AI 与机器学习2019arxiv.org
Hanabi 挑战：AI 研究的新前沿
The Hanabi Challenge: A New Frontier for AI Research
进阶 · 约 35 分钟读完前置知识：强化学习基础、多智能体系统、不完全信息博弈
导读
这篇论文由 DeepMind 提出将合作卡牌游戏 Hanabi 作为 AI 研究的新基准。在 Hanabi 中，玩家看不到自己的牌，只能用有限的提示动作向队友传递信息，因此需要不完全信息下的多智能体协作、心智理论（推断队友知道什么）以及通过行动隐式沟通。论文给出了该环境的描述与基线结果，主张它是继国际象棋、围棋之后检验这类能力的测试平台。
HN 关注点：焦点在于 DeepMind 又一次自行定义“AI 新前沿”是否合理，以及社区中已有 Hanabi 机器人强基线，同时好奇为什么选 Hanabi 而不是更普及的桥牌。
后来怎么样
Hanabi 成为合作式多智能体强化学习、不完全信息决策与心智理论研究中的常用基准，并催生了一系列 Hanabi 智能体与算法工作。
HN 评论精选
这几条评论里：质疑 1 补充 1 亲历 1
补充详细解释 Hanabi 的规则与高阶推理机制，说明为什么它是检验心智理论的合适游戏。
给不了解这个游戏的人补充背景：1) 每位玩家有一组牌，每张牌有特定颜色和数字。2) 玩家自己的牌只有别人看得到（牌面朝外摆放）。3) 轮到你时，你可以打出一张自己的牌，或就某位玩家的牌给出提示。4) 提示只能是“你有恰好 K 张 C 颜色的牌〈指出这 K 张〉”或“你有恰好 K 张 N 数字的牌〈指出这 K 张〉”这两种形式；提示次数有限，也有办法“赚回”更多提示。4b) 你能做出的推理可以超过提示本身给出的信息：每种牌数量固定，比如 RED-4 只有两张，所以如果你看到另外两位玩家各有一张 RED-4，就意味着你不可能有 RED-4；于是当有人提示你有 *-4 时，你就知道它不是红色的。但那两位拿着 RED-4 的玩家看不到自己的牌，做不出这个推理。而如果你用行动表明你已经做出了这个推理，且他们足够聪明，就能从你的行动反推出：你只有在确知自己没有 RED-4 时才会这么做；既然他们只看得到另一张 RED-4，就能推断出自己手里拿的是那张。5) 不允许其他任何交流。6) 桌子中央每种颜色有一叠起始为空的牌堆，游戏目标是……（原文在此截断）
gamegoblin查看英文原文
Context for those that don't know the game: 1. Each player has a set of tiles that each have a particular color and a particular number. 2. The tiles a player has are known only to other players (they are placed to face away from the player) 3. On your turn, you may either place one of your tiles, or give a hint to another player about their tiles. 4. Hints can only take the form of "you have exactly K tiles of C color <indicate the K tiles>" or "you have exactly K tiles of N number <indicate the K tiles>". There are some rules for how many hints can be given and ways to "earn back" more hints to give. 4b. You can make more deductions than what the hints provide. There are a fixed number of each tile. e.g. There are 2 RED-4 tiles. So if you can see that two other players both have a RED-4 tile, this means that you cannot have a RED-4 tile. So if someone gives you the hint that you have a *-4 tile, you know it is not RED. But the two players with the RED-4 tiles cannot see their own, so they cannot make this deduction. But if you act in a way that shows that you have made the deduction, and they are really smart, they can observe your act and deduce that you would have only made that act if you knew that you didn't have a RED-4 tile. And since they can only see 1 other RED-4 tile, they can deduce that they must hold the other. 5. No other communication is allowed. 6. In the center of the table there are stacks for each color that start empty. The goal of the game is to populate the stacks by playing the color tiles of that stack in increasing number. (e.g. RED-1, RED-2, RED-3 ...). If a player tries to play a tile that is invalid (e.g. the RED stack is only up to RED-2 and the player selects their RED-4 tile), a counter is incremented. 7. The game ends when the counter reaches a threshold or all tiles are played. The score is equal to the number of tiles played. Higher is better. I have long thought that it should be possible to write an AI that does superhuman hint-giving and deduction, provided that all of the AIs can collude beforehand. That is, if all of the AIs have a giant lookup table that essentially says "If the board is in this particular state, and I give hint H, that actually means FOO" where FOO contains more bits of information than hint H would by itself.
亲历开源 Hanabi 机器人作者现身，指出自己仍保持多人局 SOTA，并愿与作者交流。
从没想过我的 GitHub 仓库会被 DeepMind/Brain 的论文引用。我在 2 人以上对局的 SOTA 还保持着很大领先 :) 如果他们不用任何 Hanabi 专用技巧就能打败我的 5 人策略，我会相当佩服，不过我猜那不是他们的目标。无论如何祝 DeepMind/Brain 好运！作者们：如果你们在读这条，很希望能聊聊！如果我以后回去继续搞 Hanabi 机器人，可能会尝试用学到的策略替换我机器人里那些随机、非常不优化的启发式。
wuthefwasthat查看英文原文
Never thought I'd see my github repo cited in a DeepMind/Brain paper. I still have SOTA for >2 player by a large margin :) I'd be pretty impressed if they beat my 5p strategy without Hanabi-specific stuff, but I assume that's not their goal. Good luck to DeepMind/Brain either way! Authors: would love to chat sometime if you're reading this! If I get back to Hanabi botting, I might try to use a learned policy to replace some of the random (very non-optimized) heuristics my bot had.
质疑引用论文原文质疑：为何不直接用更广为人知的桥牌作为基准。
论文里写道：“Hanabi 的主要区别在于没有‘廉价交谈’的沟通渠道：任何信号都必须通过打出的行动来传递。因此在精神上更接近学习在桥牌的叫牌阶段向搭档传递信息。”Jakob，你为什么选择 Hanabi 而不是桥牌本身？桥牌显然知名度高得多，而且似乎带来类似的挑战。
te查看英文原文
From the paper: > The main difference in Hanabi is that there is no “cheap-talk” communication channel: any signalling must be done through the actions played. It is therefore more similar in spirit to learning how to relay information to partners in the bidding round of bridge. Jakob, why did you choose Hanabi instead of bridge itself? Bridge is obviously much more widely known and seems to present similar challenges.
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
- 2019Hacker News 讨论帖150 分 · 66 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