We present DPBench, a benchmark for evaluating coordination in multi-agent systems built from large language models. Existing benchmarks measure task-level success under a fixed protocol; the structural conditions under which coordination succeeds or fails at all have not been characterised. DPBench adapts the Dining Philosophers problem into a controlled testbed where the action protocol, the communication structure, and the group size each vary independently. We evaluate six agents: GPT-5.2, Claude Opus 4.5, Grok 4.1, Gemini 2.5 Flash, Llama 4 Maverick, and a uniform-random baseline. Under simultaneous action at N=5 with the default prompt, deadlock ranges from 25.0% (95% Wilson CI [11.2, 46.9]) for GPT-5.2 to 90.0% [74.4, 96.5] for Gemini 2.5 Flash; sequential action is solved by four of the six. Holding the model fixed at Gemini 2.5 Flash, three protocol variables drive deadlock from 90% to within CI of zero: three rounds of pre-commitment communication (0.0% vs. single-round 86.7%), a prompt encoding a classical concurrency primitive (0.0% for resource-ordering and symmetry-breaking, against 100% for the minimal prompt), or doubling the group from N=5 to N=10 (90.0% to 10.0%). Single-round messaging and memory of past timesteps do not change the rate at the sample size we ran. Whether the same model coordinates or deadlocks is determined by the protocol, not by the model's capability.

翻译：本文提出DPBench，一个用于评估基于大语言模型构建的多智能体系统协调能力的基准测试。现有基准在固定协议下衡量任务级成功率，但尚未系统刻画协调成功或失败的结构条件。DPBench将哲学家就餐问题改造为可控测试平台，其中行动协议、通信结构和群体规模可独立变化。我们评估了六个智能体：GPT-5.2、Claude Opus 4.5、Grok 4.1、Gemini 2.5 Flash、Llama 4 Maverick以及统一随机基线。在默认提示下N=5同时行动时，死锁率从GPT-5.2的25.0%（95% Wilson置信区间[11.2,46.9]）到Gemini 2.5 Flash的90.0%[74.4,96.5]不等；六种智能体中有四种可解决顺序行动问题。固定模型为Gemini 2.5 Flash时，三个协议变量将死锁率从90%降至置信区间包含零：三轮预承诺通信（0.0%对比单轮86.7%）、编码经典并发原语的提示（资源排序和对称破坏提示均为0.0%，而最小化提示为100%），或群体规模从N=5翻倍至N=10（90.0%降至10.0%）。在当前样本量下，单轮消息传递和过去时间步记忆未改变死锁率。同一模型协调还是死锁取决于协议，而非模型能力。