mlsanityData to 7 Oct 2026

Local Edits, Global Ripples: Replay-Informed Policy Adaptation for Workflow Synthesis

arXiv 2609.12127 ↗First posted 10 Sep 2026cs.CL cs.SEcited by 0

Not on the map. Why?

Abstract (arXiv)

Prompt-policy editing offers a practical way to improve agents that synthesize executable workflows without updating the underlying model. However, persistent prompt editing has two coupled properties. First, edit locality does not imply effect locality: an edit confined to one policy segment can ripple through downstream execution, altering behavior beyond the edited segment. Second, edit effects are composition-sensitive: edits that work in isolation can interfere after composition, causing one or both to lose their benefit or become harmful. Persistent adaptation must therefore support two distinct decisions: identifying where the policy should change from execution feedback, and determining whether the resulting edit remains safe to persist after composition. To address these challenges, we introduce RIPPLE (Replay-Informed Persistent Policy Localization and Editing), which separates where an edit is made from whether it remains safe after composition. It diagnoses failed trajectories, maps each actionable failure to a predefined policy segment, and restricts the correction to that part of the policy. RIPPLE then evaluates candidates against the same iteration-start policy to compare their isolated gains, before replaying promising edits after previously accepted updates to expose downstream effects and interactions. Only edits that remain safe under composition are retained. We evaluate RIPPLE on Flow-HO, a synthetic held-out benchmark for executable workflow synthesis. RIPPLE improves validation success by up to 23.1% and yields positive gains on two additional frozen language-model backbones, while maintaining edit efficiency and low execution cost. Targeted interaction analysis further demonstrates both properties: a segment-local tool-use edit changes downstream resource resolution and validation, while an edit beneficial in isolation becomes harmful after composition.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 40 of its 42 references that are in the map. First posted on arXiv 10 Sep 2026.

Builds on

40 of 42 references.

References · 40

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  3. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  4. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  5. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  6. 2020Conservative Q-Learning for Offline Reinforcement Learningcited by 1,777
  7. 2024SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineeringcited by 1,628
  8. 2024τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domainscited by 1,280
  9. 2022Large Language Models Are Human-Level Prompt Engineerscited by 1,020
  10. 2024Tulu 3: Pushing Frontiers in Open Language Model Post-Trainingcited by 988
  11. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 965
  12. 2023Teaching Large Language Models to Self-Debugcited by 963
28 more
  1. 2023Large Language Models as Optimizerscited by 830
  2. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 814
  3. 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 803
  4. 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 778
  5. 2023CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquingcited by 751
  6. 2023Automatic Prompt Optimization with "Gradient Descent" and Beam Searchcited by 628
  7. 2024Executable Code Actions Elicit Better LLM Agentscited by 537
  8. 2023Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolutioncited by 521
  9. 2024Generative Verifiers: Reward Modeling as Next-Token Predictioncited by 465
  10. 2025GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learningcited by 449
  11. 2024AFlow: Automating Agentic Workflow Generationcited by 395
  12. 2024Automated Design of Agentic Systemscited by 360
  13. 2024Agent Workflow Memorycited by 342
  14. 2025rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinkingcited by 308
  15. 2024Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programscited by 270
  16. 2024TextGrad: Automatic "Differentiation" via Textcited by 250
  17. 2023Evaluating the Ripple Effects of Knowledge Editing in Language Modelscited by 238
  18. 2023PromptAgent: Strategic Planning with Language Models Enables Expert-level Prompt Optimizationcited by 226
  19. 2024V-STaR: Training Verifiers for Self-Taught Reasonerscited by 214
  20. 2023EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizerscited by 189
  21. 2026SkillOpt: Executive Strategy for Self-Evolving Agent Skillscited by 122
  22. 2024Symbolic Learning Enables Self-Evolving Agentscited by 104
  23. 2024AutoGuide: Automated Generation and Selection of Context-Aware Guidelines for Large Language Model Agentscited by 69
  24. 2024Trace is the Next AutoDiff: Generative Optimization with Rich Feedback, Execution Traces, and LLMscited by 62
  25. 2025Self-Supervised Prompt Optimizationcited by 37
  26. 2026SkillGen: Verified Inference-Time Agent Skill Synthesiscited by 29
  27. 2026GRASP: Gated Regression-Aware Skill Proposer for Self-Improving LLM Agentscited by 10
  28. 2026PACE: Anytime-Valid Acceptance Tests for Self-Evolving Agentscited by 8

Built on by

Not cited yet.