Tag
提出 Population-Centred Evolution (PCE)，通过将整个 agent 种群的成功、失败与自修改历史蒸馏为“population block”来指导父代 harness 的改进，在 SWE-bench Verified 上达到 72.1%、Polyglot 上达到 96.7%，显著优于群体进化基线。
RSIHub 是一个开源框架，为智能体进化（agent-evolution）研究提供可组合的共享基础设施，标准化方法的构建、比较与扩展，并在 Terminal Bench 2 和 τ3 Banking 上评测了 AHE、Hyperagents、A-Evolve 和 GEPA 等方法。
Mara Chain 提出了一种新的 AI 系统自动进化框架，将被淘汰的候选方案及其失败分析保留下来迭代优化，而非直接丢弃。在 AppWorld、TerminalBench 2.1 和 MuSiQue 等基准上，以更少的 rollout 预算显著超越了 GEPA、AHE 等先前优化器，并在 GLM-5、DeepSeek、Qwen 等多种骨干模型上验证了通用性。
MOAE proposes a Pareto-preserving evolutionary search method to simultaneously optimize multiple objectives like task performance, trajectory quality, and safety for LLM agents without fixed scalarization during search.
This survey paper connects self-evolving LLM-based agents with dynamic graph transformation, proposing a framework to model agent state as dynamic graphs and organizing existing methods for their evolution.
This paper introduces Latent On-Policy Self-Distillation (LOPD), a method that makes the teacher's privileged context learnable end-to-end from experience, providing dense token-level supervision to enhance agent performance and efficiency in agentic tool use and code generation.
This paper proposes the Hybrid Open-Ended Tri-Evolution (HOTE) framework, which uses hybrid-mode reinforcement learning to evolve a proposer, solver, and judge collaboratively for deep research tasks, achieving state-of-the-art results with an 8B model surpassing larger static models.
MetaEvo proposes a two-stage framework for continual evolution of LLM-based agents, using preference-based optimization to enhance principle abstraction and modular architecture for experience reuse, outperforming strong baselines on reasoning benchmarks.
EvoMap introduces GEP (Genome Evolution Protocol), a network protocol enabling agents to convert successful strategies into genes and capsules for self-evolution, reducing repeated exploration.
awesome-autoresearch list updated, adding 6 application cases based on Karpathy's autoresearch pattern, covering scenarios such as customer service agent self-evolution, Shell integration, code configuration self-optimization, RAG tuning, and ASO.