Towards Communication-Efficient Social Intelligence in Language Agents

2026-09-28

AI Summary

This study addresses the challenge faced by language agents in balancing goal achievement against communication costs during social interactions. To this end, we propose TACT, a Teacher-Assisted Communication Training framework. TACT introduces a novel online distillation mechanism that integrates action correction, partner response testing, and local goal-cost trade-offs. By combining policy and expression expert modules, it enables on-policy knowledge distillation, ensuring that the student model achieves both efficient expression and optimal decision-making upon independent deployment. Experimental results demonstrate that TACT significantly improves goal success rates on the SOTOPIA and AgentSense benchmarks while substantially reducing token consumption and interaction turns, thereby effectively enhancing the social interaction efficiency of language agents.

Abstract

Socially intelligent language agents must negotiate, coordinate, and resolve conflicting preferences while respecting the time and attention of both participants. Balancing these demands is challenging because agents must convey enough to address a partner's constraints and advance their goals without adding words that do not help the interaction. In this paper, we propose Teacher-Assisted Communication Training (TACT) to improve social goal attainment while reducing communication cost, making interactions with agents more productive and less demanding. We first characterize communication efficiency in terms of action strategy and expression, whose effects extend beyond the current utterance to the partner's response and subsequent exchanges. We design TACT to revise student-generated actions, test the revisions through partner responses, and distill useful feedback into the student. An expression specialist removes unnecessary detail while preserving the intended action, while a strategy specialist proposes alternatives that may better address the partner's constraints. To determine which revision helps, TACT samples a partner response for each candidate and selects a teacher reference by balancing local goal support against action-token cost. That reference guides on-policy distillation on the student's own generation prefixes, allowing the student to act independently at deployment. We evaluate TACT on SOTOPIA and AgentSense. On SOTOPIA, it achieves the highest Goal among the evaluated methods on All and Hard while using substantially fewer target tokens than SFT+SDPO. On AgentSense, it improves goal success over the initial student while reducing target tokens and interaction messages.

Problem

• Social Intelligence
• Language Agents
• Communication Efficiency
• Goal Attainment
• Communication Cost

Innovation

• Communication Efficiency
• Teacher-Assisted Communication Training
• On-policy Distillation
• Social Intelligence
• Language Agents