SKILL-KD: Contrastive Skill Distillation for LLM Agents
Paper • 2607.28048 • Published • 8
🔄 In a Training Loop
Urro
urroxyz
AI & ML interests
computational linguistics major 🤖🔎ðŸ”
i am autistic. if i come off rude, i probably didn't mean to. please feel free to ask me for clarification.
Recent Activity
upvoted a paper about 1 hour ago
SKILL-KD: Contrastive Skill Distillation for LLM Agents updated a collection about 1 hour ago
WTF GENIUS PAPERS updated a collection about 1 hour ago
WTF GENIUS PAPERSOrganizations
upvoted 5
papers about 1 hour ago
Know When to Stop: Segment-Level Credit Assignment for Reducing Overthinking
Paper • 2607.00482 • Published • 6
ARCHead: Activation-Metric Residual Correction for Large Language Model Output Heads
Paper • 2608.02703 • Published • 5
Distill Where You Fail: Recovering Learning Signals of Negative RL-Groups from Adaptive Teacher Guidance
Paper • 2608.00782 • Published • 11
Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning
Paper • 2608.05139 • Published • 20
upvoted a paper about 15 hours ago
upvoted 5
papers 1 day ago
RestoreKV: Recovering Full-Cache Behavior Under Aggressive Query-Agnostic KV Cache Eviction
Paper • 2608.01247 • Published • 10
LLaDA MoE v2: Scaling Mixture-of-Experts Diffusion Language Models
Paper • 2608.03457 • Published • 27
MemSFT: Mitigating Alignment Tax with an External Parametric Memory
Paper • 2607.25614 • Published • 22
PCSD: Persistent Consistency for Self-Distillation in Agentic Reinforcement Learning
Paper • 2608.01837 • Published • 37
TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning
Paper • 2608.04007 • Published • 17
upvoted 4
papers 2 days ago
Zero-Mem: Zero-Token Memory Operations for LLM Agents
Paper • 2607.29377 • Published • 9
SKT: Skill-Use Training at Scale via Verified Synthetic Data Generation
Paper • 2608.02287 • Published • 29
GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning
Paper • 2608.02585 • Published • 23
DAPD: Dual-Anchored Policy Distillation
Paper • 2608.01735 • Published • 100
upvoted a paper 3 days ago
upvoted 4
papers 6 days ago
MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis
Paper • 2607.27146 • Published • 27
Multi-Head Attention Residuals
Paper • 2607.27230 • Published • 10
Metis: Memory Foundation Model
Paper • 2607.26760 • Published • 268
Flux-OPD: On-Policy Distillation with Evolving Contexts
Paper • 2607.28022 • Published • 43