---
title: "The Interaction Tax: When Communication Erases Diversity in Multi-Agent Teams"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.23541v1"
permalink: "https://feed.astralor.com/items/itm_9fc77a42cff048b5d9b5f527"
published: "2026-08-24T17:45:15.000Z"
score: 75
channel: "ai"
tags: ["LLM","Interaction Tax","Multi-agent Systems","Agent Collaboration"]
---

# The Interaction Tax: When Communication Erases Diversity in Multi-Agent Teams

## AI commentary

> This paper is a must-read for anyone building multi-agent systems. It uses a simple but profound experiment to puncture the intuition that more interaction equals better results. For engineers designing agent frameworks, the takeaway is direct: sample independently first, then decide whether to interact, and be cautious about letting agents read each other's full answers.

## At a glance

Does multi-agent LLM interaction help or hurt? This paper argues that contradictory findings stem from a missing distinction: not all multi-agent communication is equal. The authors show that different model families find structurally different solutions, but when agents read each other's complete outputs, their proposals converge within one round, erasing the diversity that motivates using multiple models—termed the interaction tax. Testing 11 verifier-scored optimization tasks under matched budgets, they find full-solution interaction is a weak default, while independent proposal generation avoids collapse. Critique helps only when the violated rule is easy for the LLM to find and fix. Performance depends less on the number of agents than on the information exchanged.

## Key points

- Introduces the 'interaction tax': full-solution interaction causes multi-agent proposals to converge within one round, erasing diversity.
- Across 11 verifier-scored tasks, full-solution interaction is a weak default under matched budgets compared to independent generation.
- Critique helps only when the violated rule is easy for the LLM to find and fix.
- Multi-agent performance depends less on the number of agents than on the type and timing of information exchanged.

## Full text

Does multi-agent LLM interaction help or hurt? Some work reports gains from debate (Du et al., 2024), critique loops (Chen et al., 2025), and mixture-of-agents synthesis (Wang et al., 2025), while other work finds that interaction adds cost without improving quality under equal budgets (Tran & Kiela, 2026; Xu et al., 2026; Jarrett et al., 2025), or that independent sampling already captures multi-agent gains (Li et al., 2024). We argue this contradiction partly reflects a missing distinction, because not all multi-agent communication is equal. Different model families find structurally different solutions, but when agents read each other's complete outputs, their proposals converge within one round, erasing the diversity that motivates using multiple models. We call this the interaction tax. We test 11 verifier-scored optimization tasks under matched budgets and find that full-solution interaction is a weak default. Independent proposal generation avoids this collapse. Full-solution interaction mainly makes agents stay close to the first solution they see instead of trying different approaches, and critique helps only if the violated rule is easy for the LLM to find and fix. These results suggest that multi-agent performance depends less on the number of agents than on the information they exchange, and interaction helps only when agents share the right information at the right time.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_9fc77a42cff048b5d9b5f527