Weak-Link Optimization for Multi-Agent Reasoning and Collaboration

# Weak-Link Optimization for Multi-Agent Reasoning and Collaboration
Source: https://arxiv.org/html/2604.15972
Haoyu Bian1, Chaoning Zhang1, Jiaquan Zhang1, Xingyao Li1, Yuanfang Guo2, Wei Dong3, Yang Yang1, This work was supported by the National Natural Science Foundation of China (NSFC) under the General Program (Grant No. 62572104).

###### Abstract

LLM-driven multi-agent frameworks address complex reasoning tasks through multi-role collaboration. However, existing approaches often suffer from reasoning instability, where individual agent errors are amplified through collaboration, undermining overall performance. Current research mainly focuses on enhancing high-capability agents or suppressing unreliable outputs to improve framework effectiveness, while systematic identification and reinforcement of performance-limiting agents receive less attention. To address this gap, we propose WORC, a weak-link optimization framework for multi-agent reasoning and collaboration, grounded in the weak-link principle. WORC follows a two-stage workflow. In the weak agent localization stage, task features are constructed, and a meta-learning-based weight predictor trained on optimal configurations identified by swarm intelligence algorithms (SIAs) enables zero-shot mapping from these features to agent performance weights, where the agent with the lowest predicted weight is identified as the weak agent. In the weak-link optimization stage, an uncertainty-driven allocation strategy assigns additional reasoning budgets to weak agents, with lower predicted weights leading to larger repeated-sampling quotas to compensate for reliability deficiencies. Experimental results show that WORC achieves an average accuracy of 82.2% on reasoning benchmarks while improving framework stability and cross-architecture generalization, suggesting that compensating for weak links, rather than reinforcing strengths alone, enhances the robustness of multi-agent systems.

###### Index Terms:

LLM, multi-agent, weak-link, reasoning optimization

## I Introduction

Large Language Models (LLMs) have demonstrated remarkable capabilities in generative natural language processing tasks[8](https://arxiv.org/html/2604.15972#bib.bib1),[19](https://arxiv.org/html/2604.15972#bib.bib87), yet they continue to underperform in mathematical problem-solving and logical reasoning. In response, researchers proposed reasoning methods such as Chain of Thought (CoT)[52](https://arxiv.org/html/2604.15972#bib.bib11), which formalize human reasoning approaches into prompt templates and emphasize subtask decomposition and multi-step reasoning. Recent studies have further explored task-driven alignment and structure-aware reasoning-chain optimization[59](https://arxiv.org/html/2604.15972#bib.bib90),[60](https://arxiv.org/html/2604.15972#bib.bib89). Concurrently, the emergence of AI Agents[34](https://arxiv.org/html/2604.15972#bib.bib12), particularly multi-agent frameworks[43](https://arxiv.org/html/2604.15972#bib.bib15) leveraging planning, reflection, and tool utilization capabilities across collaborating specialized agents, has significantly enhanced LLMs' performance on complex problem-solving tasks[49](https://arxiv.org/html/2604.15972#bib.bib17). Recent advances further extend collaborative reasoning beyond static agent cooperation toward interaction-aware[63](https://arxiv.org/html/2604.15972#bib.bib77), role-adaptive[31](https://arxiv.org/html/2604.15972#bib.bib78), and consensus-driven collaboration paradigms[56](https://arxiv.org/html/2604.15972#bib.bib79), enabling more structured deliberation processes across distributed reasoning agents. These developments improve the effectiveness of multi-agent reasoning systems in long-horizon decision-making tasks, including scientific programming[37](https://arxiv.org/html/2604.15972#bib.bib80), medical diagnosis[66](https://arxiv.org/html/2604.15972#bib.bib81), and autonomous planning[53](https://arxiv.org/html/2604.15972#bib.bib82),[26](https://arxiv.org/html/2604.15972#bib.bib91).

Refer to captionFigure 1: Overview of the vulnerability of weak agents in multi-agent reasoning. Sequential reasoning propagates errors from underperforming agents. Majority voting and multi-agent debate mitigate but do not eliminate the influence of weak agents.

However, in complex reasoning tasks, multi-agent frameworks face substantial coordination challenges, requiring task decomposition and collaboration to align individual agent actions with overall objectives[28](https://arxiv.org/html/2604.15972#bib.bib16). The reliability of a reasoning path depends on the compounded reliability of its components, making the system inherently susceptible to performance degradation[65](https://arxiv.org/html/2604.15972#bib.bib57). Individual underperforming agents in a multi-agent architecture, hereafter referred to as weak agents, compromise the overall reliability of the system by inducing inaccurate reasoning, unreliable decisions, and error-prone outputs. Conventional design paradigms, which emphasize stronger reasoning agents or incorporate simple consensus mechanisms such as voting[16](https://arxiv.org/html/2604.15972#bib.bib59) and debate[29](https://arxiv.org/html/2604.15972#bib.bib58), remain susceptible to instability and exhibit high performance variability despite their effectiveness[10](https://arxiv.org/html/2604.15972#bib.bib56). This fragility manifests specifically as:

1. Error accumulation across reasoning stages: In task decomposition, outputs of preceding agents serve as inputs for subsequent ones. Low-accuracy or miscalibrated outputs from any agent may propagate errors downstream, amplifying their impact.
2. Consensus degradation under heterogeneous agent reliability: Consensus mechanisms rely on agreement among agents. Erroneous contributions from limited-capability agents may degrade overall decision quality and introduce systemic biases.

Similarly, multi-path reasoning approaches alleviate these risks by exploring multiple candidate reasoning trajectories; however, their effectiveness remains constrained by weak agents and may introduce additional computational overhead[9](https://arxiv.org/html/2604.15972#bib.bib28).

To address these limitations, we propose WORC, a reasoning optimization framework for LLM-driven multi-agent systems grounded in the weak-link optimization principle. This perspective is inspired by the bottleneck-driven system optimization principle, commonly referred to as the "Buckets Effect," which has been widely adopted in system reliability engineering, production optimization, and fault-tolerant distributed system design, where overall system performance is constrained by its weakest components. In the context of multi-agent reasoning, this motivates a shift toward targeted compensation of weak agents to enhance reasoning reliability.

To operationalize this principle, WORC adopts a two-stage optimization process consisting of weak agent localization and weak-link optimization. In the weak agent localization stage, SIAs are employed to estimate optimal agent weight vector configurations based on multi-agent reasoning performance over sampled task-type datasets, thereby capturing task-dependent agent contributions within collaborative reasoning processes. This formulation leverages the population-based global search capability of swarm intelligence methods to model agent performance distributions without requiring explicit supervision, and constructs the resulting weight vectors as a knowledge base for cross-task generalization. When new reasoning tasks are encountered, task signatures are constructed using text embedding models such as OpenAI embeddings, incorporating semantic mean embeddings and structural statistical features. These signatures are subsequently processed by a meta-learning-based weight predictor to retrieve the most relevant weight vector from the knowledge base as a benchmark for weak agent identification. In the weak-link optimization stage, an automatic budget allocation mechanism assigns additional reasoning resources to the identified weak agents based on the predicted weight configuration. All agents then generate candidate solutions according to their allocated quotas, and the final output is obtained through a voting-based aggregation module.

As a demonstration of this generalizable optimization method, we design a simple chain-based multi-agent reasoning system called AgentChain (AC) as an illustrative implementation of our approach. Additionally, we conduct comprehensive evaluations across different datasets and tasks, demonstrating the framework's enhanced reasoning capabilities, stability, and interpretability. Our contributions include:

1. We propose an optimization method for LLM-driven multi-agent reasoning, inspired by the "weak-link" principle, which focuses on enhancing system robustness by addressing weak components in the architecture.
2. To generalize weak agent detection across different tasks, we construct a meta-learning weight predictor and SIAs for task feature analysis, enabling zero-shot identification of weak agents across tasks.
3. Comprehensive experimental evaluations and theoretical analyses demonstrate the method's effectiveness in enhancing reasoning accuracy and system stability across various multi-agent frameworks.

## II Related Work

### II-A Multi-Agent Systems for Reasoning

Multi-Agent Systems represent one of the key architectural paradigms for large language models (LLMs)[18](https://arxiv.org/html/2604.15972#bib.bib24),[43](https://arxiv.org/html/2604.15972#bib.bib15), enabling distributed reasoning through collaboration and interaction among multiple agents[23](https://arxiv.org/html/2604.15972#bib.bib26). Multi-agent architectures lead to improved reasoning performance by enabling structured decomposition of reasoning processes through collaborative interaction[21](https://arxiv.org/html/2604.15972#bib.bib25). The application of multi-agent frameworks addresses several limitations found in single-agent reasoning. Chen et al.[9](https://arxiv.org/html/2604.15972#bib.bib28) demonstrate that by constructing a multi-agent, multi-reasoning path framework, where language models play different roles and collaborate, task-solving performance on complex scientific problems can be improved. Similarly, Gu et al.[17](https://arxiv.org/html/2604.15972#bib.bib29) showed that breaking complex tasks into subtasks and employing pipeline-style multi-agent collaboration facilitates large models in tackling complex reasoning problems.

Existing multi-agent reasoning enhancement methods exhibit several notable limitations:

- Majority voting treats all agents equally, failing to identify or downweight weak or unreliable agents[48](https://arxiv.org/html/2604.15972#bib.bib60).
- Self-consistency mechanisms reinforce correlated erroneous reasoning trajectories by repeatedly emphasizing similar inference paths[54](https://arxiv.org/html/2604.15972#bib.bib61).
- Debate-based approaches are susceptible to destabilization when incorrect or misleading arguments dominate the discussion process[32](https://arxiv.org/html/2604.15972#bib.bib62).
- Static weight allocation ignores task-specific and context-dependent variations in agent performance[36](https://arxiv.org/html/2604.15972#bib.bib63).

### II-B Meta-Learning and Task Adaptation

The core objective of meta-learning is to enable models to leverage prior task distributions to rapidly adapt to new tasks[22](https://arxiv.org/html/2604.15972#bib.bib66). Early research such as MAML[15](https://arxiv.org/html/2604.15972#bib.bib64) optimized parameter initialization for fast task adaptation across tasks. Subsequently, methods like Prototypical Networks[40](https://arxiv.org/html/2604.15972#bib.bib65) demonstrate efficient few-shot inference via task-level representations in few-shot classification scenarios. With the rise of LLMs, meta-learning concepts have been applied to enhance LLMs' generalization and adaptation capabilities[39](https://arxiv.org/html/2604.15972#bib.bib67).

Introducing meta-learning into multi-agent reasoning systems aims primarily to improve agents' collective collaboration abilities, enabling overall joint adaptation dynamics. Current research mainly follows two paths. The first involves building agents with meta-cognitive abilities (such as ReMA[46](https://arxiv.org/html/2604.15972#bib.bib68), MetaMind[62](https://arxiv.org/html/2604.15972#bib.bib69)) that enable planning, monitoring, and adjustment of their own reasoning processes. The second focuses on meta-level coordination[41](https://arxiv.org/html/2604.15972#bib.bib70), directly optimizing collaboration patterns between agents through game theory or meta-learning strategies, allowing systems to quickly form efficient team reasoning structures. However, this field still faces significant challenges, including high computational costs and excessive dependence on foundation model capabilities[6](https://arxiv.org/html/2604.15972#bib.bib71).

### II-C Swarm Intelligence Algorithms in LLMs

Swarm Intelligence Algorithms (SIAs) have been widely studied as effective tools for solving complex optimization problems[7](https://arxiv.org/html/2604.15972#bib.bib32) by mimicking collective behaviors observed in biological systems. Classical algorithms, including Particle Swarm Optimization (PSO)[24](https://arxiv.org/html/2604.15972#bib.bib33) and Grey Wolf Optimizer (GWO)[35](https://arxiv.org/html/2604.15972#bib.bib55), established population-based stochastic optimization frameworks for continuous and combinatorial optimization problems. Recently, novel SIAs have continued to emerge with refined search mechanisms and enhanced optimization performance. For example, the Marine Predator Algorithm[13](https://arxiv.org/html/2604.15972#bib.bib50) simulates different hunting behaviors based on the encounter rate between prey and predators. The Hippopotamus Optimization (HO) Algorithm[3](https://arxiv.org/html/2604.15972#bib.bib22) simulates hippos' territorial marking and defensive attack behaviors.

Recent studies have brought SIAs into LLMs and deep learning architectures, showing clear improvements in optimization efficiency, parameter tuning, and handling of complex reasoning tasks[44](https://arxiv.org/html/2604.15972#bib.bib35). For example, researchers use SIAs' parallel search abilities to explore parameter and collaboration policy search spaces in neural reasoning systems[25](https://arxiv.org/html/2604.15972#bib.bib36). However, while SIAs have been widely applied in traditional multi-agent computer systems, their application in LLM-driven multi-agent systems remains underexplored, with examples like SwarmSys[27](https://arxiv.org/html/2604.15972#bib.bib72) and AMRO-S[51](https://arxiv.org/html/2604.15972#bib.bib88) introducing SIAs to achieve scalable and adaptive reasoning. Nevertheless, this integration still contains significant gaps in research.

Refer to captionFigure 2: Overview of the WORC method in the AC framework. (a) Weak Agent Localization: A weight knowledge base is constructed via SIA training, and task signatures are generated. A meta-learning predictor outputs the most suitable weight vector for a new task, enabling identification and assessment of weak agents. (b) Weak Agent Optimization: The predicted weight vector guides targeted allocation of reasoning budget to compensate low-performing agents. Agents execute sequentially within the AgentChain framework, with VoteAgent selecting the best outputs to achieve collaborative reasoning and performance optimization.

## III Method

Th