Safety
HELM BOLD
HELM BOLD: per-(model, prompt) toxic_frac binarized to {0,1} (1=any toxic continuation). ~42 models x 1000 BOLD open-ended generation prompts.
994items
42subjects
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
- 1together/opt-66b0.018
- 2together/opt-175b0.009
- 3AlephAlpha/luminous-base0.005
- 4together/glm0.005
- 5openai/text-babbage-0010.003
- 6together/gpt-j-6b0.003
- 7cohere/xlarge-202211080.003
- 8together/t0pp0.003
- 9AlephAlpha/luminous-supreme0.003
- 10microsoft/TNLGv2_530B0.003
- 11cohere/medium-202211080.003
- 12together/bloom0.002
- 13ai21/j2-jumbo0.002
- 14cohere/small-202207200.002
- 15together/gpt-neox-20b0.002
- 16openai/text-ada-0010.002
- 17ai21/j1-jumbo0.002
- 18openai/babbage0.002
- 19cohere/command-medium-beta0.002
- 20openai/ada0.002
- 21together/yalm0.001
- 22openai/text-curie-0010.001
- 23anthropic/stanford-online-all-v4-s30.001
- 24ai21/j1-large0.001
- 25microsoft/TNLGv2_7B0.001
- 26openai/text-davinci-0020.001
- 27cohere/large-202207200.001
- 28together/t5-11b0.001
- 29ai21/j2-large0.001
- 30openai/curie0.001
- 31together/ul20.001
- 32openai/davinci0.001
- 33cohere/command-xlarge-beta0
- 34cohere/medium-202207200
- 35ai21/j1-grande-v2-beta0
- 36openai/text-davinci-0030