ScaleMAS allows asynchronous LLM agents to interact through three operations.
Shared message search lets LLM agent groups scale without flooding readers
ScaleMAS coordinates asynchronous agents through searchable, linked messages and remains effective with as many as 128 agents.
Big Tech
Zhenting Qi · Zhuokai Zhao · Rulin Shao · Kerem Oktar · Zhiyuan Yao · Himabindu Lakkaraju · +4 more
FAIR at Meta · Harvard · Meta AI · Kempner Institute
Research Digest··3 min read
Thread:Agent Harness Optimization
Qi and colleagues replace fixed multi-agent communication structures with a shared datastore modeled loosely on web search and hyperlinks.
Why this paper
From Meta AI and 3 others · Part of Agent Harness Optimization, now 133 papers
In one line
Multi-agent LLM systems scale better when agents coordinate through a shared datastore with bounded reads and citation-based ranking.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§