models
Open weights, fine-tunes and adapters. Every listing here comes live from the Hugging Face Hub, attributed to it, and links back to the source.
dpo-harmless_saferlhfdpo-mistral-7b-harmless-v2helpful_harmless_honestRM-HH-Gemma_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenTrueRM-HH-Gemma_harmless_gpt3_20000_gemma2b_shuffleTrue_extractchosenTrueRM-HH-Gemma_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenFalseRM-HH-AllMix_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenTrueRM-harmless_harmless_gpt4_loraR64_20000_gemma2b_lr5e-06_bs2_g4dpo_harmlessharmless_gpt3_gamma0.0_beta0.1_subset20000_modelmistral7b_maxsteps5000_bz8_lr5e-06RM-HH-Mix_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenFalseRM-harmless_harmless_contrast_loraR64_20000_gemma2b_lr5e-05_bs2_g4REPROD_dpo_harmlessharmless_gpt4_subset-1_modelgemma7b_maxsteps10000_bz8_lr5e-06RM-HH-Mix_harmless_gpt3_20000_gemma2b_shuffleTrue_extractchosenTrueRM-HH-AllMix_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenFalseRM-HH-AllMix_harmless_gpt3_20000_gemma2b_shuffleTrue_extractchosenFalsedpo_harmlessharmless_human_gamma5.0_beta0.1_subset-1_modelmistral7b_maxsteps5000_bz8_lr1e-06RM-harmless_harmless_human_loraR64_20000_gemma2b_lr1e-06_bs2_g4RM-harmless_harmless_human_loraR64_20000_gemma2b_lr1e-05_bs2_g4dpo_harmlessharmless_gpt4_subset20000_modelgpt2_maxsteps5000_bz8_lr5e-06RM-harmless_harmless_contrast_loraR64_20000_gemma2b_lr1e-06_bs2_g4RM-harmless_harmless_gpt4_loraR64_20000_gemma2b_lr1e-05_bs2_g4REPROD_dpo_harmlessharmless_human_subset-1_modelgemma2b_maxsteps6000_bz8_lr5e-05RM-HH-Gemma_harmless_human_20000_gemma2bRM-HH-Mix_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenTruedpo_harmlessharmless_human_subset20000_modelgpt2_maxsteps5000_bz8_lr5e-06RM-harmless_harmless_gpt3_loraR64_20000_gemma2b_lr1e-06_bs2_g4RM-harmless_harmless_gpt4_loraR64_20000_gemma2b_lr1e-06_bs2_g4dpo_harmlessharmless_contrast_subset20000_modelgpt2_maxsteps5000_bz8_lr5e-06REPROD_dpo_harmlessharmless_human_subset-1_modelgemma2b_maxsteps6000_bz8_lr1e-05retrain_dpo_mistral_7B_helpful_harmless
