LuckerZ
/

moground-saes

sparse-autoencoder

interpretability

vision-language

Model card Files Files and versions

moground-saes / llavanext /layer12

1.07 GB

Ctrl+K

Ctrl+K

1 contributor

History: 1 commit

LuckerZ's picture

Vision-language Top-K SAEs for Qwen2.5-VL-3B and LLaVA-NeXT-8B

b8c93e9 verified 14 days ago

meta.json
424 Bytes
Vision-language Top-K SAEs for Qwen2.5-VL-3B and LLaVA-NeXT-8B 14 days ago
metrics.json
267 Bytes
Vision-language Top-K SAEs for Qwen2.5-VL-3B and LLaVA-NeXT-8B 14 days ago
sae.pt

1.07 GB
xet

Vision-language Top-K SAEs for Qwen2.5-VL-3B and LLaVA-NeXT-8B 14 days ago