Jun 2026· arXiv.org· Vol abs/2606.27492· 1 citation· 26 references
Computer Science
TL;DR
QueenBee Planner is introduced, a framework that treats inter-agent communication topology as a retrievable and self-improving design skill and suggests that multi-agent systems can learn reusable architectural design knowledge rather than merely memorizing task answers.
Abstract
Large language model (LLM) multi-agent systems increasingly depend not only on how individual agents reason, but also on how agents are connected. This paper introduces QueenBee Planner, a framework that treats inter-agent communication topology as a retrievable and self-improving design skill. A pool of worker agents, the task adapter, and the scoring function are frozen; only an outer LLM planner learns to generate temporal communication DAGs specifying who sends information to whom, in which round, who merges messages, and who emits the final answer. Execution traces are distilled into evidence-backed design rules with three actions: \emph{Preserve}, \emph{Modify}, and \emph{Avoid}. To prevent self-evolution from turning lucky runs or plausible but false explanations into policy, QueenBee uses held-out acceptance gates, variance-aware credit, motif-level attribution, transfer trust, insight falsification, and structural deduplication. We evaluate the method on Count-Frequency aggregation and Silo-Bench-style distributed coordination tasks. With fixed workers, self-evolved graph generation produces communication structures that improve over fixed topologies and cold generation. In the CF fulltest setting, the best generated graph reduces RMSE from 12.53 for the strongest fixed topology to 7.87 while also reducing messages, model calls, and token cost; Silo-style results show the same direction of improvement over cold and fixed-topology baselines. These results suggest that multi-agent systems can learn reusable architectural design knowledge rather than merely memorizing task answers.
Roles provide an interpretable interface for organizing language-model agents, yet most multi-agent systems treat them as hand-written prompt labels disconnected from learned behavior and parameter updates. We argue that a useful role should instead be an executable control variable: it should summarize behavior predictive of future utility, guide subsequent interaction, and identify the trainable capacity responsible for that behavior. We introduce ExRole, a trajectory-to-role framework that learns future-aware role prototypes from prefix-local team traces, resolves them into readable instructions and token-aligned role markers, and optionally routes shared LoRA rank slots with turn-aligned credit. Across MuSiQue and 2WikiMultiHopQA, ExRole improves over single-agent search by 15.0/14.4 and 13.5/16.1 EM/F1 points, respectively. Against the strongest non-ExRole controls, the corresponding gains remain 11.5/11.6 and 7.7/9.7 points. Across both benchmarks, the controlled results consistently favor trajectory-induced role conditioning over role-free, manual, random, and shuffled alternatives. Role-Agent-Turn interventions further show that the induced roles capture transferable behavioral specialization beyond fixed agent identities or turn positions.
Zhou Liu, Chaoyang Han, Zewei Pan et al.· 0 citations
Training LLM-based multi-agent systems with multi-agent reinforcement learning with multi-agent reinforcement learning is rapidly gaining traction, and a parallel line of work argues that such systems should be judged by their behavior, not only their reward.
Multi-agent LLM pipeline systems break down the task among multiple roles for better reasoning, but are benchmarked mainly with large-scale commercial models. In this study, we investigate Parishad, a structured multi-agent system involving five roles, by deploying it on Qwen2.5-7B-Instruct, a local model, on two datasets: GSM8K (500 questions) and HumanEval (164 questions), compared with prompting directly and two-call self-refinement. The multi-agent system drops GSM8K accuracy from 75.0\% to 45.0\% with JSON data format due to the error accumulation problem. With plaintext format, the accuracy is restored to 82.0\%. A two-call self-refinement strategy (V1) can achieve 86.2\% accuracy on GSM8K, with 7.4$\times$ lower token usage. However, the same V1 implementation on HumanEval---where direct accuracy is already 96.3\%---actively destroys performance (66.5\%). A task-aware gated redesign (V2) applied to HumanEval preserves accuracy at 95.1\%. Our results demonstrate that communication format and implementation details determine outcomes more than architectural complexity, and that simpler approaches match or outperform multi-agent pipelines for local 7B model deployment. All code and data are released.
Multi-agent ensembling multiplies active parameters and inference cost without answering three basic questions: which agents to consult, how deeply a query should traverse a hierarchy of agents, and when inter-agent communication is worth its cost. We present GRADE (Gated Routing and Adaptive Depth for Efficient Reasoning), a hierarchical multi-agent system in which four lightweight learned gates jointly govern agent selection, hierarchy depth, inter-agent communication, and branch pruning. Training uses CoGRPO (Collaborative Group-Relative Policy Optimization), a novel critic-free recipe that adapts GRPO to multi-agent hierarchies and assigns a shared advantage signal to every gate and agent that participated in a rollout. Agent models are drawn from a hot-swappable Expert Registry; per-agent calibration maps allow experts to be replaced at inference time without retraining. At $\sim$17B average active parameters, GRADE outperforms all baselines on GSM8K, MMLUPro, and GPQA, surpassing the strongest baseline by 4.8 points on MMLUPro at half the active compute. On AIME-2025, where model depth dominates, GRADE remains competitive to existing frameworks. Ablations isolate the hierarchy and masked cross-attention as the largest contributors to accuracy, and show that per-agent calibration is necessary for safe hot-swapping.
Improving the task performance of Large Language Models (LLMs) is essential, yet scaling these models faces significant challenges such as diminishing returns and high costs. Multi-Agent Systems (MAS) offer a promising solution by distributing tasks among specialized agents to improve the overall task performance. This can reduce training costs at the expense of increased test time due to the discussion and decision-making process. The decision protocol is a critical component of MAS because it specifies how multiple agents collaborate to create a final solution. This thesis introduces the Multi-Agent LLM (MALLM) framework, which implements and evaluates various decision protocols, namely voting, consensus, and judge decision mechanisms, to simulate multi-agent discussions for conversational task solving. Unlike previous work that used a single decision protocol or tested them on limited datasets, this study systematically examines their impact on a diverse set of tasks, ranging from knowledge-based datasets (MMLU, MMLU-Pro, GPQA) and logic-based datasets (StrategyQA, MuSR, Math-lvl-5, SQuAD 2.0). The results indicate that consensus protocols excel in knowledge-intensive domains while voting and judge protocols are more effective for logic-based tasks. Increasing response diversity through independent solution generation improves decision quality, while changes in information access during the decision process have minimal impact.
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.
Weixiang Sun, Zehong Wang, Hong Huang et al.· 0 citations