mlsanityData to 7 Oct 2026

Component-Aware Feedback for Self-Evolving Programs

arXiv 2609.38639 ↗First posted 29 Sep 2026cs.AI cs.IRcited by 0

On the map: Synthetic data and self-improvement for LLMs → Self-evolving LLM agents

Abstract (arXiv)

LLM-guided evolutionary search can discover complex programs, but existing methods mostly only save candidate programs and fitness scores while discarding which component edits produced which fitness metric changes. Existing methods force the mutator LLM to infer the effect of prior edits from cluttered histories, making program search slow and unstable. This is especially true for locally servable LLMs to evolve multi-component systems. We introduce component-aware feedback, which compares each evaluated program with its parent, identifies the components that changed, and logs them with the associated metric differences into an attribution memory that later mutations read. The memory keeps each change in two reference frames, local against the parent it came from and global against the seed program, which shows both the immediate effect of a change and the cumulative progress made since the seed. We study this on LLM reranking, a multi-objective optimization problem where a multi-stage pipeline must balance quality against serving cost. Across twelve textscBright datasets, our method reaches the strongest baseline's final quality after a median of one third of the search budget and ends 7.2% higher in held-out nDCG@10, and under a cost-aware objective it finds pipelines that are on average more accurate while using 11% fewer tokens per query, showing component-aware feedback to be a promising direction for more efficient self-evolving systems.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Self-evolving LLM agents, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 29 of its 37 references that are in the map. First posted on arXiv 29 Sep 2026.

Builds on

29 of 37 references.

References · 29

  1. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  2. 2025GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learningcited by 449
  3. 2025LLMs Get Lost In Multi-Turn Conversationcited by 400
  4. 2020Document Ranking with a Pretrained Sequence-to-Sequence Modelcited by 373
  5. 2023Is ChatGPT Good at Search? Investigating Large Language Models as Re-Ranking Agentscited by 356
  6. 2023Large Language Models are Effective Text Rankers with Pairwise Ranking Promptingcited by 302
  7. 2024Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programscited by 270
  8. 2024ReEvo: Large Language Models as Hyper-Heuristics with Reflective Evolutioncited by 266
  9. 2024Evolution of Heuristics: Towards Efficient Automatic Algorithm Design Using Large Language Modelcited by 258
  10. 2022Improving Passage Retrieval with Zero-Shot Question Generationcited by 164
  11. 2024BRIGHT: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrievalcited by 153
  12. 2023RankZephyr: Effective and Robust Zero-Shot Listwise Reranking is a Breeze!cited by 151
17 more
  1. 2025ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolutioncited by 146
  2. 2023Beyond Yes and No: Improving Zero-Shot LLM Rankers via Scoring Fine-Grained Relevance Labelscited by 86
  3. 2025Rank1: Test-Time Compute for Reranking in Information Retrievalcited by 68
  4. 2024Trace is the Next AutoDiff: Generative Optimization with Rich Feedback, Execution Traces, and LLMscited by 62
  5. 2025Rank-R1: Enhancing Reasoning in LLM-based Document Rerankers via Reinforcement Learningcited by 58
  6. 2024FIRST: Faster Improved Listwise Reranking with Single Token Decodingcited by 54
  7. 2023A Setwise Approach for Effective and Highly Efficient Zero-shot Ranking with Large Language Modelscited by 53
  8. 2024TourRank: Utilizing Large Language Models for Documents Ranking with a Tournament-Inspired Strategycited by 47
  9. 2024Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankerscited by 42
  10. 2025ReasonRank: Empowering Passage Ranking with Strong Reasoning Abilitycited by 39
  11. 2026CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discoverycited by 38
  12. 2025DIVER: A Multi-Stage Approach for Reasoning-intensive Information Retrievalcited by 35
  13. 2024JudgeRank: Leveraging Large Language Models for Reasoning-Intensive Rerankingcited by 34
  14. 2025REARANK: Reasoning Re-ranking Agent via Reinforcement Learningcited by 24
  15. 2026DeltaEvolve: Accelerating Scientific Discovery through Momentum-Driven Evolutioncited by 23
  16. 2025ERank: Fusing Supervised Fine-Tuning and Reinforcement Learning for Effective and Efficient Text Rerankingcited by 6
  17. 2026SwarmResearch: Orchestrating Coding Agents for Open-Ended Discoverycited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Evolution or Illusion? Rethinking Evaluation in LLM Evolutionary Search
  2. 2026SEDIMA: Cross-Run Hierarchical Insight Memory for Evolutionary Search Agents
  3. 2025Evolution without an Oracle: Driving Effective Evolution with LLM Judges
  4. 2025Self-Improving Language Models for Evolutionary Program Synthesis: A Case Study on ARC-AGI
  5. 2026What Do Evolutionary Coding Agents Evolve?
  6. 2026Deliberate Evolution: Agentic Reasoning for Sample-Efficient Symbolic Regression with LLMs
  7. 2026Teaching LLMs to Self-Evolve: Cultivating Core Meta-Skills with Reinforcement Learning
  8. 2026What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search