Backbone-Adaptive Evidence Routing for Robust Pairwise LLM Judging
Not on the map. Why?
Abstract (arXiv)
Pairwise language-model judges can gather evidence through direct comparison, reasoning, or reference-based verification, but no single protocol is best across benchmarks and judge backbones. We introduce Backbone-Adaptive Evidence Routing (BAER), which adapts the evidence mechanism while preserving candidate symmetry: swapping the two responses may reverse the preference but cannot change its strength. BAER separates each expert's signed preference from candidate-invariant reliability and builds three symmetric heads: evidence stacking, reliability-based expert routing, and candidate-blind reference verification. Development data select one head for each benchmark--backbone condition, and that choice is frozen before testing. Across four benchmarks and two 8B judge backbones, BAER achieves the highest test accuracy among the compared methods in all eight conditions, with full prediction coverage and gains of 0.87--7.32 points over the strongest external baseline. The results show that adapting how evidence is gathered is more reliable than fixing one judging protocol everywhere.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 21 of its 23 references that are in the map. First posted on arXiv 25 Sep 2026.
Builds on
21 of 23 references.
References · 21
- 2024The Llama 3 Herd of Modelscited by 14,192
- 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
- 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
- 2025Qwen3 Technical Reportcited by 8,195
- 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
- 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,634
- 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,205
- 2024Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preferencecited by 1,151
- 2024Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluatorscited by 865
- 2023Can Large Language Models Be an Alternative to Human Evaluations?cited by 729
- 2024RewardBench: Evaluating Reward Models for Language Modelingcited by 480
- 2023Large Language Models Are Not Robust Multiple Choice Selectorscited by 454
9 more
- 2023UltraFeedback: Boosting Language Models with Scaled AI Feedbackcited by 394
- 2024Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Modelscited by 301
- 2024JudgeBench: A Benchmark for Evaluating LLM-based Judgescited by 294
- 2023Evaluating Large Language Models at Evaluating Instruction Followingcited by 290
- 2021Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Controlcited by 256
- 2024Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judgecited by 249
- 2024CalibraEval: Calibrating Prediction Distribution to Mitigate Selection Bias in LLMs-as-Judgescited by 32
- 2026SCOPE: Selective Conformal Optimized Pairwise LLM Judgingcited by 12
- 2024The Comparative Trap: Pairwise Comparisons Amplifies Biased Preferences of LLM Evaluatorscited by 11
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judge
- 2026CAPS: Cascaded Adaptive Pairwise Selection for Efficient Parallel Reasoning
- 2026JudgeStealer: Extracting LLM Judging Capabilities across Evaluation Protocols
- 2026Judge, Retrieve, or Abstain: Uncertainty-Guarded LLM Judging with Provable Risk Guarantees
- 2026ODAR: Principled Adaptive Routing for LLM Reasoning via Active Inference
- 2026TARo: Token-level Adaptive Routing for LLM Test-time Alignment
- 2025Verdict: A Library for Scaling Judge-Time Compute
- 2025Confidence-Guided Stepwise Model Routing for Cost-Efficient Reasoning