CoWeaver: A Bi-directional, Learnable and Explainable Matching Engine for Mixed Human-Agent Science Collaboration
Multiagent SystemsArtificial IntelligenceComputation and Language
Summary
The authors created COWEAVER, a smart method to help scientists find the best collaborators by matching their skills and needs. It works by identifying gaps in abilities and ranking candidates in two steps, while also learning from feedback to improve its guesses about new people. Their tests showed that combining exploring new options and picking the best known choices works better or as well as just picking the known best. Overall, COWEAVER matched collaborators more accurately and efficiently than other methods they tried.
Authors
Jiayao Gu, Kexin Chu, Peidong Liu, Yue Yang, Lynn Ai, Qi Zhang, Ling Yang, Tianyu Shi
Abstract
LLM-based agents excel at writing articles, coding and information retrieval. However, they fail to form strong collaborations within the scientific community due to the bidirectional, dynamic nature of the problem and a high demand of decision interpretability. We proposed COWEAVER, a bidirectional, learnable and explainable algorithm to match scientists and form strong collaborations within a human-agent network. COWEAVER matches candidates and requesters through filling capability gaps and filters candidates through a two-stage ranking step. Finally, the model explores newcomers by maintaining uncertainty-aware capability estimates and updating them through requester's feedback. We show that the selection mechanism of combining both exploration (UCB) and greedy of COWEAVER exceeds the greedy-only mechanism - the analytical best solution - on 6 out of the 20 tasks and performed on par with the greedy-only mechanism in terms of selecting the best candidate. We compared COWEAVER baselines in terms of matching quality and efficiency. COWEAVER outperforms baselines on all metrics.