# The Collaboration Tax: How Much LLM Multi-Agent Systems Pay to Coordinate

arXiv: https://arxiv.org/abs/2608.22152 · HTML: https://mlsanity.com/paper/2608.22152/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Social simulation and multi-agent debate, one of 16 directions in LLM agents and tool use.
- Builds on 19 of its 40 references that are in the map.
- First posted on arXiv 23 Aug 2026.

## Placed in

- LLM agents and tool use / [Social simulation and multi-agent debate](https://mlsanity.com/agents/social-simulation-and-multi-agent-debate/)

## Builds on (most cited first, 19 of 19)

- [Generative Agents: Interactive Simulacra of Human Behavior](https://mlsanity.com/paper/2304.03442/) (2023)
- [Improving Factuality and Reasoning in Language Models through Multiagent Debate](https://mlsanity.com/paper/2305.14325/) (2023)
- [CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society](https://mlsanity.com/paper/2303.17760/) (2023)
- [On the Utility of Learning about Humans for Human-AI Coordination](https://mlsanity.com/paper/1910.05789/) (2019)
- [Evaluating Large Language Models in Theory of Mind Tasks](https://mlsanity.com/paper/2302.02083/) (2023)
- [The Hanabi Challenge: A New Frontier for AI Research](https://mlsanity.com/paper/1902.00506/) (2019)
- [SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks](https://mlsanity.com/paper/2503.15478/) (2025)
- [Collab-Overcooked: Benchmarking and Evaluating Large Language Models as Collaborative Agents](https://mlsanity.com/paper/2502.20073/) (2025)
- [A Natural Language Corpus of Common Grounding under Continuous and Partially-Observable Context](https://mlsanity.com/paper/1907.03399/) (2019)
- [Why Reasoning Fails to Plan: A Planning-Centric Analysis of Long-Horizon Decision Making in LLM Agents](https://mlsanity.com/paper/2601.22311/) (2026)
- [The Collaboration Gap: Exploration and Benchmarking of Open-World Agentic Cooperation](https://mlsanity.com/paper/2511.02687/) (2025)
- [MT-PingEval: Evaluating Multi-Turn Collaboration with Private Information Games](https://mlsanity.com/paper/2602.24188/) (2026)
- [More Capable, Less Cooperative? When LLMs Fail At Zero-Cost Collaboration](https://mlsanity.com/paper/2604.07821/) (2026)
- [A Generalist Hanabi Agent](https://mlsanity.com/paper/2503.14555/) (2025)
- [Policy4OOD: A Knowledge-Guided World Model for Policy Intervention Simulation against the Opioid Overdose Crisis](https://mlsanity.com/paper/2602.12373/) (2026)
- [Food4All: An Agentic Framework and Benchmark for Food Resource Navigation with Adaptive User Understanding](https://mlsanity.com/paper/2510.18289/) (2025)
- [PreScam: A Benchmark for Predicting Scam Progression from Early Conversations](https://mlsanity.com/paper/2605.12243/) (2026)
- [ProPlay: Procedural World Models for Self-Evolving LLM Agents](https://mlsanity.com/paper/2606.12780/) (2026)
- [MDForge: Agentic Molecular Dynamics Pipeline Design under Sparse Simulator Feedback](https://mlsanity.com/paper/2606.12916/) (2026)

## Built on by (most cited first, 2 of 2)

- [Pay for the Fault, Not the Flow: Label-Free In-Flow Multi-Agent Workflow Optimization](https://mlsanity.com/paper/2610.01017/) (2026)
- [Training LLMs for Multi-Agent Collaboration with Verifiable Information Rewards](https://mlsanity.com/paper/2610.17893/) (2026)

## Abstract (arXiv)

Multi-agent systems built from large language models are deployed widely, yet how much performance is lost when two LLMs must coordinate rather than act alone remains unclear. We formulate the collaboration tax as the team-decentralisation loss of a two-player cooperative game with private information, with two propositions characterising its sign and its equivalence to a max-superadditivity violation. We operationalise this definition on 32 solo-tractable tasks grouped by source of grounding friction and measure it on 11 models from 7 providers. The tax is structured along two no-exception axes: a category ordering across every model and a monotonic decrease with capability. The proximate mechanism is not a reasoning deficit but a four-stage conversational cascade in which agents make ungrounded claims, fail to query the partner, skip integrating both views, and accept the answer without re-derivation. The tax is mechanically predictable from conversation features and partly tractable: a prompt intervention targeting all four stages closes a substantial fraction of the gap, with the dominant bottleneck differing across categories. In heterogeneous pairs the tax is pulled toward the stronger partner rather than the additive midpoint, empirically realising the max-superadditivity violation predicted by our framework. Together these results recast collaboration in LLM systems as a measurable, predictable, and partly tractable cost.