RLTasksby ulam.ai
Models / OpenAI

OpenAI / MODEL EVIDENCE

GPT-3.5-turbo + CoT + EthicsPrompt

1 benchmarks · 4 observations · 1 sources

Scores at a glance

1 benchmarks · large dot: GPT-3.5-turbo + CoT + EthicsPrompt; small dots: closest models where reported

MACHIAVELLI ↗21.5 points
16.835.2

gpt-3.5-turbo + CoT + EthicsPrompt · MACHIAVELLI · Normalized reward · agent comparison ↗

3 more reported results

94 points · gpt-3.5-turbo + CoT + EthicsPrompt · MACHIAVELLI · Normalized power · agent comparison · Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark ↗

97 points · gpt-3.5-turbo + CoT + EthicsPrompt · MACHIAVELLI · Normalized disutility · agent comparison · Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark ↗

83 points · gpt-3.5-turbo + CoT + EthicsPrompt · MACHIAVELLI · Normalized ethical violations · agent comparison · Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark ↗

Closest models by score

Similar positions on shared benchmark leaderboards. Smaller gaps are closer.

6.3 percentile-point gap

1 of 1 shared benchmark comparisons

How similarity is measured

Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.

Filter comparisons to this model →

Sources

Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.

  1. Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark — published result table ↗

    Alexander Pan, Jun Shern Chan, Andy Zou, Nathaniel Li, Steven Basart, Thomas Woodside et al. · 2023

    All 10 authors

    Alexander Pan, Jun Shern Chan, Andy Zou, Nathaniel Li, Steven Basart, Thomas Woodside, Jonathan Ng, Hanlin Zhang, Scott Emmons, Dan Hendrycks

Names used in source tables

gpt-3.5-turbo + CoT + EthicsPrompt