mlsanityData to 7 Oct 2026

After the Fix: Transfer of Corrected Agent Experience

arXiv 2609.34603 ↗First posted 28 Sep 2026cs.AI cs.CL cs.SEcited by 0

Not on the map. Why?

Abstract (arXiv)

Does repairing an episode make its experience a better memory for the next task? We transfer the same failed source before and after accepted repair to a fixed target, alongside independent execution. Our 3,300 runs cover 100 ThinkingBox pairs and the same 100 APEX pairs with and without source-state inheritance, under eleven conditions. ThinkingBox's Full/Skill/Hybrid correction gains are 44/29/32 percentage points, with corrected performance 25/22/18 points above independence; inference weakens at the task-family level. Yet 12 of Full's 15-point larger correction gap over Skill come from worse uncorrected performance, not better corrected memory. Moreover, 22 of Full's 46 upward transitions restore observed baseline success. Neither APEX regime establishes comparable aggregate correction benefits. Action evidence connects workflow gains with reusable obligations and convention conflicts with source-local choices. Text APEX's accepted execution reaches 52% versus its summary's 40%, without robust global/group-level superiority or an estab- lished advantage over independence. Smaller handoffs reduce input but increase calls. The value of repairing experience is therefore distinct from the value of reusing it: memory updates require both a previous-version reference and a fresh-start reference.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 20 of its 21 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

20 of 21 references.

References · 20

  1. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  2. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  3. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  4. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 965
  5. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 814
  6. 2025LLMs Get Lost In Multi-Turn Conversationcited by 400
  7. 2024Agent Workflow Memorycited by 342
  8. 2026SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Taskscited by 289
  9. 2023LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Modelscited by 279
  10. 2025ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memorycited by 253
  11. 2026Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skillscited by 161
  12. 2023CLIN: A Continually Learning Language Agent for Rapid Task Adaptation and Generalizationcited by 88
8 more
  1. 2026MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Taskscited by 72
  2. 2025The Complexity Trap: Simple Observation Masking Is as Efficient as LLM Summarization for Agent Context Managementcited by 40
  3. 2026APEX-Agentscited by 29
  4. 2026Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?cited by 19
  5. 2025How Reliable is Language Model Micro-Benchmarking?cited by 8
  6. 2026One Success Isn't Reliability: Thinkingbox, a Sandbox and Benchmark for Agents in Stateful Business Workflowscited by 2
  7. 2026Break It Down, Pass It On: Cross-Task Skill Transfer in LLM Agentscited by 1
  8. 2026Trace as State: Reasoning Traces as Conditional States for Long-Context Transformerscited by 1

Built on by

Not cited yet.