zELO: Turning Pairwise Preferences into Absolute Scores for Rerankers
CShorten30 · x · 2026-11-13
On the Weaviate Podcast, Ghita Houir Alami explains zELO, a method for learning absolute scores from only sparse pairwise comparisons.
A Thurstone model converts pairwise preferences into Elo-style scores, and sparse sampling still provides enough signal to fit them. Those scores then train a pointwise reranker, powering zerank-1 and zerank-2.
Related event: ZeroEntropy co-founder explains zELO training behind zerank-2 reranker(2 posts)→
More from Research
- Ai2's XYBench Finds Five Top LLMs Fail to Correct Users' Misguided Questions — allen_ai · 2026-11-14
- Ai2 Asks Whether Prompting or Training Can Make LLMs Give More Useful Advice — allen_ai · 2026-11-14
- Ai2's XYBench Wins EMNLP Best Resource Award, Exposing LLMs' XY Problem Blind Spot — allen_ai · 2026-11-14
- G-STAR paper: 32–64x token compression gives 20x faster autoregressive image generation — _akhaliq · 2026-11-14
- OpenCLIP 4.2 ships with CLAP2 variable-length audio and NaFlexViT support — wightmanr · 2026-11-14
- VIALS benchmark exposes weaknesses of vision encoders and tool-use agents — AllThingsApx · 2026-11-14