CRAFT: Grounded Multi-Agent Coordination Under Partial Information

TL;DR

CRAFT introduces a benchmark for multi-agent coordination with language models under partial info, revealing current models' limitations in pragmatic communication and collaboration.

Contribution

It formalizes a multi-sender pragmatic communication problem, provides a diagnostic framework, and evaluates diverse models revealing coordination challenges.

Findings

Stronger reasoning does not always improve coordination.

Smaller open-weight models can outperform larger frontier models.

Improved individual communication does not ensure successful collaboration.

Abstract

We introduce CRAFT, a multi-agent benchmark for evaluating pragmatic communication in large language models under strict partial information. In this setting, multiple agents with complementary but incomplete views must coordinate through natural language to construct a shared 3D structure that no single agent can fully observe. We formalize this problem as a multi-sender Bounded Pragmatic Speaker problem and provide a diagnostic framework that decomposes failures into spatial grounding, belief modeling and pragmatic communication errors, including a taxonomy of behavioral failure profiles in both frontier and open-weight models. Across a diverse set of models, including 8 open-weight and 7 frontier including reasoning models, we find that stronger reasoning ability does not reliably translate to better coordination: smaller open-weight models often match or outperform frontier systems, and improved individual communication does not guarantee successful collaboration. These results suggest that multi-agent coordination remains a fundamentally unsolved challenge for current language models. Our code can be found at https://github.com/csu-signal/CRAFT