AI 与机器学习2024arxiv.org
More Agents Is All You Need：LLM 性能随 agent 数量扩展
More Agents Is All You Need: LLMs performance scales with the number of agents
入门 · 约 30 分钟读完前置知识：LLM 推理与采样（temperature）基础、多数投票和集成学习思想、文本相似度度量（如 BERTScore）
导读
这篇论文研究 LLM 性能是否随智能体数量增加而提升。核心做法非常简单：对同一模型重复运行同一查询多次（智能体之间不共享上下文），再用相似度算法选出出现最多的答案作为最终结果。实验显示这种朴素方法随智能体数量增加性能持续提升，效果不输甚至有时优于 CoT、LLM-Debate 等复杂多智能体框架。
HN 关注点：是否暗示复杂的多智能体协作框架其实没有特别之处，收益主要来自多次采样加多数投票，同时讨论其计算成本是否可持续。
HN 评论精选
这几条评论里：支持 1 质疑 2 补充 1
补充论文的简单重复采样+多数投票方法效果不输复杂多智能体框架，暗示后者的收益其实来自多次运行而非巧妙设计
我不确定评论区的人是否正确理解了这篇论文。它实际上可能推翻了 CoT、LLM-Debate 这类多智能体方案的理念。论文提出的方法只是对同一个 LLM 把同一查询跑多次，各次之间不共享任何上下文，然后对答案跑相似度算法、选出最常见的答案（这说得通：如果 LLM 输出的是幻觉和正确答案的混合，正确答案会彼此相似，而幻觉往往是混乱的）。这个简单算法的表现不输甚至有时优于其他多智能体算法。这暗示那些带巧妙提示词的多智能体方案并没有什么特别之处，其提升主要来自把 LLM 跑了多次、并让 LLM 挑选最佳答案。
phire查看英文原文
I'm not sure people in these comments are reading this paper correctly. This seems to essentially disprove the whole idea of multi-agent setups like Chain-of-thought and LLM-Debate. Because this paper introduces their alternative method which simply runs the same query multiple times on the same LLM, without any context shared across queries. And then they run a similarity algorithm on the answers and pick the most common answer. (Which makes sense to me. If an LLM is giving you a mixture of "hallucinations" and correct answers, the correct answers will similar and the hallucinations will hopefully be chaotic) And this simple algorithm preform just as well (and sometimes better) than all the other multi-agent algorithms. This suggests that the other multi-agent schemes with their clever prompts aren't really doing anything special; Their improve results are coming mostly from the fact that the LLM is run multiple times, that the prompt asks the LLM to pick the best answer.
支持支持多智能体叠加的方向，指出收益递减拐点与人类理想会议规模相似，期待差异化微调智能体的后续研究
终于来了。我大约 16 个月前就说过，我们不该只盯着单个智能体把所有事做对，而应该分层叠加智能体，现在终于有论文可以引用了。有趣的是，任务收益递减的拐点大致与人类理想会议规模相当。如果用更细粒度的智能体数量做实验，我很想知道能多接近那些数字。我还很期待后续研究：如果每个智能体朝略微不同的目标微调，能再提升多少性能？我预计哪怕只是给各智能体设置不同的 temperature 都会有提升。很高兴看到研究界开始朝这个方向走。
kromem查看英文原文
Finally. I've been saying that we need to stop focusing on a single agent getting everything right and instead layer agents for about 16 months now, but it's great to have a paper to point to. It's interesting that the diminishing returns for tasks flatten out rapidly around the same size as the ideal human meeting sizes: https://www.researchgate.net/figure/18-Optimal-Meeting-Sizes-1_tbl6_3889414 If this was done at more granular steps of agent quantity I'm curious just how closely it would match those numbers. I'd also really love to see the eventual follow-up where we see how much more performance can be obtained when the agents are each fine tuned towards slightly different aims. I'd expect there'd even be a performance lift from just having the agents each set at different temperature levels. Very happy to see the research community starting to step in this direction!
质疑质疑该方法成本过高：单次查询计算量增加 7-15 倍，难以持续
这是不是一种极其昂贵、不可持续的方法？我同意 MoE 才是方向、新模型可能面临收益递减的看法。但单个 prompt 的计算量会突然增加 7-15 倍？
trash_cat查看英文原文
Is this not an incredibly expensive/unsustainable method? I agree with the sentiment that MoE is the way to go as the newer models will probably see diminishing returns. But the compute for a single prompt will suddenly increase 7-15 fold?
质疑批评论文标题滥用 "All You Need" 梗，分散了对真正洞见的注意力
论文标题里这种 "All You Need" 的风潮该消停了。最初的 "Attention is All You Need" 用这个标题是因为在他们的场景里字面上成立。现在太多论文只是把它当梗用，反而掩盖了论文真正的洞见。
spencerchubb查看英文原文
This trend of "All You Need" in paper titles needs to die. The original "Attention is All You Need" used that title because it is literally true in their case. So many papers just use it as a meme now, and it distracts from the true insight of the paper.
相关资源
从这篇论文所有 HN 讨论帖的全部评论里整理出来的链接。
历次讨论
- 2024Hacker News 讨论帖288 分 · 206 评
同领域还值得读
DeepSeek-R1：通过强化学习激励 LLM 的推理能力
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL
研究如何用强化学习激励大语言模型的推理能力，提出 DeepSeek-R1，并探讨 RL 训练对模型推理表现的影响。
AI 与机器学习2025 · 1,351 分 · 1,056 评 · 引用 30进阶 · 约 80 分钟编辑精选1-bit LLM 时代：面向低成本计算的三值参数
The Era of 1-bit LLMs: ternary parameters for cost-effective computing
提出用三值参数 {-1, 0, 1} 替代 LLM 中的浮点权重，即所谓“1-bit LLM”。
AI 与机器学习2024 · 1,040 分 · 447 评 · 引用 19进阶 · 约 45 分钟编辑精选DeepSeek-V3.2：推进开放大语言模型的前沿
DeepSeek-v3.2: Pushing the frontier of open large language models
这是 DeepSeek-V3.2 的技术报告，延续了 DeepSeek 开源权重模型的路线，目标是让开放模型继续逼近甚至追平前沿闭源模型。
AI 与机器学习2025 · 982 分 · 465 评进阶 · 约 90 分钟编辑精选通过自我对弈与通用强化学习掌握国际象棋和将棋（AlphaZero）
Mastering Chess and Shogi by Self-Play with General Reinforcement Learning
DeepMind 的 AlphaZero 论文：用通用强化学习算法（自我博弈 + MCTS + 神经网络），从零开始仅数小时训练便在国际象棋和将棋上击败了世界顶级引擎 Stock…
AI 与机器学习2017 · 539 分 · 270 评 · 引用 13进阶 · 约 90 分钟编辑精选用强化学习发现更快的矩阵乘法算法
Discovering faster matrix multiplication algorithms with reinforcement learning
论文提出 AlphaTensor，用强化学习搜索矩阵乘法算法。
AI 与机器学习2022 · 518 分 · 111 评专业 · 约 90 分钟编辑精选