AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization
We introduce AtomSearch, which keeps revision actions local while using subsequent skill states to guide search.
ProofPaper ↗
Key points
- Agent skills adapt frozen models through editable external instructions.
- Contrastive atomic editing links each primitive rule edit to failed and successful executions, a behavioral hypothesis, and an intended scope.
- Skill-State Tree Search (STS) evaluates short edit compositions and selectively retains them as executable states that can be revisited and extended.
- Outcome memory supplies later proposals with observed repair and regression statistics.
Sources (1)
- [1]AtomSearch: Evidence-Grounded Tree Search for Agent Skill OptimizationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 10:33 AM
We introduce AtomSearch, which keeps revision actions local while using subsequent skill states to guide search.
Agent skills adapt frozen models through editable external instructions.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 14, 2026MolHop: Scaffold-Hopping Molecular Design with Agentic Reinforcement Learning
- Oct 14, 2026Beyond the Trial-and-Error Loop: Hybrid Projection and Automated Tuning for Distributed Training
- Oct 13, 2026Agentic Self-Awareness: Aligning Knowing, Acting, and Holding
- Oct 13, 2026Discovering Undisclosed Concepts in LoRA Adapters for Diffusion-Based AIGC Models
- Oct 13, 2026AlignPred: Error-Controlled KV Cache Compression via Alignment and Predictive Coding
- Oct 13, 2026Risk-Averse Inference-Time Diffusion Model Alignment with Finite-Sample Guarantees