# Calibrating Criterion Revision in LLM Agents: Failure Modes and a Trace-Anchored Protocol

arXiv: https://arxiv.org/abs/2608.20729 · HTML: https://mlsanity.com/paper/2608.20729/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 15 of its 38 references that are in the map.
- First posted on arXiv 21 Aug 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 15 of 15)

- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning](https://mlsanity.com/paper/2508.19828/) (2025)
- [Dynamic Cheatsheet: Test-Time Learning with Adaptive Memory](https://mlsanity.com/paper/2504.07952/) (2025)
- [Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory](https://mlsanity.com/paper/2511.20857/) (2025)
- [Measuring what Matters: Construct Validity in Large Language Model Benchmarks](https://mlsanity.com/paper/2511.04703/) (2025)
- [How Memory Management Impacts LLM Agents: An Empirical Study of Experience-Following Behavior](https://mlsanity.com/paper/2505.16067/) (2025)
- [Agentic Memory: Learning Unified Long-Term and Short-Term Memory Management for Large Language Model Agents](https://mlsanity.com/paper/2601.01885/) (2026)
- [Mem2ActBench: A Benchmark for Evaluating Long-Term Memory Utilization in Task-Oriented Autonomous Agents](https://mlsanity.com/paper/2601.19935/) (2026)
- [EvoMemBench: Benchmarking Agent Memory from a Self-Evolving Perspective](https://mlsanity.com/paper/2605.18421/) (2026)
- [Task Matters: Knowledge Requirements Shape LLM Responses to Context-Memory Conflict](https://mlsanity.com/paper/2506.06485/) (2025)
- [UMEM: Unified Memory Extraction and Management Framework for Generalizable Memory](https://mlsanity.com/paper/2602.10652/) (2026)
- [Causal Intervention-Based Memory Selection for Long-Horizon LLM Agents](https://mlsanity.com/paper/2605.17641/) (2026)
- [ImplicitMemBench: Measuring Unconscious Behavioral Adaptation in Large Language Models](https://mlsanity.com/paper/2604.08064/) (2026)

## Abstract (arXiv)

Language-model agents can improve after failure or carry text across episodes without revising what counts as success. We study the narrower attribution problem of criterion revision: when criterion K0 accepts an outcome violating a broader commitment B, what observations justify saying that the system formed and persistently used K1? We require five non-compensatory conditions: criterion-failure detection, a model-emitted proposal, new-episode transfer, intervention sensitivity on the claimed carrier, and preservation. We evaluate CMB-0.1 on twelve cross-domain cases and four arms: stateless inference, append-only history, model-generated but harness-committed state, and evaluator-written oracle state. Seven mechanism fixtures yield 84 deterministic scorer trials; four local quantized artifacts yield 96 calls and 192 model-case-arm trials. No model trial satisfies all five conditions, but this zero does not establish general capability absence. Eleven calls remain invalid after one retry; several commitments disclose the target distinction; the harness performs commits; deletion reuses a stateless call; and conflict changes multiple factors. Qwen2.5-7B answers every transfer and preservation item without revision state, exposing zero-state reconstruction. These failures make CMB-0.1 an instrument-calibration result rather than a model ranking. We derive a prospective, trace-anchored CMB-0.4 protocol requiring concealed transfer, explicit WRITE/NO-WRITE/ESCALATE actions, a separately logged policy-selected commit, matched interventions, repeated hidden items, and a frozen executable oracle. It is a successor design, not a completed confirmatory result. The paper contributes a measurement chain, an empirical diagnosis of its first implementation, and a more discriminating protocol for future tests of criterion revision.