refusals

community

AI & ML interests

None defined yet.

avrecum

authored a paper 3 months ago

Cannot or Should Not? Automatic Analysis of Refusal Composition in IFT/RLHF Datasets and Refusal Behavior of Black-Box LLMs

Paper • 2412.16974 • Published Dec 22, 2024 • 3

avrecum

updated 2 datasets 3 months ago

refusals/refusal_dataset_ultra

Viewer • Updated Nov 8, 2025 • 7.2M • 55 • 3

refusals/refusal_dataset_100k

Viewer • Updated Nov 8, 2025 • 104k • 9 • 1

inxoy

updated a dataset 4 months ago

refusals/refusal_dataset_100k

Viewer • Updated Nov 8, 2025 • 104k • 9 • 1

inxoy

in refusals/refusal_dataset_ultra 4 months ago

Create LICENSE

#3 opened 4 months ago by

Update README.md

#2 opened 4 months ago by

avrecum

updated 13 datasets about 1 year ago

refusals/human_multi_classifications_500

Viewer • Updated Dec 3, 2024 • 501 • 4

refusals/human_multi_classifications

Viewer • Updated Dec 3, 2024 • 667 • 3

refusals/gemini_1_5_pro_classifications_multi_human

Viewer • Updated Dec 3, 2024 • 661 • 5

refusals/command_r_plus_classifications_multi_human

Viewer • Updated Dec 3, 2024 • 661 • 6

refusals/mistral_large_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 661 • 2

refusals/qwen2_72b_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 661 • 6

refusals/llama_3_1_70b_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 661 • 4

refusals/llama_3_1_405b_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 661 • 5

refusals/llama_3_1_8b_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 660 • 2

refusals/gpt_4o_mini_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 661 • 2

refusals/gpt_4o_classifications_multi_human

Viewer • Updated Dec 2, 2024 • 661 • 2

refusals/predictions_logistic_classifier

Viewer • Updated Dec 2, 2024 • 607 • 2

refusals/human_classifications

Viewer • Updated Dec 2, 2024 • 8.65k • 9

mvh

authored a paper about 2 years ago

Black-Box Access is Insufficient for Rigorous AI Audits

Paper • 2401.14446 • Published Jan 25, 2024 • 3