AI 与机器学习2025arxiv.org
Hierarchical Reasoning Model
Hierarchical Reasoning Model
进阶 · 约 50 分钟读完前置知识：深度学习基础、循环神经网络/Transformer、ARC-AGI 与约束求解
导读
HRM（层级推理模型）提出一种不依赖预训练和思维链监督的循环层级架构，仅用约 1000 个输入输出样例、27M 参数就能在数独、30x30 迷宫寻路和 ARC-AGI 等任务上取得亮眼成绩（论文报告 ARC 约 40.3%）。其核心卖点是通过分层、多次迭代的推理来规避沿时间反向传播的高昂代价。
HN 关注点：标题和摘要里“小模型、无预训练、碾压 CoT 大模型”的强对比引发关注，但评论者普遍质疑其评测方式、可比性与过度宣传。
HN 评论精选
这几条评论里：支持 1 质疑 2 补充 1
质疑论文未经同行评审、评测细节缺失，且 ARC2 成绩远非榜首。
我建议保持怀疑。这项工作确实有些很有意思的想法，尤其是规避了沿时间反向传播的开销。但它似乎没有经过同行评审。结果部分很奇怪：没有说明评测是怎么做的，唯一的数值只出现在首页那张图里。而且 ARC2 的成绩（与该图相反）并不是榜首——目前榜首是 19%，而 HRM 只有 5%。
JonathanRaines查看英文原文
I advise scepticism. This work does have some very interesting ideas, specifically avoiding the costs of backpropagation through time. However, it does not appear to have been peer reviewed. The results section is odd. It does not include include details of how they performed the assesments, and the only numerical values are in the figure on the front page. The results for ARC2 are (contrary to that figure) not top of the leaderboard (currently 19% compared to HRMs 5% https://www.kaggle.com/competitions/arc-prize-2025/leaderboard)
质疑HRM 是针对特定谜题训练的求解器，与通用 CoT 模型不可比，属于过度宣传。
我欣赏它与神经科学的联系，论文本身也没让我感到警报，如果落到我手上评审，我不会拒稿。但我对其结论的适用性极度怀疑。从文中看，他们似乎造出了一个通用的（至少是可适配的）约束满足求解器，能从少量样例中学习约束满足问题的规则。如果属实确实很酷（我还没空复现他们的例子并迁移到别的问题上），但我不理解为什么要和 CoT 模型比较。CoT 模型原则上能解决任何复杂任务；而它需要针对特定谜题训练后才能求解，并不声称通用，甚至不清楚它是否被设计成能适配任意谜题。根据论文内容和所选示例，我猜并非如此。这有点像声称 Stockfish 比当前最先进 LLM 聪明得多，因为下棋能把它们打得落花流水。作者有好想法，但宣传得……过于慷慨了。
malcontented查看英文原文
I appreciate the connections with neurology, and the paper itself doesn't ring any alarm bells. I don't think I'd reject it if it fell to me to peer review. However, I have extreme skepticism when it comes to the applicability of this finding. Based on what they have written, they seem to have created a universal (maybe; adaptable at the very least) constraint-satisfaction solver that learns the rules of the constraint-satisfaction problem from a small number of examples. If true (I have not yet had the leisure to replicate their examples and try them on something else), this is pretty cool, but I do not understand the comparison with CoT models. CoT models can, in principle, solve _any_ complex task. This needs to be trained to a specific puzzle which it can then solve: it makes no pretense to universality. It isn't even clear that it is meant to be capable of adapting to any given puzzle. I suspect this is not the case, just based on what I have read in the paper and on the indicative choice of examples they tested it against. This is kind of like claiming that Stockfish is way smarter than current state of the art LLMs because it can beat the stuffing out of them in chess. I feel the authors have a good idea here, but that they have marketed it a bit too... generously.
支持引述论文摘要中的关键数据，认为结果惊人并打算细读。
初步扫了一遍摘要和引言，HRM 的结果看起来不可思议：只用 1000 个输入输出样例，无预训练、无 CoT 监督，就能学会解决即使最先进 LLM 也束手无策的问题。比如在复杂数独（Sudoku-Extreme Full）上接近完美，在 30x30 迷宫上能找最优路径，而最先进的 CoT 方法完全失败（0% 准确率）。在 ARC AGI Challenge 上，仅用官方数据集（约 1000 例）从头训练、27M 参数、30x30 网格上下文（900 token），就达到 40.3%，明显超过 o3-mini-high（34.5%）和 Claude 3.7 8K 上下文（21.2%），尽管后者参数与上下文规模大得多。我要把全文仔细读一遍。感谢分享到 HN！
cs702查看英文原文
Based on a quick first skim of the abstract and the introduction, the results from hierarchical reasoning (HRM) models look incredible: > Using only 1,000 input-output examples, without pre-training or CoT supervision, HRM learns to solve problems that are intractable for even the most advanced LLMs. For example, it achieves near-perfect accuracy in complex Sudoku puzzles (Sudoku-Extreme Full) and optimal pathfinding in 30x30 mazes, where state-of-the-art CoT methods completely fail (0% accuracy). In the Abstraction and Reasoning Corpus (ARC) AGI Challenge 27,28,29 - a benchmark of inductive reasoning - HRM, trained from scratch with only the official dataset (~1000 examples), with only 27M parameters and a 30x30 grid context (900 tokens), achieves a performance of 40.3%, which substantially surpasses leading CoT-based models like o3-mini-high (34.5%) and Claude 3.7 8K context (21.2%), despite their considerably larger parameter sizes and context lengths, as shown in Figure 1. I'm going to read this carefully, in its entirety. Thank you for sharing it on HN!
补充观察到符号式/搜索式方法似乎正在回归 AI 研究。
是我错觉吗，还是符号式（我更愿称之为“电子游戏式”）AI 正在重新渗回 AI 领域？
torginus查看英文原文
Is it just me or are symbolic (or as I like to call it 'video game') AI is seeping back into AI?
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。
- 代码
- github.com/sapientinc/HRM被提到 7 次
- 评论区提到的论文
历次讨论
- 2025Hacker News 讨论帖339 分 · 106 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