多智能体协作提升数学解题准确率，但对恶意干扰仍不鲁棒。
More Agents Improve Math Problem Solving but Adversarial Robustness Gap Persists
- 用投票框架让多个智能体协同解题，数量越多越准
- 50%标点噪声下准确率下降超30%，人写错字仍是最大漏洞
- 25个智能体也难防攻击，适合研究模型安全的读者
当大语言模型智能体协同工作时，其在数学问题求解上的表现优于单个模型。然而，它们对对抗性输入是否更鲁棒？我们通过扰动数学题目进行研究，包括三种强度的标点噪声（10%、30%、50%），以及真实世界和类人错字（WikiTypo、R2ATA）。采用统一的采样-投票框架（Agent Forest），评估六种开源模型（Qwen3-4B/14B、Llama3.1-8B、Mistral-7B、Gemma3-4B/12B）在四个基准（GSM8K、MATH、MMLU-Math、MultiArith）上的表现，设置不同智能体数量 n = {1,2,5,10,15,20,25}。结果表明：1）噪声类型影响显著，标点噪声危害随强度增加，而人类错字仍是主要瓶颈，导致与干净输入准确率差距最大，且攻击成功率（ASR）最高，即使智能体数量庞大；2）协作能稳定提升准确率，从n=1到n=5增益最大，超过n≈10后收益递减；但对抗鲁棒性差距始终存在，不随智能体数量增加而缩小。
原文摘要 · Abstract (English)
When LLM agents work together, they seem to be more powerful than a single LLM in mathematical question answering. However, are they also more robust to adversarial inputs? We investigate this question using adversarially perturbed math questions. These perturbations include punctuation noise with three intensities (10%, 30%, 50%), plus real-world and human-like typos (WikiTypo, R2ATA). Using a unified sampling-and-voting framework (Agent Forest), we evaluate six open-source models (Qwen3-4B/14B, Llama3.1-8B, Mistral-7B, Gemma3-4B/12B) across four benchmarks (GSM8K, MATH, MMLU-Math, MultiArith), with various numbers of agents n = {1,2,5,10,15,20,25}. Our findings show that 1) Noise type matters: punctuation noise harm scales with its severity, and the human typos remain the dominant bottleneck, yielding the largest gaps to Clean accuracy and the highest attack success rate (ASR) even with a large number of agents; 2) Collaboration reliably improves accuracy as the number of agents, n, increases, with the largest gains from n=1 to n=5 and diminishing returns beyond n$\approx$10. However, the adversarial robustness gap persists regardless of the agent count.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。