arxiv:2606.12138
Gleb Gerasimov
gudleifrr
AI & ML interests
NLP, interpretability
Recent Activity
authored a paper 4 days ago
Small Vectors, Big Effects: A Mechanistic Study of RL-Induced Reasoning
via Steering Vectors authored a paper 4 days ago
Train One Sparse Autoencoder Across Multiple Sparsity Budgets to
Preserve Interpretability and Accuracy authored a paper 4 days ago
Teach Old SAEs New Domain Tricks with BoostingOrganizations
None yet
models 214
gudleifrr/gpt2_saes
Updated
gudleifrr/sae_Qwen_Qwen2.5-Math-7B_diff_blocks.15.hook_resid_post_16384_batchtopk_64_0.001_9715
Updated
gudleifrr/sae_Qwen_Qwen2.5-Math-7B_diff_blocks.10.hook_resid_post_16384_batchtopk_64_0.001_1376
Updated
gudleifrr/sae_Qwen_Qwen2.5-Math-7B_diff_blocks.10.hook_resid_post_16384_batchtopk_64_0.001_3866
Updated
gudleifrr/sae_Qwen_Qwen2.5-Math-7B_diff_blocks.10.hook_resid_post_16384_batchtopk_64_0.001_9689
Updated
gudleifrr/sae_Qwen_Qwen2.5-7B_diff_blocks.10.hook_resid_post_16384_batchtopk_64_0.001_8634
Updated
gudleifrr/sae_Qwen_Qwen2.5-Math-7B_diff_blocks.10.hook_resid_post_16384_batchtopk_64_0.001_r1
Updated
gudleifrr/sae_Qwen_Qwen2.5-7B_diff_blocks.10.hook_resid_post_16384_batchtopk_64_0.001_r1
Updated
gudleifrr/sae_Qwen_Qwen2.5-7B_default_ln_final.hook_normalized_16384_batchtopk_64_0.001
Updated
gudleifrr/sae_Qwen_Qwen2.5-7B_default_blocks.10.hook_resid_post_16384_batchtopk_64_0.001
Updated
datasets 9
gudleifrr/gpt2_saes
Updated • 3
gudleifrr/interpretations
Updated • 6
gudleifrr/OpenThoughts-114k-full-fix
Viewer • Updated • 114k • 10
gudleifrr/steering_diff
Preview • Updated • 117
gudleifrr/MATH-500
Viewer • Updated • 500 • 50
gudleifrr/OpenThoughts-114k-full
Viewer • Updated • 114k • 52
gudleifrr/linear_probing
Viewer • Updated • 8 • 10
gudleifrr/OpenThoughts-114k-thinking
Viewer • Updated • 114k • 17
gudleifrr/text-correction-en
Viewer • Updated • 784k • 16