Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

OSC: Cognitive Orchestration through Dynamic Knowledge Alignment in Multi-Agent LLM Collaboration

Jusheng Zhang, Yijia Fan, Kaitong Cai, Xiaofei Sun, Keze Wang

TL;DR

Experiments on complex reasoning and problem-solving benchmarks demonstrate that OSC significantly improves task performance and communication efficiency, transforming parallel-working individuals into a deeply collaborative cognitive team.

Abstract

This paper introduces OSC (Orchestrating Cognitive Synergy), a knowledge-aware adaptive collaboration framework designed to enhance cognitive synergy in multi-agent systems with large language models. While prior work has advanced agent selection and result aggregation, efficient linguistic interactions for deep collaboration among expert agents remain a critical bottleneck. OSC addresses this gap as a pivotal intermediate layer between selection and aggregation, introducing Collaborator Knowledge Models (CKM) to enable each agent to dynamically perceive its collaborators' cognitive states. Through real-time cognitive gap analysis, agents adaptively adjust communication behaviors, including content focus, detail level, and expression style, using learned strategies. Experiments on complex reasoning and problem-solving benchmarks demonstrate that OSC significantly improves task performance and communication efficiency, transforming "parallel-working individuals'' into a "deeply collaborative cognitive team.'' This framework not only optimizes multi-agent collaboration but also offers new insights into LLM agent interaction behaviors.

OSC: Cognitive Orchestration through Dynamic Knowledge Alignment in Multi-Agent LLM Collaboration

TL;DR

Experiments on complex reasoning and problem-solving benchmarks demonstrate that OSC significantly improves task performance and communication efficiency, transforming parallel-working individuals into a deeply collaborative cognitive team.

Abstract

This paper introduces OSC (Orchestrating Cognitive Synergy), a knowledge-aware adaptive collaboration framework designed to enhance cognitive synergy in multi-agent systems with large language models. While prior work has advanced agent selection and result aggregation, efficient linguistic interactions for deep collaboration among expert agents remain a critical bottleneck. OSC addresses this gap as a pivotal intermediate layer between selection and aggregation, introducing Collaborator Knowledge Models (CKM) to enable each agent to dynamically perceive its collaborators' cognitive states. Through real-time cognitive gap analysis, agents adaptively adjust communication behaviors, including content focus, detail level, and expression style, using learned strategies. Experiments on complex reasoning and problem-solving benchmarks demonstrate that OSC significantly improves task performance and communication efficiency, transforming "parallel-working individuals'' into a "deeply collaborative cognitive team.'' This framework not only optimizes multi-agent collaboration but also offers new insights into LLM agent interaction behaviors.

Paper Structure

This paper contains 45 sections, 7 equations, 4 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. LLM-Driven Multi-Agent Systems
  4. Agent Selection and Result Aggregation
  5. Inter-Agent Communication
  6. Method
  7. OSC Framework
  8. Dynamic Collaborator Knowledge Model (CKM)
  9. Learned Cognitive Gap Analysis and Adaptive Communication Objectives
  10. Adaptive Communication Strategy pi_comm
  11. Strategically Guided Linguistic Realization
  12. Reinforcement Learning Optimization
  13. Experiment
  14. Main Results and Analysis
  15. Experimental Setup
  16. ...and 30 more sections

Figures (4)

  • Figure 1: The OSC framework uses Collaborator Knowledge Models (CKMs)
  • Figure 2: Price-performance trade-off on AlpacaEval 2.0. OSC configurations (hexagons) are compared against KABB (Full) (circle), individual single-models (triangles), and proprietary models (stars). OSC demonstrates a strong Pareto frontier, optimizing performance relative to cost. The dashed line connects OSC configurations, highlighting improved performance with increasing, yet efficiently managed, expert collaboration.
  • Figure 3: A comparison showing that fine-tuning the CKM and $f_{gap}$ modules improves task success (LC Win Rate) and communication efficiency (Avg. Rounds and Tokens) over a pretraining-only approach and the KABB baseline.
  • Figure 4: Hyperparameter tuning for communication rounds ($N_{round}$) on AlpacaEval 2.0 shows that $N_{round}=4$ achieves the optimal balance between task success (LC Win Rate) and communication cost (Avg. Tokens).