Add CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery

[BobbyZhouZijian]

Hi — thanks for maintaining this awesome list!

I'd like to suggest adding our recent paper, CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery, which we think fits the scope of this list (self-evolving / autonomous multi-agent systems for open-ended discovery).

• Paper: https://arxiv.org/abs/2604.01658
• Project page: https://human-agent-society.github.io/CORAL/
• Code: https://github.com/Human-Agent-Society/CORAL

TL;DR: CORAL is the first framework for autonomous multi-agent evolution on open-ended problems. It replaces fixed heuristics and hard-coded exploration rules with long-running agents that explore, reflect, and collaborate through shared persistent memory, asynchronous multi-agent execution, and heartbeat-based interventions. It sets new SOTA on 10 math / algorithmic / systems-optimization tasks, with 3–10× higher improvement rates than fixed evolutionary-search baselines — and on Anthropic's kernel-engineering task, four co-evolving agents improve the best-known score from 1363 → 1103 cycles.

BibTeX:

@article{coral2026,
  title  = {CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery},
  author = {Qu, Ao and Zheng, Han and Zhou, Zijian and Yan, Yihao and Tang, Yihong and Ong, Shao Yong and Hong, Fenglu and Zhou, Kaichen and Jiang, Chonghe and Kong, Minwei and Zhu, Jiacheng and Jiang, Xuan and Li, Sirui and Wu, Cathy and Low, Bryan Kian Hsiang and Zhao, Jinhua and Liang, Paul Pu},
  journal = {arXiv preprint arXiv:2604.01658},
  year   = {2026},
  url    = {https://arxiv.org/abs/2604.01658}
}

Happy to reformat the entry or move it to a different section if you'd prefer — just let me know. Thanks for considering!

[gemini-code-assist]

Code Review

This pull request adds the 'CORAL' paper to the list of multi-agent system papers in the README. Feedback suggests cross-listing this entry in the math exploration and memory management sections to improve discoverability and align with the repository's categorization practices.

[gemini-code-assist]

The paper's focus on shared persistent memory and its significant results on math and kernel-engineering tasks suggest it would be highly relevant for cross-listing in other sections of this repository. To improve discoverability, consider also adding this entry to:

• ### 💻 Math Exploration & Vibe Coding Agents -> #### Collective Multi-agent Reasoning (around line 834).
• #### Multi-agent Memory Management for Evolution (around line 742).

This would align with the repository's practice of listing papers in both methodological and application-specific categories.