← Models

Model profile

GPT J 6B

eleutheraideveloper
2021-06-09release date
Not rankedSafety rank
#394 / 645Freedom rank

Evidence summary

Safety. GPT J 6B does not meet the evidence gate for a Safety rank. Published Safety evidence spans 2 eval lineages and 1 of 7 components. Its strongest relative result is HELM Classic RealToxicityPrompts (toxic_prompt_toxic_fraction, #17 of 40); its weakest is ChineseSafe (score, #11 of 22).

Freedom. GPT J 6B has an estimated Freedom rank of #394; its 90% source-sensitivity interval is #206–#525. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is HELM Classic RealToxicityPrompts (toxic_prompt_toxic_fraction, #24 of 40); its weakest is HELM Classic RealToxicityPrompts (toxic_prompt_toxic_fraction, #24 of 40).

Compare this model

Only models sharing at least one published sub-eval are listed.

Official and reference links

Safety evals

Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.

Eval / sub-evalRankValueDistributionSource
ChineseSafescore#11 / 2255.98Source ↗official
HELM Classic RealToxicityPromptsnon_toxic_prompt_toxic_fraction#17 / 400.01905Source ↗official
HELM Classic RealToxicityPromptstoxic_prompt_toxic_fraction#17 / 400.0852Source ↗official

Freedom evals

Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.

Eval / sub-evalRankValueDistributionSource
HELM Classic RealToxicityPromptsnon_toxic_prompt_toxic_fraction#24 / 400.01905Source ↗official
HELM Classic RealToxicityPromptstoxic_prompt_toxic_fraction#24 / 400.0852Source ↗official