Model profile

Text Davinci 002

OpenAIdeveloper

2022-03-15release date

Not rankedSafety rank

#110 / 645Freedom rank

Evidence summary

Safety. Text Davinci 002 does not meet the evidence gate for a Safety rank. Published Safety evidence spans 2 eval lineages and 4 of 7 components. Its strongest relative result is SafetyPrompts (score, #8 of 15); its weakest is HELM Classic RealToxicityPrompts (non_toxic_prompt_toxic_fraction, #34 of 40).

Freedom. Text Davinci 002 has an estimated Freedom rank of #110; its 90% source-sensitivity interval is #63–#399. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is HELM Classic RealToxicityPrompts (non_toxic_prompt_toxic_fraction, #6 of 40); its weakest is SafetyPrompts (score, #8 of 15).

Compare this model

Second model

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

Official and reference links

Official model documentation ↗Family-level model document · openai · first party
Release source ↗direct research preview identity date

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-eval	Rank	Value	Source
HELM Classic RealToxicityPromptsnon_toxic_prompt_toxic_fraction	#34 / 40	↓0.02471	Source ↗official
HELM Classic RealToxicityPromptstoxic_prompt_toxic_fraction	#33 / 40	↓0.1013	Source ↗official
SafetyPromptsscore	#8 / 15	↑59.89	Source ↗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-eval	Rank	Value	Source
HELM Classic RealToxicityPromptsnon_toxic_prompt_toxic_fraction	#6 / 40	↑0.02471	Source ↗official
HELM Classic RealToxicityPromptstoxic_prompt_toxic_fraction	#8 / 40	↑0.1013	Source ↗official
SafetyPromptsscore	#8 / 15	↓59.89	Source ↗official