Research map / Reinforcement learning / Visual and representation RL
ZetaEvolve: Learning to Search through History-Conditioned Potential Value
Jiyang Shen, Xiaojing Zhang, Bochen Lyu, Zhanxing Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes ZetaEvolve, a framework that models self-evolving LLM agents for mathematical discovery and code optimization through a partially observable search lens with history-conditioned potential values.
Problem
Self-evolving agents for mathematical discovery and code optimization face search spaces that are too large and unstructured for exhaustive exploration, and existing methods lack systematic modeling of self-evolution, leading to inefficient and redundant component designs.
Approach
Models self-evolution as a Partially Observable (search) process and learns to search via history-conditioned potential values, offering a systematic alternative to ad hoc component designs.
Closest papers by meaning
- Evolution Fine-Tuning: Learning to Discover Across 371 Optimization Tasks (NeurIPS 2026 · similarity 0.87)
- ThetaEvolve: Test-time Learning on Open Problems (ICML 2026 · similarity 0.86)
- Geometric Active Exploration in Markov Decision Processes: the Benefit of Abstraction (ICML 2024 · similarity 0.86)
- Improving Monte Carlo Tree Search for Symbolic Regression (NeurIPS 2025 · similarity 0.86)
- Simple Baselines are Competitive with Code Evolution (NeurIPS 2026 · similarity 0.86)
- SeeA*: Efficient Exploration-Enhanced A* Search by Selective Sampling (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
