Delaying early consensus: ArcticSwarm hits 82.6% on BrowseComp-Plus with gated agent isolation
rohanpaul_ai · x · 2026-09-08
The arXiv paper ArcticSwarm tackles a key failure mode in long-horizon multi-agent research: when agents share partial findings too early, one plausible answer attracts all the search effort before alternatives are tested.
Key design:
- Separating evidence gathering from integration: subagents publish to a shared bulletin board instead of syncing in real time
- Gated isolation: selected search tasks keep their own prior, blocking premature consensus
- Structured review at three commitment boundaries: only confident candidates propagate
Results: 82.6% on the full BrowseComp-Plus with open-weight Qwen 3.5-27B (78.8% without gated isolation, 74.5% without structured review, vs. MiroFlow baseline at 70.6%). On live-web BrowseComp with GPT-5 it reaches 73.6%, well above the reported provider system (54.9%) and MiroFlow (63.4%).
Related event: Delayed Consensus Boosts Multi-Agent Research Performance(2 posts)→
More from coding & agent
- Grok Build ships triple daily updates: first-party MCP server, persistent subagents push toward full agent workspace — elonmusk · 2026-09-08
- A 'block first, generate second' AI video tool seeks feedback on its 3D pre-vis workflow — KeyCod3923 · 2026-09-08
- Mastra launches remote filesystem support for agents across S3, GCS, Azure and more — glcst · 2026-09-08
- How GPT-6 Astra's computer use loop powers its viral Blender 3D world generation — iamrobotbear · 2026-09-08
- Open-source Google Workspace CLI lets AI run your Gmail, Docs and Calendar — Aizkmusic · 2026-09-08
- Dev uses GPT-6 Astra to run local Windows AAA games like Skyrim on an iPad mini — ammaar · 2026-09-08