Safety
HELM BBQ
HELM BBQ: per-(model, question) exact_match in {0,1} (1=correct) on 1000 BBQ social-bias multiple-choice questions. 87 models.
999items
81subjects
Apache-2.0license
safetydomain
textmodality
item-level responses released
Saturation status: Unknown
Response matrix
Fit to width. Hover for subject & item; click a cell for details.

Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect
Subjects
- 1anthropic/claude-opus-4-202505140.993
- 2anthropic/claude-sonnet-4-5-202509290.989
- 3openai/gpt-oss-120b0.985
- 4google/gemini-3-pro-preview0.984
- 5anthropic/claude-sonnet-4-202505140.979
- 6openai/o3-2025-04-160.979
- 7zai-org/glm-4.5-air-fp80.978
- 8openai/gpt-5-nano-2025-08-070.976
- 9google/gemini-2.5-flash-preview-04-170.975
- 10openai/o1-2024-12-170.973
- 11anthropic/claude-opus-4-20250514-thinking-10k0.971
- 12openai/o1-mini-2024-09-120.969
- 13openai/gpt-5-2025-08-070.968
- 14openai/gpt-oss-20b0.967
- 15xai/grok-3-mini-beta0.967
- 16deepseek-ai/deepseek-v30.967
- 17deepseek-ai/deepseek-r10.966
- 18deepseek-ai/deepseek-r1-hide-reasoning0.966
- 19qwen/qwen3-235b-a22b-fp8-tput0.966
- 20anthropic/claude-sonnet-4-20250514-thinking-10k0.966
- 21google/gemini-2.5-pro-preview-03-250.964
- 22deepseek-ai/deepseek-r1-05280.963
- 23openai/gpt-5-mini-2025-08-070.963
- 24qwen/qwen3-235b-a22b-instruct-2507-fp80.962
- 25writer/palmyra-x-0040.955
- 26meta/llama-3.1-70b-instruct-turbo0.954
- 27qwen/qwen2-72b-instruct0.951
- 28openai/gpt-4o-2024-05-130.951
- 29google/gemini-2.5-flash-lite0.949
- 30moonshotai/kimi-k2-instruct0.949
- 31anthropic/claude-3-5-sonnet-202406200.949
- 32writer/palmyra-x50.948
- 33google/gemini-1.5-flash-0010.947
- 34meta/llama-3.1-405b-instruct-turbo0.945
- 35google/gemini-1.5-pro-0010.945
- 36writer/palmyra-fin0.942