# Combined Win Rate Evaluation on PKU-SafeRLHF Random prompts n = 100 samples Source: [Combined Win Rate Evaluation on PKU-SafeRLHF Random prompts n = 100 samples]() Collection index: [llms.txt]() Data updated: 2026-05-12T14:02:24.561Z Evaluation of model performance using the Combined Win Rate metric, specifically calculated via CVaR(0.125), applied to a small, randomly sampled subset (n=100) of the PKU-SafeRLHF dataset. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Combined win rate evaluation | | Dataset (including any stated split) | PKU-SafeRLHF Random prompts n = 100 samples | | Metric | CVaR(0.125) Combined Win Rate | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 10 of 10 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [CVaR(0.125) Combined Win Rate](): Higher is better; unit: PERCENT. ## Recorded results | Method | CVaR(0.125) Combined Win Rate | Result date | Report paper | | --- | --- | --- | --- | | IPO Entropic τ=10 | 37.1 | Not recorded | 6a02b491dbc836b87d824d17 | | EG Entropic τ=5 | 36.8 | Not recorded | 6a02b491dbc836b87d824d17 | | IPO CVaR α=0.25 | 36.8 | Not recorded | 6a02b491dbc836b87d824d17 | | IPO Entropic τ=5 | 30.8 | Not recorded | 6a02b491dbc836b87d824d17 | | IPO CVaR α=0.125 | 30.8 | Not recorded | 6a02b491dbc836b87d824d17 | | EG k = 1 Neutral | 27.9 | Not recorded | 6a02b491dbc836b87d824d17 | | IPO k = 1 Neutral | 27.1 | Not recorded | 6a02b491dbc836b87d824d17 | | NMD k = 1 Neutral | 22.9 | Not recorded | 6a02b491dbc836b87d824d17 | | EGPO† (HF) | 4.8 | Not recorded | 6a02b491dbc836b87d824d17 | | SFT Base | 2.5 | Not recorded | 6a02b491dbc836b87d824d17 | ### IPO Entropic τ=10 CVaR(0.125) Combined Win Rate: 37.1. Result ID: 6a03328b47f34ce924e7dafe. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"tau (τ)":"10","objective":"Entropic"}. This record does not assert independent reproduction. ### EG Entropic τ=5 CVaR(0.125) Combined Win Rate: 36.8. Result ID: 6a03328b47f34ce924e7daff. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"tau (τ)":"5","objective":"Entropic"}. This record does not assert independent reproduction. ### IPO CVaR α=0.25 CVaR(0.125) Combined Win Rate: 36.8. Result ID: 6a03328b47f34ce924e7db00. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"alpha (α)":"0.25","objective":"CVaR"}. This record does not assert independent reproduction. ### IPO Entropic τ=5 CVaR(0.125) Combined Win Rate: 30.8. Result ID: 6a03328b47f34ce924e7dafd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"tau (τ)":"5","objective":"Entropic"}. This record does not assert independent reproduction. ### IPO CVaR α=0.125 CVaR(0.125) Combined Win Rate: 30.8. Result ID: 6a03328b47f34ce924e7db01. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"alpha (α)":"0.125","objective":"CVaR"}. This record does not assert independent reproduction. ### EG k = 1 Neutral CVaR(0.125) Combined Win Rate: 27.9. Result ID: 6a03328b47f34ce924e7dafc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"k":"1","objective":"Neutral"}. This record does not assert independent reproduction. ### IPO k = 1 Neutral CVaR(0.125) Combined Win Rate: 27.1. Result ID: 6a03328b47f34ce924e7dafa. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"k":"1","objective":"Neutral"}. This record does not assert independent reproduction. ### NMD k = 1 Neutral CVaR(0.125) Combined Win Rate: 22.9. Result ID: 6a03328b47f34ce924e7dafb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"k":"1","objective":"Neutral"}. This record does not assert independent reproduction. ### EGPO† (HF) CVaR(0.125) Combined Win Rate: 4.8. Result ID: 6a03328b47f34ce924e7daf9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"source":"Hugging Face"}. This record does not assert independent reproduction. ### SFT Base CVaR(0.125) Combined Win Rate: 2.5. Result ID: 6a03328b47f34ce924e7daf8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"type":"Baseline"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.