6arXiv cs.CL (Computation and Language)·
Double Ratchet: Co-evolving evaluation metrics and skills for self-improving LLM agents
A new arXiv preprint introduces Double Ratchet, a system that co-evolves both evaluation metrics and agent skills in settings where no reliable automatic verifier exists. The metric loop uses evolutionary search over small drawback detectors anchored to a small reference set, while the skill loop uses a lifecycle-managed approach; together they retain 88–110% of the performance lift achievable with ground-truth metrics across code generation (MBPP+), text-to-SQL (Spider 2.0-Snow), and report generation tasks. The paper also addresses safety, showing that anchor discipline and outer audits can catch and repair cases where evolved skills game the rubric. This work directly addresses a core bottleneck in self-improving agent systems: the chicken-and-egg problem of needing a reliable evaluator to improve.
Related guides (3)
Related events (8)
Rubrics on Trial: Self-evolving evaluation rubrics from a single query via synthetic pairwise evidence
A new arXiv preprint introduces 'Rubrics on Trial,' a query-only framework that automatically constructs and validates evaluation rubrics for LLMs without external annotations or model training. The system evolves rubric sets from scratch using synthetic rubric-conditioned response pairs, screening out non-discriminative, over-specific, or style-only rubrics. Evaluated across five preference benchmark suites, it achieves best average accuracy and leads on six of seven evaluation sets, addressing a key bottleneck in scalable LLM evaluation.
Biased LLM judges silently disable skill retirement in self-evolving agents
A new arXiv paper analyzes how biased LLM judges used as reward signals in self-evolving agent systems can silently disable the 'skill retirement' mechanism that prevents skill libraries from degrading below a no-skill baseline. The authors show that false-pass bias—where failures are incorrectly scored as passes—disables contribution-based retirement past a sharp threshold that cannot be overcome with more data, while leaving aggregate performance metrics unchanged, making the failure invisible. The paper proposes a defect-injection audit to determine pre-deployment whether a judge falls above or below the critical threshold.
LLMs judge worse than they generate: empirical challenge to self-evaluation pipeline assumptions
A new arXiv preprint tests the implicit assumption that LLM evaluation is easier than generation, using a controlled in-context QA setup across four benchmarks (SQuAD 2.0, DROP, HotpotQA, MuSiQue) and two models. Results show generation accuracy exceeds self-evaluation accuracy on three of four benchmarks, with attention analysis revealing that evaluation attends to context 3–5x less than generation does. LoRA fine-tuning experiments confirm the asymmetry is not a training artifact, with cross-task interference observed in both directions. The findings directly challenge assumptions underlying LLM-as-a-Judge and self-evaluation pipelines widely used in RLHF and agentic systems.
EvoPolicyGym benchmark evaluates autonomous policy evolution by LLM agents in RL environments
A new arXiv preprint introduces EvoPolicyGym, a benchmark for evaluating how LLM-based agents iteratively improve executable policies in compact interactive RL environments under a fixed interaction budget. The benchmark provides trajectory-level diagnostics beyond aggregate scores, distinguishing how agents allocate budget and convert feedback into parametric tuning. GPT-5.5 achieves the strongest aggregate rank score and top-two performance across all 16 environments. The work targets a gap in agent evaluation where iterative policy refinement is conflated with open-ended software engineering progress.
Continual-learning evaluation on Terminal-Bench 2.0 tests whether agent optimizer gains compound across tasks
A new arXiv paper introduces a two-phase continual-learning evaluation framework built on Terminal-Bench 2.0 to test whether agent-optimization gains persist and compound when new tasks arrive over time. Three agent-harness optimization methods — GEPA, Meta Harness, and RELAI-VCL — are compared under identical budgets; all improve in static single-phase settings but diverge sharply under continual optimization. RELAI-VCL is the only method that both transfers positively to unseen tasks and continues improving, reaching a 76.4% lifelong average pass rate versus 58.7% for the unoptimized baseline. The key finding is that compounding gains require regression control built into the optimization loop to prevent shortcut solutions.
LLM-as-a-Verifier: Training-free verification framework scales along granularity, repetition, and criteria decomposition
Researchers introduce LLM-as-a-Verifier, a general-purpose verification framework that treats verification as a new scaling axis for LLMs, computing continuous scores from token logit distributions rather than discrete judge outputs. The framework scales along three dimensions—score granularity, repeated evaluation, and criteria decomposition—and achieves state-of-the-art results on Terminal-Bench V2 (86.5%), SWE-Bench Verified (78.2%), RoboRewardBench (87.4%), and MedAgentBench (73.3%) without requiring additional training. The authors also demonstrate that the framework's fine-grained signals can serve as dense RL feedback, improving sample efficiency for SAC and GRPO on robotics and math benchmarks, and build a Claude Code extension for monitoring agentic systems.
Role-Agent: Bootstrapping LLM Agents via Dual-Role Evolution
Role-Agent is a new framework that uses a single LLM simultaneously as both agent and environment, enabling self-bootstrapped co-evolution without external environment feedback. The system has two components: World-In-Agent (WIA), which uses predicted vs. actual state alignment as a process reward, and Agent-In-World (AIW), which reshapes training data by retrieving tasks with similar failure patterns. Experiments across multiple benchmarks show an average performance gain of over 4% over strong baselines. The approach addresses key limitations in LLM agent training: inefficient feedback and static environments.
RePro: Retrospective Progress-Aware Self-Refinement for LLM Agent Training
Researchers introduce RePro (Retrospective Progress-Aware Training), a framework addressing the gap between step-wise RL optimization and metacognitive task-progress awareness in LLM agents. The approach uses a forward-then-reflect rollout paradigm where agents execute actions online and then retrospectively assess step-wise progress given the completed trajectory and known outcome. Evaluated on WebShop, ALFWorld, and Sokoban, RePro achieves up to 12% absolute success rate gains over baseline Qwen-family models without requiring continuous external supervision.