AI

Inherit-MAS: Test-Time Evolution of Multi-Agent Systems through Workflow and Execution Inheritance

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 08/10/2026, 21:18:24 UTC.

Briefings

Inherit-MAS improves multi-agent system performance by explicitly applying inheritance and selection principles at workflow and execution levels. Using GPT-4o-mini workers, the system achieved 55.4% completion on WorkBench and 49.7% joint F1 on HotpotQA FullWiki, outperforming existing evolving-MAS baselines. Execution inheritance reduces total token usage by 5.3% on WorkBench and 18.1% on HotpotQA compared to rerunning the same controller without execution inheritance.

Key points

• Inherit-MAS introduces explicit inheritance at workflow and execution levels using a meta-model that synthesizes worker agent workflows with declared roles and tool permissions.
• Execution inheritance only inherits stored results if the complete resolved request and execution context match, avoiding redundant model and tool calls during refinement rounds.

Context

Multi-agent systems coordinate specialized agents to tackle complex tasks, but effective workflows are difficult to design in advance. Test-time evolution refines workflows using execution feedback, yet broad revisions can disturb useful components, while re-executing unchanged requests can incur redundant computation. This study addresses these challenges through a meta-model inspired by biological evolution principles.

Limitations

Results are based on GPT-4o-mini workers on WorkBench and Qwen3-32B workers on HotpotQA FullWiki. Performance metrics include completion rates and joint F1 scores. The study does not provide causal claims about the specific mechanisms of inheritance or execution inheritance, nor does it include additional background context beyond the provided evidence.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 12/11/2026, 23:49:07 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeifjcbxijs4hylixczmnjamv7tfzmhzurarhukjxf6ojq463gfzaxq