# ConflictGuide: AutoResearch Improves When Competing Behaviors Are Made Visible

arXiv: https://arxiv.org/abs/2609.39933 · HTML: https://mlsanity.com/paper/2609.39933/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 54 references that are in the map.
- First posted on arXiv 30 Sep 2026.

## Builds on (most cited first, 18 of 18)

- [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://mlsanity.com/paper/2506.13131/) (2025)
- [Factorized Fourier Neural Operators](https://mlsanity.com/paper/2111.13802/) (2021)
- [AIDE: AI-Driven Exploration in the Space of Code](https://mlsanity.com/paper/2502.13138/) (2025)
- [Toward a Better Understanding of Fourier Neural Operators from a Spectral Perspective](https://mlsanity.com/paper/2404.07200/) (2024)
- [AutoSOTA: An End-to-End Automated Research System for State-of-the-Art AI Model Discovery](https://mlsanity.com/paper/2604.05550/) (2026)
- [AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces](https://mlsanity.com/paper/2608.23041/) (2026)
- [Self-EvolveRec: Self-Evolving Recommender Systems with LLM-based Directional Feedback](https://mlsanity.com/paper/2602.12612/) (2026)
- [AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents](https://mlsanity.com/paper/2607.18754/) (2026)
- [NOVA: A Verification-Aware Agent Harness for Architecture Evolution in Industrial Recommender Systems](https://mlsanity.com/paper/2606.27243/) (2026)
- [ASI-Evolve: AI Accelerates AI](https://mlsanity.com/paper/2603.29640/) (2026)
- [Automated Researchers Can Mitigate Well-characterized Alignment Failures](https://mlsanity.com/paper/2608.28945/) (2026)
- [Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents](https://mlsanity.com/paper/2607.12790/) (2026)
- [EvoPINN: Agentic Discovery of Executable Algorithms for Physics-Informed Neural Networks](https://mlsanity.com/paper/2607.26490/) (2026)
- [Mechanist: AI as a Scientific Instrument for Discovering the Mechanisms of Intelligence](https://mlsanity.com/paper/2608.12036/) (2026)
- [LLaMEA-SAGE: Guiding Automated Algorithm Design with Structural Feedback from Explainable AI](https://mlsanity.com/paper/2601.21511/) (2026)
- [Physics-Audited Agentic Discovery in Scientific Machine Learning](https://mlsanity.com/paper/2607.07379/) (2026)
- [GoalEvolve: From Handcrafted Algorithm Priors to Goal-Driven Evolution of Physical Design Algorithms](https://mlsanity.com/paper/2608.16733/) (2026)
- [AgentFold: Closed-Loop Agentic Search for Protein Folding Model Design](https://mlsanity.com/paper/2608.26747/) (2026)

## Abstract (arXiv)

When designing machine learning models, desirable properties are often in tension: improving one behavior can impair another, so task progress can depend on alleviating the conflict. LLM-based AutoResearch systems, which iteratively edit model code and retain edits based on scalar task-performance feedback, have largely ignored this trade-off. We find that scalar feedback supports broad exploration early in search, but it does not reveal how edits affect competing behaviors. In matched-budget experiments, introducing competing-behavior feedback as task gains diminish increases the share of proposals that improve both behaviors and sustains progress beyond scalar-only plateaus. Obtaining this feedback for a given model requires identifying its competing behaviors and designing probes to measure them. To make competing-behavior feedback actionable, we introduce ConflictGuide. Its reusable ConflictGuide-Skill combines a literature-grounded taxonomy with model-specific evidence to identify competing behaviors and specify probes for a code agent to implement as metrics. Evolution proceeds in two stages: Stage I explores with task feedback; Stage II uses probe feedback to steer proposals toward conflict alleviation and retains marginal-gain edits only when probes indicate sufficient alleviation. Across five diverse model families, ConflictGuide reduces task and conflict-related errors by up to 28% and 14%, respectively, relative to scalar-only AutoResearch, with gains extending to other code agents.