Relic: From Multi-Agent Collaboration to Persistent Organizational Capability
Relic turns multi-agent failures into executable protocols, raising CooperBench delivery to 76.9%.
Relic converts recurring multi-agent collaboration failures into organization-owned executable protocols that bind triggers, responsibilities, evidence, and runtime consequences, while remaining open to revision. Across 360 controlled runs over ten software workloads and three models, complete-contract delivery rose from 14.06% to 19.76%. Under fresh-member transfer, behavioral correctness was 41.2% with executable bindings versus 34.6% with the same rules as text and 25.4% with no inherited protocol. On CooperBench, after excluding broken pairs, Relic reached 367/477 (76.9%).
- Relic turns recurring collaboration failures into revisable, executable team protocols.
- Complete-contract delivery rose from 14.06% to 19.76% across 360 runs.
- Executable protocol bindings beat readable text under fresh-member transfer, 41.2% vs 34.6%.
- On filtered CooperBench, Relic scored 367/477 (76.9%), the best reported peer-structured result.
- Products
- CooperBench
From the original article
1 min read at huggingface.co
Multiple agents may often conflict in an organization: for example, one coding agent changes an interface in a repository, but another continues to develop on the old version where existing tests become stale. A conversation can resolve the episode, but when the participants change, what makes the lesson continue to govern the team?
By Hongyi Du, Tianyi Zhang, Weijia Zhang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.