using belief propagation on cavity graphs",
"relationship between loop correction and expectation propagation for nonlinear models"
] | [
true,
true,
true
] | In this paper, we derive the equations for Loop Corrected Belief Propagation on a continuous variable Gaussian model. Leveraging the exactness of belief propagation averages for Gaussian models, we identify an alternative method for obtaining covariances based on belief propagation on cavity graphs. Furthermore, we dis... | arxiv_abstracts | 4 | {
"id": "0705.4566",
"categories": [
"cs.AI",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
6470,
6207,
36563,
4819,
969,
6574,
24066,
4490,
15751
],
[
41076,
5709,
41424,
12706,
46423,
58086,
55763,
36563,
39731
],
[
37716,
36563,
4819,
4678,
46653,
4490,
969,
12753,
22680
]
] |
[
"chi square transformation cluster crosstable matching algorithm",
"probabilistic component for breaking ties in cluster matching"
] | [
true,
true
] | Cluster matching by permuting cluster labels is important in many clustering contexts such as cluster validation and cluster ensemble techniques. The classic approach is to minimize the Euclidean distance between two cluster solutions, which induces inappropriate stability in certain settings. Therefore, we present the... | arxiv_abstracts | 4 | {
"id": "0705.4302",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
55899,
271,
42729,
18447,
55825,
1545,
25922,
36438,
57992
],
[
10754,
6584,
13099,
440,
35696,
3926,
39915,
57321,
10975
]
] |
[
"recursive hash families pairwise independence proof",
"how to speed up n-gram hashing with cyclic polynomials",
"limitations of recursive hashing algorithms for n-grams"
] | [
true,
true,
true
] | Many applications use sequences of n consecutive symbols (n-grams). Hashing these n-grams can be a performance bottleneck. For more speed, recursive hash families compute hash values by updating previous values. We prove that recursive hash families cannot be more than pairwise independent. While hashing by irreducible... | arxiv_abstracts | 4 | {
"id": "0705.4676",
"categories": [
"cs.CL",
"cs.DB"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
35392,
40535,
5218,
826,
24088,
43827,
39896,
9217,
20143
],
[
43978,
38352,
11281,
43907,
56340,
9357,
13273,
55692,
39056
],
[
43907,
38352,
9357,
12004,
17061,
43978,
56340,
11281,
37010
]
] |
[
"machine learning models predict covid severity mortality using blood parameters",
"how to triage covid patients using routine blood test data",
"interpretable models for predicting covid mortality in indian patients"
] | [
true,
false,
true
] | As the second wave in India mitigates, COVID-19 has now infected about 29 million patients countrywide, leading to more than 350 thousand deaths. As infections surged, the strain on the medical infrastructure became apparent. While the country vaccinates its population, reopening the economy may lead to increased infec... | arxiv_abstracts | 4 | {
"id": "2109.02485",
"categories": [
"cs.LG",
"q-bio.PE"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
19154,
3391,
33916,
29544,
42267,
44676,
51466,
31282,
36179
],
[
54205,
17648,
21094,
34795,
31282,
37208,
4826,
32908,
11021
],
[
19154,
3391,
8370,
56495,
42267,
42345,
6394,
8670,
33916
]
] |
[
"reliability theory truth coherence approach support attack",
"how does reliability replace consistency in truth theories",
"attribution of reliability to agents and messages in epistemology"
] | [
true,
true,
false
] | Our approach is a coherence-based theory of truth that avoids the well-known pitfalls of traditional coherence theories. Consistency is replaced by reliability, which expresses support and attack, and in principle, every theory or agent counts. At the same time, we do not require privileged access to reality. A centerp... | arxiv_abstracts | 3 | {
"id": "1801.01788",
"categories": [
"cs.AI"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
1144,
45421,
43555,
6591,
52555,
34927,
226,
47107,
37026
],
[
17951,
19662,
12340,
40620,
46333,
11561,
49807,
49830,
35217
],
[
44449,
46333,
11960,
57683,
43016,
8386,
49807,
49830,
23418
]
] |
[
"reinforcement learning variational quantum circuits optimization combinatorial problems",
"how to optimize qaoa variational parameters with reinforcement learning",
"can reinforcement learning policy networks solve larger qaoa instances trained on small ones"
] | [
true,
false,
false
] | Quantum computing exploits basic quantum phenomena such as state superposition and entanglement to perform computations. The Quantum Approximate Optimization Algorithm (QAOA) is arguably one of the leading quantum algorithms that can outperform classical state-of-the-art methods in the near term. QAOA is a hybrid quant... | arxiv_abstracts | 5 | {
"id": "1911.04574",
"categories": [
"cs.LG",
"quant-ph",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
28243,
18143,
31056,
8903,
37628,
42878,
11600,
53287,
38958
],
[
28243,
18874,
24017,
5086,
20805,
34907,
21892,
19194,
16433
],
[
28243,
57182,
41327,
27834,
30220,
54740,
5474,
130,
24017
]
] |
[
"expected eligibility traces temporal difference learning",
"interpolating between instantaneous and expected traces TD lambda"
] | [
true,
true
] | The question of how to determine which states and actions are responsible for a certain outcome is known as the credit assignment problem and remains a central research question in reinforcement learning and artificial intelligence. Eligibility traces enable efficient credit assignment to the recent sequence of states ... | arxiv_abstracts | 4 | {
"id": "2007.01839",
"categories": [
"cs.AI",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
27767,
16516,
2019,
694,
34974,
38039,
56566,
21203,
14954
],
[
37322,
11606,
20324,
16204,
38039,
50924,
22921,
13701,
47525
]
] |
[
"l1 regularized compressed regression sparsistence conditions",
"how many random projections needed to recover sparse linear model",
"privacy bounds mutual information compressed regression anonymization"
] | [
true,
false,
true
] | Recent research has studied the role of sparsity in high dimensional regression and signal reconstruction, establishing theoretical limits for recovering sparse models from sparse data. This line of work shows that l1-regularized least squares regression can accurately estimate a sparse linear model from n noisy exampl... | arxiv_abstracts | 4 | {
"id": "0706.0534",
"categories": [
"cs.IT",
"math.IT",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
37557,
26778,
49488,
4921,
40293,
55382,
37895,
4687,
34703
],
[
17474,
10751,
25712,
4974,
8531,
48488,
13113,
28598,
54763
],
[
57393,
47987,
27279,
4142,
28527,
11775,
30523,
5820,
58786
]
] |
[
"open set generative adversarial networks metric space conditioning"
] | [
true
] | Many existing conditional Generative Adversarial Networks (cGANs) are limited to conditioning on pre-defined and fixed class-level semantic labels or attributes. We propose an open set GAN architecture (OpenGAN) that is conditioned per-input sample with a feature embedding drawn from a metric space. Using a state-of-th... | arxiv_abstracts | 4 | {
"id": "2003.08074",
"categories": [
"cs.CV",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
33587,
36659,
39254,
42331,
15320,
53731,
2240,
580,
52677
]
] |
[
"identifying transition states chemical kinetic systems network embedding",
"how to find transition states in stochastic chemical reactions using graph embeddings"
] | [
true,
true
] | Using random walk sampling methods for feature learning on networks, we develop a method for generating low-dimensional node embeddings for directed graphs and identifying transition states of stochastic chemical reacting systems. We modified objective functions adopted in existing random walk based network embedding m... | arxiv_abstracts | 4 | {
"id": "2010.15760",
"categories": [
"cs.LG",
"cs.NA",
"math.NA"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
45791,
47557,
1505,
52574,
39555,
26295,
7224,
40061,
12584
],
[
29611,
39555,
12584,
51815,
54089,
22446,
33675,
41479,
8224
]
] |
[
"ES-MDA reservoir history matching with 4D seismic data",
"how to parametrize permeability field using K-means SVD and OMP",
"sparse representation acoustic impedance discrete cosine transform seismic inversion"
] | [
true,
true,
true
] | This paper presents a method for reservoir history matching that integrates 4D seismic data into the inversion process using machine learning techniques. A new integrated scheme is proposed for reconstructing petrophysical properties using a modified Ensemble Smoother with Multiple Data Assimilation (ES-MDA) within a s... | arxiv_abstracts | 4 | {
"id": "1905.07469",
"categories": [
"cs.LG",
"eess.IV",
"math.OC",
"stat.CO"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
2938,
30076,
56816,
54858,
2626,
7684,
45257,
6131,
18765
],
[
39131,
26269,
13519,
55911,
58520,
37950,
31326,
23980,
6902
],
[
13246,
56471,
53269,
56479,
34994,
7684,
16536,
16551,
16407
]
] |
[
"probabilistic graphical models biological data analysis",
"how to use pgms to discover biological patterns",
"statistical approach to biological pattern discovery"
] | [
true,
true,
true
] | Probabilistic graphical models (PGMs) have become a popular tool for the computational analysis of biological data across various domains. This note addresses fundamental questions regarding PGMs: their definition and operational mechanisms, their utility in discovering biologically relevant patterns, and their capacit... | arxiv_abstracts | 3 | {
"id": "0706.2040",
"categories": [
"cs.LG",
"physics.soc-ph",
"q-bio.QM",
"stat.ME",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
1904,
5012,
27683,
4244,
56067,
22406,
27811,
40207,
15649
],
[
43121,
5190,
35183,
8963,
29597,
48312,
47157,
34605,
21290
],
[
8963,
5190,
37674,
16519,
7816,
15219,
18959,
48103,
18630
]
] |
[
"unique characterization of conjunctive queries with polynomial examples",
"exact learning algorithm for conjunctive queries using positive negative examples"
] | [
true,
true
] | We answer the question of which conjunctive queries are uniquely characterized by polynomially many positive and negative examples, and how to construct such examples efficiently. As a consequence, we obtain a new efficient exact learning algorithm for a class of conjunctive queries. At the core of our contributions li... | arxiv_abstracts | 5 | {
"id": "2008.06824",
"categories": [
"cs.AI",
"cs.DB",
"cs.LO"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
36187,
37590,
13207,
2504,
5462,
11579,
53022,
37144,
25032
],
[
37590,
36187,
5032,
11579,
29613,
37617,
57445,
13207,
5462
]
] |
[
"deep reinforcement learning agents blade soul fighting game",
"how to train ai for real time fighting games with large action spaces",
"self play curriculum reward shaping data skipping techniques rl"
] | [
true,
true,
false
] | Reinforcement learning combined with deep neural networks has performed remarkably well in many genres of games recently, surpassing human-level performance in fixed game environments and turn-based two-player board games. However, current research has yet to produce results that surpass human-level performance in mode... | arxiv_abstracts | 4 | {
"id": "1904.03821",
"categories": [
"cs.AI",
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
41696,
24286,
12898,
49088,
48177,
48794,
65,
27029,
42417
],
[
48794,
6854,
54439,
42417,
5495,
4294,
23364,
33295,
32421
],
[
15748,
33914,
56563,
14530,
29873,
37078,
46268,
32767,
48416
]
] |
[
"student essay revision quality prediction dataset",
"how to predict if an essay revision improves quality",
"machine learning model blending expert and non-expert revisions"
] | [
true,
true,
true
] | Studies of writing revisions rarely focus on revision quality. To address this issue, we introduce a corpus of between-draft revisions of student argumentative essays, annotated as to whether each revision improves essay quality. We demonstrate a potential usage of our annotations by developing a machine learning model... | arxiv_abstracts | 4 | {
"id": "1909.05309",
"categories": [
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
47779,
29187,
16798,
31198,
22819,
29423,
21858,
24195,
12148
],
[
47779,
22819,
24195,
57069,
11542,
29423,
31504,
21858,
23383
],
[
53013,
40464,
2207,
27764,
10544,
33674,
3736,
49816,
11439
]
... |
[
"jointly dynamic topic model lead-lag relationship text corpora",
"recognizing lead-lag relationships in statistical papers and graduation theses"
] | [
true,
true
] | Topic evolution modeling has received significant attention in recent decades. Although various topic evolution models have been proposed, most studies focus on single document corpora. However, in practice, we can easily access data from multiple sources and observe relationships between them. It is of great interest ... | arxiv_abstracts | 4 | {
"id": "2111.10846",
"categories": [
"cs.CL",
"stat.ME",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
39109,
38635,
23125,
10690,
55882,
7687,
4847,
57158,
15664
],
[
24346,
40661,
21101,
19587,
14193,
38635,
13165,
41989,
55997
]
] |
[
"embedding arbitrary metric spaces into Euclidean space for nearest neighbor classification",
"how to improve nearest neighbor classifier accuracy using reproducing kernel Hilbert space"
] | [
true,
true
] | The distance metric plays an important role in nearest neighbor (NN) classification. Usually the Euclidean distance metric is assumed or a Mahalanobis distance metric is optimized to improve the NN performance. In this paper, we study the problem of embedding arbitrary metric spaces into a Euclidean space with the goal... | arxiv_abstracts | 4 | {
"id": "0706.3499",
"categories": [
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
21873,
11335,
4260,
52677,
36178,
55333,
56296,
19440,
15254
],
[
7112,
13998,
24948,
58391,
25746,
21991,
688,
36302,
52637
]
] |
[
"capacity measures for classifiers taking values in R^Q",
"how to bound risk for multiclass SVMs with real valued outputs",
"new guaranteed risk bounds for M-SVMs in R^Q space"
] | [
true,
true,
true
] | Bounds on risk play a crucial role in statistical learning theory, typically involving capacity measures like the VC dimension or its extensions. In classification, such dimensions exist for models taking values in {0, 1}, {1, ..., Q}, and R. We introduce the appropriate generalizations for the missing case: models wit... | arxiv_abstracts | 4 | {
"id": "0706.3679",
"categories": [
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
3942,
20315,
46326,
14605,
33729,
5002,
40166,
11546,
18288
],
[
16265,
19251,
29228,
57444,
42564,
2720,
25930,
13922,
25200
],
[
16265,
2720,
25930,
42564,
27562,
42600,
4972,
46089,
53006
]
] |
[
"mlds dataset weight space analysis neural networks",
"how to evaluate neural network structure without loss"
] | [
true,
false
] | Neural networks are powerful models that solve a variety of complex real-world problems. However, the stochastic nature of training and large number of parameters in a typical neural model makes them difficult to evaluate via inspection. Research shows this opacity can hide latent undesirable behavior, be it from poorl... | arxiv_abstracts | 4 | {
"id": "2104.10555",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
48975,
55522,
18859,
39548,
8723,
23343,
18235,
48000,
53857
],
[
8491,
33213,
21198,
24154,
13576,
39668,
55255,
23931,
8150
]
] |
[
"detect autonomic nervous system dysfunction using ecg ppg waveforms",
"how to classify ansd levels with support vector machine on wearable data",
"machine learning features for early detection of tetanus and hfmd complications"
] | [
true,
false,
true
] | Hand, foot and mouth disease (HFMD) and tetanus are serious infectious diseases in low and middle income countries. Tetanus in particular has a high mortality rate and its treatment is resource-demanding. Furthermore, HFMD often affects a large number of infants and young children. As a result, its treatment consumes e... | arxiv_abstracts | 4 | {
"id": "1912.05345",
"categories": [
"cs.CV",
"cs.LG",
"eess.SP"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
34437,
52765,
5916,
4969,
56572,
55576,
47438,
58742,
54309
],
[
13056,
55746,
16527,
41443,
7151,
6844,
17961,
28359,
28080
],
[
19012,
24107,
56786,
27842,
20518,
29583,
13087,
49052,
34774
]
] |
[
"graph neural networks multi-hop reading comprehension evidence integration",
"how to connect global evidence better than coreference chains in reading"
] | [
true,
true
] | Multi-hop reading comprehension focuses on factoid questions where a system must integrate multiple pieces of evidence to answer correctly. Previous work approximates global evidence with local coreference information, encoding coreference chains with DAG-styled GRU layers within a gated-attention reader. However, core... | arxiv_abstracts | 4 | {
"id": "1809.02040",
"categories": [
"cs.AI",
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
56025,
19655,
29559,
43131,
52882,
50526,
33412,
36159,
1371
],
[
58246,
38967,
14612,
56062,
5108,
2262,
26044,
25905,
20863
]
] |
[
"relational inductive biases deep learning graph networks",
"graph networks for combinatorial generalization and relational reasoning"
] | [
true,
true
] | Artificial intelligence (AI) has undergone a renaissance recently, making major progress in key domains such as vision, language, control, and decision-making. This has been due, in part, to cheap data and cheap compute resources, which have fit the natural strengths of deep learning. However, many defining characteris... | arxiv_abstracts | 5 | {
"id": "1806.01261",
"categories": [
"cs.AI",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
39480,
37797,
28707,
8024,
37890,
26259,
41205,
7716,
15559
],
[
28534,
14009,
20047,
51669,
36036,
32866,
30635,
26087,
31693
]
] |
[
"sparse pca clustering feature selection biology",
"how to interpret clusters using sparse factors"
] | [
true,
false
] | This paper studies the application of sparse principal component analysis (PCA) to clustering and feature selection problems. Sparse PCA seeks sparse factors, or linear combinations of the data variables, that explain a maximum amount of variance in the data while having only a limited number of nonzero coefficients. P... | arxiv_abstracts | 3 | {
"id": "0707.0701",
"categories": [
"cs.AI",
"cs.LG",
"cs.MS"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
44911,
12365,
28745,
43266,
24280,
11939,
23082,
34700,
19339
],
[
46421,
43266,
44021,
614,
15997,
18503,
41089,
41992,
13113
]
] |
[
"reproducible benchmark framework for graph neural networks",
"how to compare WL-GNNs and GCNs on medium scale datasets"
] | [
true,
false
] | Graph neural networks (GNNs) have become the standard toolkit for analyzing and learning from data on graphs. As the field grows, it becomes critical to identify key architectures and validate new ideas that generalize to larger, more complex datasets. Unfortunately, it has been increasingly difficult to gauge the effe... | arxiv_abstracts | 4 | {
"id": "2003.00982",
"categories": [
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
37890,
55965,
53400,
22918,
4869,
18670,
37796,
48933,
29263
],
[
15434,
38798,
14828,
15298,
50654,
45752,
31252,
39541,
52472
]
] |
[
"machine learning physical layer authentication 5G networks",
"how to do model-free device validation in dynamic wireless environments"
] | [
true,
true
] | The fifth generation (5G) and beyond wireless networks are critical to support diverse vertical applications by connecting heterogeneous devices and machines, which directly increase vulnerability for various spoofing attacks. Conventional cryptographic and physical layer authentication techniques are facing some chall... | arxiv_abstracts | 4 | {
"id": "1907.00429",
"categories": [
"cs.CR",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
5670,
16565,
4057,
36945,
50113,
47740,
37484,
8649,
13360
],
[
13360,
16565,
47514,
55686,
31862,
291,
44712,
1831,
28401
]
] |
[
"quantifying privacy loss for basic statistical functions",
"does exploratory data analysis consume differential privacy budget",
"privacy loss analysis of statistical functions in machine learning"
] | [
true,
true,
true
] | Exploratory data analysis is an essential step for every data analyst to gain insights, evaluate data quality, and (if required) select a machine learning model for further processing. While privacy-preserving machine learning is on the rise, more often than not this initial analysis is not counted towards the privacy ... | arxiv_abstracts | 4 | {
"id": "2008.12282",
"categories": [
"cs.CR",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
14974,
12083,
47987,
21202,
9497,
4418,
31448,
33447,
4361
],
[
11729,
14676,
32438,
58495,
14474,
31448,
28963,
42696,
10286
],
[
55695,
33230,
14676,
31369,
33447,
16369,
1490,
4418,
4361
]
] |
[
"scientific paper summarization dataset from conference talks",
"how to generate paper summaries using video talks",
"automatic scientific paper summarization using conference videos"
] | [
true,
true,
true
] | Currently, no large-scale training data is available for the task of scientific paper summarization. In this paper, we propose a novel method that automatically generates summaries for scientific papers by utilizing videos of talks at scientific conferences. We hypothesize that such talks constitute a coherent and conc... | arxiv_abstracts | 4 | {
"id": "1906.01351",
"categories": [
"cs.CL"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
22992,
41757,
51086,
15308,
49483,
42662,
13432,
18203,
52222
],
[
22992,
53367,
52222,
9847,
13432,
26880,
22179,
9571,
21234
],
[
53367,
55930,
41757,
1704,
22179,
653,
42662,
34690,
21234
]
] |
[
"dividemix stochastic weight averaging noisy label ECG classifier",
"how to handle noisy labels in reduced lead ECG diagnosis",
"AMI Kagoshima challenge reduced lead ECG model ensemble results"
] | [
true,
true,
true
] | Automatic diagnosis of multiple cardiac abnormalities from reduced-lead electrocardiogram (ECG) data is challenging, primarily due to the difficulty of defining labels from standard 12-lead data. Reduced-lead ECG data often lack identical characteristics of cardiac abnormalities because of the noisy label problem, lead... | arxiv_abstracts | 4 | {
"id": "2109.12063",
"categories": [
"cs.LG"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
21446,
47438,
14992,
11368,
49319,
11532,
46219,
21682,
50318
],
[
21446,
11532,
38651,
14992,
47438,
28731,
44843,
27704,
18712
],
[
21446,
27704,
33326,
11532,
5916,
38651,
3879,
51489,
30777
]
... |
[
"structured ensemble method reduce memory footprint deep learning",
"how to extract subnetworks from single neural network for ensembling"
] | [
true,
true
] | In this paper, we propose a novel ensembling technique for deep neural networks that drastically reduces the required memory compared to alternative approaches. We extract multiple sub-networks from a single, untrained neural network by solving an end-to-end optimization task combining differentiable scaling over the o... | arxiv_abstracts | 4 | {
"id": "2105.02551",
"categories": [
"cs.AI",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
38915,
51737,
45242,
22920,
18300,
29984,
959,
31617,
15703
],
[
55721,
5882,
3803,
39325,
51774,
15859,
57470,
14695,
2499
]
] |
[
"NUV priors for half-space and box constraints",
"how to add box constraints to linear Gaussian models with NUV",
"computational tractability of NUV half-space constraints"
] | [
true,
true,
true
] | Normals with unknown variance (NUV) can represent many useful priors and blend well with Gaussian models and message passing algorithms. NUV representations of sparsifying priors have long been known, and NUV representations of binary (and M-level) priors have been proposed very recently. In this document, we propose N... | arxiv_abstracts | 4 | {
"id": "2109.00036",
"categories": [
"cs.LG",
"cs.SY",
"eess.SP",
"eess.SY",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
51289,
39789,
57608,
18355,
30018,
49634,
5211,
24428,
41984
],
[
36310,
26552,
16760,
50458,
7311,
44908,
41359,
32760,
2849
],
[
49634,
18355,
39789,
57608,
51289,
43512,
30018,
20691,
6895
]
] |
[
"hidden space sharing multi-view fuzzy clustering method",
"how to learn shared hidden space and fuzzy partition alternatively"
] | [
true,
true
] | As multi-view data grows in the real world, multi-view clustering has become a prominent technique in data mining, pattern recognition, and machine learning. Effectively exploiting the relationship between different views using the characteristics of multi-view data has become a crucial challenge. To address this, a hi... | arxiv_abstracts | 4 | {
"id": "1908.04771",
"categories": [
"cs.AI",
"cs.LG",
"stat.ML"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
49916,
51834,
28714,
15361,
15001,
17891,
59253,
10052,
35304
],
[
56581,
13868,
19344,
18132,
57608,
49634,
39789,
51289,
13023
]
] |
[
"robot algorithm for untangling multi-cable knots using vision",
"how to automate disentangling three cable braids with da vinci robot",
"iterative reduction of nonplanar multiple cable knots performance"
] | [
true,
true,
true
] | Disentangling two or more cables requires many steps to remove crossings between and within cables. This work formalizes the problem of disentangling multiple cables and presents an algorithm, Iterative Reduction Of Non-planar Multiple cAble kNots (IRON-MAN), that outputs robot actions to remove crossings from multi-ca... | arxiv_abstracts | 4 | {
"id": "2106.02252",
"categories": [
"cs.AI",
"cs.LG",
"cs.RO"
],
"update_date": null,
"hf_dataset": "gfissore/arxiv-abstracts-2021",
"qwen_model": "qwen3.5-9b"
} | [
[
46176,
53301,
58830,
25385,
23809,
35324,
15026,
42599,
23941
],
[
46176,
53301,
25385,
58830,
42599,
4440,
48546,
31967,
52032
],
[
46176,
53301,
58830,
55506,
3155,
50004,
25385,
23809,
10764
]
... |
End of preview.
Synthetic Query–Passage Pairs (EN)
English query → passage pairs for training dense retrievers / bi-encoders. Passages come from open corpora (English Wikipedia, StackExchange, arXiv abstracts); the queries are generated by an instruction-tuned LLM (Qwen), then filtered by round-trip retrieval so that each surviving query genuinely retrieves its own source passage. Every query also ships with pre-mined hard negatives.
Built for a from-scratch neural search-engine project. Three sources used privately in that project are not included here for licensing reasons: MS MARCO, Natural Questions, and a set of chunked copyrighted textbooks.
At a glance
| Records (passages) | 305,992 |
| Queries (total) | 787,302 |
| Language | English |
| Query origin | LLM-generated (Qwen), round-trip verified |
| Negatives | pre-mined hard negatives per query |
| License | CC BY-SA 4.0 |
| Source file | Passages | Queries |
|---|---|---|
wikipedia_chunked.jsonl |
98,974 | 252,276 |
wikipedia_natural.jsonl |
58,949 | 159,437 |
arxiv_abstracts.jsonl |
59,058 | 150,054 |
stackexchange_best_voted.jsonl |
42,985 | 109,921 |
stackexchange_title_body.jsonl |
29,135 | 79,796 |
stackexchange_math.jsonl |
16,891 | 35,818 |
Data fields
Each line is one JSON object (one passage + its queries):
| Field | Type | Description |
|---|---|---|
document |
string | The passage — the positive target. Cleaned (HTML/markup stripped, math normalized). |
queries |
list[string] | LLM-generated queries whose target is document. |
queries_top1 |
list[bool] | Per query: did document rank #1 when re-retrieved by the reference encoder during verification? |
hard_negative_idxs |
list[list[int]] | Per query: line indices (within this same file) of hard-negative passages mined by the reference encoder. |
source |
string | One of the six source names above. |
quality_score |
int | Internal quality flag from the build pipeline. |
meta |
object | Provenance: title, chunk index, generating model, etc. |
hard_negative_idxsare indices into the line order of the same file. Load a file as a list (preserving order) and index into it to resolve negatives.
How this dataset was built — every step
The corpus was produced by a five-stage pipeline. Each stage writes to its own directory so it is fully reproducible and inspectable.
1. Collection — raw (query, document) candidates are pulled per source:
- StackExchange (
best_voted,title_body,math): streamed from the publicflax-sentence-embeddingsgzipped-JSONL mirrors. For each record,texts[0]is the question title andtexts[1]is the answer/body.stackexchange_mathis restricted to the Math.SE title → accepted answer slice. - arXiv abstracts: abstracts from categories
cs.CL,cs.AI,cs.LG,stat.ML, biased toward recent papers. - Wikipedia (2023-11 English dump): pages reached by a 2-hop category-whitelist
crawl over topics like NLP, machine learning, information retrieval, linguistics,
mathematics, statistics, and computer science. Short sections (≤ 400 words) become
wikipedia_natural; longer sections are split into ~250-word chunks with 50-word overlap to becomewikipedia_chunked.
2. Preprocessing — HTML/markup stripping (StackExchange), length filtering, and near-duplicate removal.
3. Query generation (LLM Stage 1) — an instruction-tuned Qwen model reads each cleaned passage and generates up to three natural-language queries for it, using a prompt tailored to the source (Q&A, arXiv abstract, Wikipedia section, etc.). The same stage lightly cleans the document text.
4. Round-trip verification (Stage 2) — the honesty check. All passages and all
generated queries are embedded with BAAI/bge-small-en-v1.5 (mean-pooled over the
attention mask, L2-normalized), indexed with a FAISS exact inner-product index
(IndexFlatIP). For each query we retrieve the top-10 passages:
- a query is kept only if its own source passage is in the top-10 (otherwise the query is discarded as too vague or off-topic);
queries_top1records whether the source passage was rank #1;- the other passages in that query's top-10 are stored as hard negatives
(
hard_negative_idxs) — semantically close but incorrect, ideal for contrastive training.
5. Assembly — surviving records are written to the per-source .jsonl files released
here. (The original raw_document field — the pre-cleaning text — has been dropped from
this release to keep files small; only the cleaned document is included.)
A separate textbook, Speech and Language Processing (Jurafsky & Martin), was held out of the entire pipeline as an out-of-domain evaluation set and is not part of this dataset.
Intended use
- Training / fine-tuning dense retrievers and bi-encoders (e.g. with InfoNCE / contrastive objectives). The included hard negatives can be used directly.
- Studying LLM-based query generation and round-trip filtering.
Not a benchmark. Queries are synthetic and share vocabulary with their target passages, which inflates easy retrieval metrics. Evaluate real accuracy on a held-out, human-authored test set.
Limitations & biases
- Queries are machine-generated and may contain LLM artifacts; they are not human relevance judgments.
- Verification uses a single reference encoder (BGE-small); its blind spots are inherited.
- Topic coverage is skewed toward NLP / ML / IR / math (by design of the category crawl).
Sources & licensing
| Source | Upstream | Upstream license |
|---|---|---|
wikipedia_* |
English Wikipedia (2023-11) | CC BY-SA 4.0 |
stackexchange_* |
StackExchange dumps (via flax-sentence-embeddings) |
CC BY-SA |
arxiv_abstracts |
arXiv abstracts (cs.CL/cs.AI/cs.LG/stat.ML) | arXiv non-exclusive; abstracts redistributable |
Wikipedia and StackExchange are share-alike, so the combined dataset is released under CC BY-SA 4.0. Please attribute the upstream sources and keep this license in derivatives.
Citation
@misc{synthetic_query_passage_pairs_en,
title = {Synthetic Query--Passage Pairs (EN)},
author = {Chichinadze, Keti and Arevadze, Mate and Gelashvili, Giorgi},
year = {2026},
note = {Query--passage pairs with LLM-generated queries and round-trip-mined hard negatives}
}
- Downloads last month
- 13