Research map / Large language models / Alignment and preferences
UDA: Unsupervised Debiasing Alignment for Pair-wise LLM-as-a-Judge
Yang Zhang, Cunxiang Wang, Lindong Wu, Wenbo Yu, Yidong Wang, Guangsheng Bao, Jie Tang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace fixed K Elo with UDA adaptive K and soft labels | Elo judging · ArenaHard, 10 judges | avg inter-judge std dev: 158.5 -> 64.8 | Improved |
| apply UDA adapter zero-shot to unseen prompts | Elo judging · Human-Annotated Transfer Set | avg inter-judge std dev: 193.95 -> 70.97 | Improved |
| apply UDA | Elo judging · Human-Annotated Transfer Set | Pearson correlation with human: 0.651 -> 0.812 | Improved |
| add judge self-awareness features (judge's own answer embedding) (ablation) | UDA · Human-Annotated Transfer Set | Pearson correlation with human: 0.510 -> 0.812 | Improved |
| add judge self-awareness features (ablation) | UDA · Human-Annotated Transfer Set | avg inter-judge std dev: 65.64 -> 70.97 | Hurt |
| apply UDA to reduce self-preference | Elo judging · curated dataset, self-score deviation | self-score deviation from other judges: -21% to +56% -> -15% to +4% | Improved |
| pull judges toward consensus | UDA · well-calibrated judges | individual accuracy | Hurt |
Closest papers by meaning
- Heterogeneous Judge-Aware Ranking with Sensitivity, Disagreement, and Confidence (NeurIPS 2026 · similarity 0.90)
- From Uncertain Judgments to Calibrated Rankings: Conformal Elo Estimation for LLM Evaluation (NeurIPS 2026 · similarity 0.90)
- Any Large Language Model Can Be a Reliable Judge: Debiasing with a Reasoning-based Bias Detector (NeurIPS 2025 · similarity 0.90)
- Beyond the Surface: Enhancing LLM-as-a-Judge Alignment with Human via Internal Representations (NeurIPS 2025 · similarity 0.89)
- On Evaluating LLM Alignment by Evaluating LLMs as Judges (NeurIPS 2025 · similarity 0.88)
- Style Outweighs Substance: Failure Modes of LLM Judges in Alignment Benchmarking (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
