JudgePanel: A Compact Judge with Panel Deliberation via Adaptive Multi-Reward Reinforcement Learning

arXiv:2608.29168v1 Announce Type: new Abstract: The LLM-as-a-Judge paradigm has emerged as a scalable alternative to human evaluation. However, single-model judges are limited by their inherent model biases, while multi-agent evaluation protocols that mitigate this through diverse deliberation are prohibitively expensive at inference time. To this end, we propose extbf{modelname}, which equips a compact nderline{Judge} model with multi-agent nderline{Panel} deliberation capability. Specifically, we first train on panel deliberation traces from an ensemble of strong evaluators, capturing structured patterns of discussion, disagreement, and resolution. To further improve judgment quality beyond SFT, we introduce extit{AdaReward}, an adaptive multi-reward RL algorithm that dynamically rebalances reward component weights as different objectives saturate at different rates during RL training. For practical deployment, we further design a lightweight domain specialization module for rapid adaptation to new evaluation domains with few hundred labeled samples. As a result, (i) extit{Novel}: the first framework to equip a single compact judge with multi-agent panel deliberation capability at single-model inference cost; (ii) extit{Effective & Reliable}: JudgePanel with a 14B backbone outperforms judge-specialized models up to 70B across four evaluation benchmarks, demonstrates strong position consistency, and rapidly specializes to new domains with few hundred samples.

Related

• Multi-Dimensional Behavioral Evaluation of Agentic Stock Prediction Systems Using Large Language Model Judges with Closed-Loop Reinforcement Learning Feedback
• Multi-Paradigm Agent Interaction in Practice:A Systematic Analysis of Generator-Evaluator, ReAct Loop,and Adversarial Evaluation in the buddyMe Framework
• Inducing Reward-Free Judging Rubrics that Reduce Over-Crediting in Agent Evaluation

Source: arXiv cs.AI | 2026-09-01