The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving Agents

• URL: http://arxiv.org/abs/2607.07436v1
• Date: Wed, 08 Jul 2026 14:08:04 GMT
• Title: The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving Agents
• Abstract summary: We show that a biased judge does not merely add noise; it emphsilently switches off the curator.<n>We make this precise with a corrupted-reward analysis and, isolating the causal channel by injecting corruption on top of a deterministic reward, a behavioral study on a reference-free report-writing testbed with a code-generation cross-check.
• Score: 9.989306175511238
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: A self-evolving agent retires its bad skills by watching them fail, so what happens when the judge cannot see the failures? Skill retirement is the structural constraint that keeps a growing library from drifting below the no-skill baseline, but its guarantee assumes an unbiased reward, which is false for the LLM judges that reference-free tasks force upon us. We show that a biased judge does not merely add noise; it \emph{silently switches off the curator}. We make this precise with a corrupted-reward analysis and, isolating the causal channel by injecting corruption on top of a deterministic reward, a behavioral study on a reference-free report-writing testbed with a code-generation cross-check. Symmetric noise leaves retirement intact, but \emph{false-pass} bias (failures slipping through as passes) disables contribution-based retirement past a sharp threshold that no amount of data can cross. Separating genuine retirement from cap-eviction churn shows this \emph{mechanism} failure is universal, holding across domains and failure rates and sparing only near-zero-false-pass, verifier-like graders. The downstream \emph{outcome}, though, is regime-dependent: eval quality degrades only where the same corruption also starves skill synthesis, and otherwise holds steady, so the disabled curator is \emph{silent}, surfacing in no aggregate metric. The contribution is a behavioral safety result, not a performance one. A cheap defect-injection audit then tells an operator, before deployment, which side of the threshold their judge occupies.

Related papers

• LLM-as-a-Judge Is Not an Oracle: Why Self-Improving Agents Need Deterministic Guardrails [0.0]
  Self-improving agent pipelines have a problem at their center.<n>An evaluation signal failed in eleven classes: judge bias, harness and metric failures, ground-truth errors, and reward hacking.<n>We describe PROCTOR, a Teacher-Student loop in which a stateful orchestrator holds all tool access, stateless subagents diagnose failures and draft mutations they cannot apply.
  arXiv  Detail & Related papers  (2026-09-02T07:54:23Z)
• 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)
• Judge, Retrieve, or Abstain: Uncertainty-Guarded LLM Judging with Provable Risk Guarantees [16.73004625486562]
  Using LLMs as judges has become standard practice for evaluating model outputs at scale.<n>In the absence of a reference answer, the judge evaluates factual correctness either through its parametric knowledge or through tool augmentation.<n>We propose a risk-controlled framework that calibrates uncertainty thresholds on a held-out set so that the false discovery rate among accepted verdicts remains below a user-specified level.
  arXiv  Detail & Related papers  (2026-08-18T16:42:02Z)
• Knowing When to Quit: Diagnosing and Training LLMs to Abort Futile Reasoning [109.33717747526583]
  Large language models generate computationally expensive yet semantically void reasoning on beyond-capability tasks.<n>We introduce textbfCaRL (textbfCapability-textbfaligned textbfReinforcement textbfL) to align behavior with capability boundaries.<n>Experiments demonstrate a substantial reduction in futile reasoning while preserving performance across task difficulties.
  arXiv  Detail & Related papers  (2026-07-31T09:30:33Z)
• Self-Authored Verification Is Unreliable in Heuristic Self-Improving Agents [5.91817208489443]
  We study the verifier-acceptance-deployment gap.<n>This gap refers to the discrepancy between an agent's self-authored verification signal and a sealed deployment evaluation.<n>We introduce a Sealed Exogenous Acceptance Loop (SEAL)
  arXiv  Detail & Related papers  (2026-07-27T11:45:14Z)
• Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents [9.989306175511238]
  Self-evolving agent systems improve by creating, revising, and retiring their own skills, but every such loop rests on a hidden assumption: a reliable evaluation metric already exists.<n>We make three claims. First, metrics can be emphevolved: our metric loop searches compositions of small detectors under a full evolutionary lifecycle.<n>Second, since no metric exists to beat, the yardstick is recovering what an accurate metric would have enabled, and emphDouble Ratchet, our co-evolution of the metric with a lifecycle-managed skill loop.
  arXiv  Detail & Related papers  (2026-07-14T14:02:50Z)
• Hypothesis-Driven Skill Optimization for LLM Agents [2.716362160018478]
  We propose a train-free framework in which both the skill curator and the agent executor are frozen inference endpoints.<n>On ALFWorld, HDSO improves executor-only baselines by +6.9 Avg. SR points for Qwen3-8B and +4.0 points for Qwen3.6-27B.
  arXiv  Detail & Related papers  (2026-06-21T04:43:20Z)
• LivingArena: Do LLMs Know What Other LLMs Don't? Peer-Probing as Scalable Evaluation [79.03892269184145]
  LivingArena is an automated, contamination-resistant evaluation framework.<n>Models take turns proposing questions, aiming to pose items that opponents cannot answer correctly.<n>Questioners are encouraged to actively identify and exploit opponents' knowledge boundaries.
  arXiv  Detail & Related papers  (2026-06-19T06:20:58Z)
• Signed Compression Progress on a Sealed Audit is Goodhart-Resistant [0.43830114853179497]
  Compression progress is a long-standing proposal for intrinsic motivation.<n>We prove that intrinsic reward is "credible" because it is paid only for learning.<n>We give a Lean 4 mechanization of the structural core.
  arXiv  Detail & Related papers  (2026-06-09T20:10:40Z)
• AI Alignment via Incentives and Correction [14.986111703734222]
  We study AI alignment through the lens of law-and-economics models of deterrence and enforcement.<n>A solver may benefit from producing a persuasive but incorrect answer, hiding uncertainty, or exploiting spurious shortcuts.<n>We formalize this interaction in a two-agent model in which a principal chooses rewards over joint correction outcomes.
  arXiv  Detail & Related papers  (2026-05-02T23:28:02Z)
• IMPACT-CYCLE: A Contract-Based Multi-Agent System for Claim-Level Supervisory Correction of Long-Video Semantic Memory [73.22944697933603]
  Existing pipelines produce opaque, end-to-end outputs that expose no intermediate state for inspection.<n>We present IMPACT-Cycle, a supervisory multi-agent system that reformulates long-video understanding as a multimodal iterative claim-level maintenance.
  arXiv  Detail & Related papers  (2026-04-22T03:03:33Z)
• A Coin Flip for Safety: LLM Judges Fail to Reliably Measure Adversarial Robustness [57.510025257780306]
  We show that existing validation protocols fail to account for substantial distribution shifts inherent to red-teaming.<n>We propose ReliableBench, a benchmark of behaviors that remain more consistently judgeable, and JudgeStressTest, a dataset designed to expose judge failures.
  arXiv  Detail & Related papers  (2026-02-04T15:13:35Z)
• Gaming the Judge: Unfaithful Chain-of-Thought Can Undermine Agent Evaluation [76.5533899503582]
  Large language models (LLMs) are increasingly used as judges to evaluate agent performance.<n>We show this paradigm implicitly assumes that the agent's chain-of-thought (CoT) reasoning faithfully reflects both its internal reasoning and the underlying environment state.<n>We demonstrate that manipulated reasoning alone can inflate false positive rates of state-of-the-art VLM judges by up to 90% across 800 trajectories spanning diverse web tasks.
  arXiv  Detail & Related papers  (2026-01-21T06:07:43Z)
• Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers [90.50039419576807]
  Reinforcement Learning with Verifiable Rewards (RLVR) trains policies against automated verifiers to avoid costly human labeling.<n>To reduce vulnerability to verifier hacking, many RLVR systems collapse rewards to binary $0,1$ during training.<n>This choice carries a cost: it introduces textitfalse negatives (rejecting correct answers, FNs) and textitfalse positives (accepting incorrect ones, FPs)
  arXiv  Detail & Related papers  (2025-10-01T13:56:44Z)

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.