Share the Judge, Learn the Deferral: Where Specialization Helps LLM Evaluation

• URL: http://arxiv.org/abs/2607.27984v1
• Date: Thu, 30 Jul 2026 10:29:22 GMT
• Title: Share the Judge, Learn the Deferral: Where Specialization Helps LLM Evaluation
• Abstract summary: We study 99,952 public, rubric-conditioned examples.<n>Supplying the correct rubric improves locked-test accuracy by 2.11 points over a response-only control.<n>On RewardBench 2, learned correctness heads route examples through a 0.6B-4B-8B cascade without changing any reward score.
• Score: 0.1269104766024433
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Agentic systems have widened the gap between producing candidate outputs and reviewing them. This paper asks a practical architectural question: should domain specialization be built into an evaluator's weights, or into the rule that decides when its judgment can be trusted? We study 99,952 public, rubric-conditioned examples. Supplying the correct rubric improves locked-test accuracy by 2.11 points over a response-only control; replacing it with an unrelated rubric costs 2.66 points. Dividing the same training corpus among eight criterion-family LoRA judges, however, loses 10.05 points and cuts audited coverage at a 5% risk target from 24.44% to 5.43%. Matching the bank's stored capacity with one rank-64 adapter does not reproduce this loss. Nor is the result explained by learning rate or optimizer steps. Initializing the family adapters from a shared, trained judge recovers test accuracy to 76.85%, 19.94 points above scratch training at the same learning rate (95% interval 18.88-21.02). The result changes when specialization governs deferral rather than judgment. On RewardBench 2, learned correctness heads route examples through a 0.6B-4B-8B cascade without changing any reward score. Across 20 locked repartitions, the cascade attains 89.40% accuracy, compared with 84.75% for 8B alone, at 0.415 normalized parameter compute. Every run passes an exact one-sided 95% risk audit; margin-based rules remain near 84.8% accuracy while using at least 0.94 compute. These results suggest a qualified design rule: share the learning of judgment until there is enough data to justify a split, and place domain-specific adaptation in an audited release boundary.

Related papers

• Abstention Errors and Segment Support in CUAD: An Auditable Contract-Extraction Case Study [0.0]
  This paper presents a retrospective, reproducible evaluation of a fixed public RoBERTa checkpoint on CUAD's published 102-contract test split.<n>Of 30,464 returned candidate strings, 19,562 occur on questions with no annotated answer.<n>The system answers 1,335 of 2,938 such questions, a 45.4% false-positive rate.
  arXiv  Detail & Related papers  (2026-09-07T21:26:53Z)
• A Judge Should Know What Changed:Construct Validity for LLM-as-a-Judge Evaluation [14.867869624585886]
  We formalize construct validity for an evaluator as a two-dimensional profile.<n>We show that S and R are independent and that no scalar summary preserves all relevant comparisons.<n>These results show that high judge agreement can coexist with weak sensitivity to changes in the construct being evaluated.
  arXiv  Detail & Related papers  (2026-08-25T11:32:54Z)
• RecurSE: Bounded Recursive Self-Evaluation for LLM Rubric Judges [48.831446371402414]
  RecurSE pairs a trainable judge with a synchronized policy-copy checker that audits the judge's reasoning against meta-rubrics to supply a scalar process reward.<n>We show that RecurSE achieves consistent generalization gains across held-out medical, pairwise, summarization, and professional benchmarks.
  arXiv  Detail & Related papers  (2026-08-25T08:35:04Z)
• No Judgment Without a Reason: Counterfactual Receipts for Versioned AI Evaluators [10.23869027501654]
  Evaluators often produce correct labels via flawed reasoning, a critical failure for agentic systems gating actions, routing reviews, or supplying training feedback.<n>We formalize evaluator reasoning accountability via three core sources: grounds, norms, and authority. Varying these sources yields an eight-cell counterfactual judgment cube to characterize judgment updates.<n>We show reason-aware evaluation must decouple prediction and certification, reporting transformation consistency alongside standard accuracy for trustworthy evaluator auditing.
  arXiv  Detail & Related papers  (2026-08-21T10:01:58Z)
• SVR: Self-Verifying Refinement via Joint Verdict-Confidence Reinforcement Learning for Adaptive Test-Time Compute [62.3458279176813]
  Self-Verifying Refinement is an oracle-free multi-turn reinforcement learning framework.<n>It learns to use self-verification as a compute-control policy.<n>It achieves a macro-average accuracy of 0.563 with only 2.99 inference turns on average.
  arXiv  Detail & Related papers  (2026-07-30T16:20:58Z)
• Rehearse: Stepping Back from the Confidence Cliff in Self-Improving Autoresearch [44.05969616406614]
  Autoresearch improves machine-learning code by proposing changes, running full training jobs, and keeping changes that improve the metric.<n>We study how the reliability of this pre-execution judgment changes over the course of an autoresearch trajectory.
  arXiv  Detail & Related papers  (2026-07-30T05:08:15Z)
• Theoria: Rewrite-Acceptability Verification over Informal Reasoning States [0.0]
  We present Theoria, a verification architecture that closes this gap.<n>A candidate solution is rewritten into a sequence of typed state transitions.<n>Every transition is independently auditable.
  arXiv  Detail & Related papers  (2026-07-01T17:56:42Z)
• Step-wise Rubric Rewards for LLM Reasoning [72.17879367869503]
  Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models.<n>We find 18.2% of steps in correct responses are wrong yet positively rewarded.<n>Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points.
  arXiv  Detail & Related papers  (2026-05-17T07:08:14Z)
• Distributional Energy-Based Models for Uncertainty-Aware Structured LLM Reasoning [40.342912574072024]
  Large Language Models produce structured outputs such as travel plans or code solutions.<n>Individual reasoning steps may appear correct while the output as a whole violates budgets, fails test cases, or contradicts earlier deductions.<n>We propose a learned quality scorer with deterministic analytical constraint penalties for verifying structured LLM outputs.
  arXiv  Detail & Related papers  (2026-05-15T17:08:27Z)
• Multi-Dimensional Behavioral Evaluation of Agentic Stock Prediction Systems Using Large Language Model Judges with Closed-Loop Reinforcement Learning Feedback [1.2362187555287152]
  Forecast evaluation in finance has relied on aggregate accuracy metrics and predictive-accuracy tests built on point-forecast errors.<n>We propose a behavioral forecast-evaluation methodology that complements accuracy tests by assessing the intermediate decision process itself.
  arXiv  Detail & Related papers  (2026-05-07T06:31:34Z)
• TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them [58.04324690859212]
  Large Language Models (LLMs) as automated evaluators (LLM-as-a-judge) has revealed critical inconsistencies in current evaluation frameworks.<n>We identify two fundamental types of inconsistencies: Score-Comparison Inconsistency and Pairwise Transitivity Inconsistency.<n>We propose TrustJudge, a probabilistic framework that addresses these limitations through two key innovations.
  arXiv  Detail & Related papers  (2025-09-25T13:04:29Z)
• ObjexMT: Objective Extraction and Metacognitive Calibration for LLM-as-a-Judge under Multi-Turn Jailbreaks [12.396822247035578]
  We present exMT, a benchmark for objective extraction and metacognition.<n>Given a multi-turn transcript, a model must output a one-sentence base objective and a self-reported confidence.<n> Accuracy is scored by similarity to gold objectives, then thresholded once on 300 calibration items.
  arXiv  Detail & Related papers  (2025-08-23T03:32:04Z)
• Tracing LLM Reasoning Processes with Strategic Games: A Framework for Planning, Revision, and Resource-Constrained Decision Making [38.75183725659772]
  Large language models (LLMs) are increasingly used for tasks that require complex reasoning.<n>We argue that measuring internal processes is essential for understanding model behavior and improving reliability.<n>We introduce a framework that evaluates LLMs along three core dimensions: planning, revision, and resource-constrained decision making.
  arXiv  Detail & Related papers  (2025-06-13T17:59:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.