Retrogradi/winogender_experiments
Preview • Updated • 21
Bias Mitigation via Mechanistic Interpretability
-
-
Retrogradi/gender_stereoset_rephrased
Viewer • Updated • 1.79k • 50 -
Retrogradi/gemma2b_best_model_wino_sft_attn_0.5_ul1.0_lr1e-05
Updated -
Retrogradi/llama3.2_best_model_wino_sft_attn_1.0_ul1.0_lr1e-06
Updated -
Retrogradi/gpt2xl_wino_dpo_mlp_from_attn_10.0_beta0.3_lr5e-06
Updated -
Retrogradi/gpt2xl_wino_dpo_attn_0.8_beta0.5_lr5e-06
Updated -
Retrogradi/llama3.2_wino_dpo_mlp_from_attn_10.0_beta0.5_lr1e-05
Updated -
Retrogradi/gemma2b_wino_dpo_mlp_impact_only_0.5_beta0.5_lr1e-06
Updated -
Retrogradi/llama3.2_wino_cmp_dpo_snr_mlp_1layers_beta0.3_lr5e-06
Updated -
Retrogradi/gemma2b_best_model_wino_cmp_sft_snr_mlp_1layers_ul1.0_lr5e-06
Updated