AI
SailSearch: Agent-Assisted Discovery in Combinatorial Optimization
Publication: cs.AI updates on arXiv.org
Author: Dhriti Verma, Kanthi Sarpatwar, Vinicius Lima, Dzung Phan, Jayant Kalagnanam
Source published: 20/11/2026, 05:00:00 UTC
Briefing created: 20/11/2026, 05:30:07 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 20/11/2026, 05:03:24 UTC.
Briefings
SailSearch is a human-guided workflow using two language-model agents to assist with algorithmic research, specifically targeting Market Split and maximum independent set problems. The system successfully solved 130 to 133 of 156 benchmark instances for Market Split within 24 hours on two CPU cores, while for MIS, it found solutions 5 to 15 vertices below the optimum on seven instances with up to 3,996 vertices. The research was published on arXiv as a preprint.
Key points
- SailSearch uses two language-model agents to assist with algorithmic research, specifically targeting Market Split and maximum independent set problems.
- For Market Split, the system solved 130 to 133 of 156 benchmark instances within 24 hours on two CPU cores.
Context
This work presents a new human-guided workflow for algorithmic research involving two language-model agents, addressing open problems in combinatorial optimization.
Limitations
Results are limited to ten random seeds per instance for Market Split and seven instances for MIS. Solutions for MIS were found 5 to 15 vertices below the optimum on seven instances with up to 3,996 vertices.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 20/11/2026, 05:30:07 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeibk6nzwwp6ohhf6niwrea3ukljjjtqtjcncsdxvrxyblvqfyu5c6y