跳到主内容
@wquguru
精选88DAIR.AI(RSS)论文研究

DAIR.AI周报:量子电路发现与混合推理智能体

🥇Top AI Papers of the Week

原文
发到 X

1. Quantum Circuit Discovery

1. 量子电路发现

Many quantum algorithms rely on a small set of computational primitives, and compiling those primitives into quantum circuits is hard because the computation must be reversible. Researchers from Anthropic, Harvard, and QuEra Computing pair an LLM that proposes circuit designs with an independent evaluator that verifies and scores every candidate.

许多量子算法依赖于少量计算原语,将这些原语编译为量子电路之所以困难,是因为计算必须是可逆的。来自 Anthropic、哈佛大学和 QuEra Computing 的研究人员将提出电路设计的大型语言模型(LLM)与独立评估器配对,后者负责验证并对每个候选方案进行评分。

  • Evolutionary search: The search runs over circuit-generating programs with OpenEvolve, using Claude Mythos 5 to propose architectures and program rewrites. Each search starts from a hand-written task specification and a blank program, and the best designs seed the next round.
  • Verified circuits: The evaluator checks each circuit for correctness before scoring it by Toffoli count and peak qubit count. Every reported gain is a circuit that passed this check.
  • Results: On seven of eight tasks spanning numerical computation, chemistry, optimization, quantum simulation, and number theory, the discovered circuits use fewer Toffoli gates and fewer qubits than the best-known designs, cutting the combined cost by 4.1x to 39x. For a Reed-Solomon decoder, the search switched to the Berlekamp-Massey recurrence and lowered the Toffoli count from 5.1 million to 440,000.
  • Why it matters: On the eighth task, the search found a new reversible architecture for prime-field modular inversion that makes divsteps, a classical algorithm previously impractical for reversible integer arithmetic, usable and improves the Toffoli-qubit trade-off. The authors suggest that some apparent costs of quantum arithmetic reflect limits of existing reversible constructions rather than of the algorithms themselves.
  • 进化搜索:使用 OpenEvolve 在生成电路的程序上进行搜索,利用 Claude Mythos 5 提出架构和程序重写。每次搜索都从手写任务规范和空白程序开始,最佳设计将作为下一轮的种子。
  • 已验证电路:评估器在通过 Toffoli 门数量和峰值量子比特数量对电路进行评分之前,会先检查其正确性。所有报告的改进均指通过了此检查的电路。
  • 结果:在涵盖数值计算、化学、优化、量子模拟和数论的八项任务中的七项上,发现的电路使用的 Toffoli 门和量子比特数量均少于已知最佳设计,使综合成本降低了 4.1 倍至 39 倍。对于里德-所罗门解码器,搜索转向了 Berlekamp-Massey 递推算法,并将 Toffoli 门数量从 510 万降低至 44 万。
  • 意义所在:在第八项任务中,搜索发现了一种用于素域模逆运算的新可逆架构,使得 divsteps(一种此前在实际可行的可逆整数算术中难以应用的经典算法)变得可用,并改善了 Toffoli 门与量子比特之间的权衡。作者指出,某些量子算术的表观成本反映的是现有可逆构造的限制,而非算法本身的限制。

Paper | Tweet

论文 | 推文

Message from the Editor

编辑寄语

We are excited to release our new MCP server that lets you search and explore all the past papers in our AI Papers of the Week index directly from Claude Code or Codex. Set it up with one command, npx @dair-ai/mcp setup.

我们很高兴发布新的 MCP 服务器,让您能够直接从 Claude Code 或 Codex 中搜索和浏览我们“每周 AI 论文”索引中的所有过往论文。只需一条命令即可设置:npx @dair-ai/mcp setup。

Get Started

开始使用

2. LongHarness

Agents that reason over document collections larger than their context window usually keep retrieved evidence and their evolving conclusions mixed into one growing interaction history. Researchers from Shanghai AI Laboratory, Fudan University, and colleagues present LongHarness, a model-agnostic harness for corpora of 100M tokens and more that keeps evidence and conclusions as explicit state.

在处理超出其上下文窗口的文档集合进行推理时,智能体通常会将检索到的证据与其不断演变的结论混合在一个不断增长的历史交互中。来自上海人工智能实验室、复旦大学及同事的研究人员提出了 LongHarness,这是一个模型无关的框架,适用于包含 1 亿个 token 及以上的语料库,它将证据和结论作为显式状态保存。

  • Two failure modes: The authors identify unbounded exploration, where retrieval keeps expanding the history without closing the information gap that matters, and evidence-state drift, where earlier evidence or later corrections stop reaching the current conclusion.
  • Evidence-Conclusion Coevolution: Open conclusions decide what to retrieve next, and each new piece of evidence strengthens or revises the conclusions that asked for it. An evidence lifecycle step keeps accumulated evidence current and traceable throughout the interaction.
  • Same model, much higher scores: With GLM-5.3-Flash, exact match on OfficeQA Pro rises from 15.8% to 60.2% on V1 and from 4.4% to 61.1% on V2 compared with direct reasoning. LongHarness beats Claude Code by 12.0 points on V1 and Codex by 18.9 points on V2, and Qwen3.8-Flash reaches 61.7% on V1.
  • Why it matters: The gains come from the harness, with the backbone held fixed. Teams building agents for long-context work over large internal corpora can get large improvements by managing evidence explicitly instead of relying on a growing chat history.
  • 两种失效模式:作者指出了无界探索,即检索不断扩展历史,却未缩小关键的信息差距;以及证据-状态漂移,即早期证据或后续修正无法再影响当前结论。
  • 证据与结论的协同演化:开放式结论决定下一步检索什么,而每条新证据都会强化或修订提出该请求的结论。一个证据生命周期步骤在整个交互过程中保持累积证据的时效性和可追溯性。
  • 同一模型,得分大幅提高:使用 GLM-5.3-Flash,相较于直接推理,OfficeQA Pro 上的精确匹配率从 V1 的 15.8% 提升至 60.2%,从 V2 的 4.4% 提升至 61.1%。LongHarness 在 V1 上比 Claude Code 高出 12.0 分,在 V2 上比 Codex 高出 18.9 分,Qwen3.8-Flash 在 V1 上达到 61.7%。
  • 为何重要:这些提升来自 Harness(框架/工具),骨干模型保持不变。为大型内部语料库构建长上下文 Agent 的团队,通过显式管理证据而非依赖不断增长的聊天历史,可以获得显著改进。

Paper | Tweet

论文 | 推文

3. Hybrid Reasoning Agents

3. 混合推理 Agent

Today’s LLM agents write out every reasoning step as text before they search, run code, or take an action, and those steps cost tokens and time. Researchers from Stanford, Together AI, and colleagues introduce Hybrid Reasoning Agents, which reason internally in latent space and switch to text only to call a tool or give an answer.

当前的 LLM Agent 在搜索、运行代码或采取行动之前,会将每个推理步骤都输出为文本,而这些步骤会消耗 Token 和时间。来自斯坦福大学、Together AI 及合作者的研究人员引入了混合推理 Agent(Hybrid Reasoning Agents),它们在潜在空间(latent space)中进行内部推理,仅在调用工具或给出答案时才切换至文本模式。

  • HybridLink: A lightweight module moves the model between latent and text modes within a single generation process. A continuous working memory keeps both kinds of context across turns, tool calls stay in plain text, and the latent steps can still be inspected.
  • Two-stage training: HybridSFT first teaches the backbone to work with HybridLink during tool-integrated generation. HybridRL then teaches the model when to think in latent space and when to act through text.
  • Results: Across 3 model scales and 11 benchmarks in math, science, medicine, search, and agentic reasoning, it improves over frontier LLM agents and agentic frameworks, both text-based and latent, by 10.4% on average. The 14B model beats the strongest agentic baseline by 17.4% on agentic reasoning tasks.
  • Why it matters: The agents use 35.5% to 54.7% fewer tokens, run 1.8x to 2.6x faster, and cost 2.7x to 3.7x less. Researchers have mostly studied latent reasoning without tools, and this work shows how to combine its efficiency with standard tool use.
  • HybridLink:一个轻量级模块在单次生成过程中使模型在潜在模式和文本模式之间切换。连续的working memory(工作记忆)在多轮对话中同时保留两类上下文,工具调用保持纯文本形式,且潜在推理步骤仍可被检查。
  • 两阶段训练:HybridSFT 首先教导骨干模型在与 HybridLink 集成的生成过程中协同工作。随后,HybridRL 教导模型何时在潜在空间中思考,何时通过文本行动。
  • 结果:在数学、科学、医学、搜索和 Agent 推理领域的 3 个模型规模和 11 个基准测试中,其平均性能比前沿 LLM Agent 和 Agent 框架(包括基于文本的和基于潜空间的)提高了 10.4%。14B 模型在 Agent 推理任务上比最强的 Agent 基线高出 17.4%。
  • 为何重要:Agent 使用的 Token 减少了 35.5% 至 54.7%,运行速度提高了 1.8 倍至 2.6 倍,成本降低了 2.7 倍至 3.7 倍。研究人员此前主要研究不带工具的潜在推理,而这项工作展示了如何将其效率与标准工具使用相结合。

Paper | Tweet

论文 | 推文

4. Harness^2

Recursive self-improvement of agent harnesses usually depends on a verifier, and open-world tasks are new at test time, so no ground-truth checker exists. Researchers from Google Cloud AI Research and Northwestern University introduce Harness^2, a plug-and-play framework that improves the harness at test time through two connected levels of recursion.

代理工具(agent harnesses)的递归自我改进通常依赖于验证器,而开放世界任务在测试时是新的,因此不存在真实标签检查器。Google Cloud AI Research 和西北大学的研究人员引入了 Harness^2,这是一个即插即用框架,通过两个相互连接的递归层级在测试时改进工具。

  • Task-level recursion: A contrastive Propose-Probe-Compose step proposes several edits to the harness for the current task, tests them against each other, and keeps the combination that works best. This guides edits reliably without a ground-truth verifier.
  • Harnessing-level recursion: Lessons from those edits update a second, continual improvement harness that the agent carries to future tasks, so improvements carry over instead of staying task-specific.
  • Results: A single step improves GLM-5 and Gemini 3.5 Flash over their base harnesses in every configuration tested, with gains of up to 11.5 points on WorkBuddy-Bench, 4.5 points on JobBench, and 11.8 points on LAB. Running five candidate edits in parallel adds another 2.9 to 3.9 points on WorkBuddy-Bench.
  • Why it matters: ReasoningBank, a memory baseline, fell up to 6.6 points below the base harness on coding tasks in the same comparison. For teams evolving their own harnesses, the paper shows that test-time harness edits can scale with more recursion budget even when no verifier is available.
  • 任务级递归:对比式的提议-探测-组合步骤为当前任务提出对工具的几种编辑方案,相互测试它们,并保留效果最佳的组合。这在没有真实标签验证器的情况下可靠地指导编辑过程。
  • 工具级递归:从这些编辑中获得的经验教训会更新第二个持续改进的工具,该工具由代理携带至未来任务,从而使改进得以延续,而非局限于特定任务。
  • 结果:在每个测试配置下,单步操作均使 GLM-5 和 Gemini 3.5 Flash 优于其基础工具,在 WorkBuddy-Bench 上提升高达 11.5 分,在 JobBench 上提升 4.5 分,在 LAB 上提升 11.8 分。并行运行五个候选编辑方案可进一步在 WorkBuddy-Bench 上增加 2.9 至 3.9 分。
  • 重要性:在同一比较中,作为记忆基线的 ReasoningBank 在编码任务上比基础工具低多达 6.6 分。对于正在开发自有工具的团队,论文表明,即使没有验证器可用,测试时的工具编辑也能随着更多递归预算的增加而扩展。

Paper | Tweet

论文 | 推文

5. Scaling Discovery Requires Forgetting

5. 扩展发现需要遗忘

Automated research with coding agents can now run for days on a single problem, yet the search often stops improving early. Researchers from UC Berkeley, Stanford, and Bespoke Labs show that the fix is to make the agents forget.

使用编码代理的自动化研究现在可以在单个问题上运行数天,但搜索往往在早期就停止改进。来自加州大学伯克利分校、斯坦福大学和 Bespoke Labs 的研究人员表明,解决方案是让代理学会遗忘。

  • The plateau: Letting one agent iterate longer, running agents in parallel, and letting parallel agents share candidates all plateau as history accumulates. Each agent anchors on its own past attempts and makes fewer, smaller edits, so extra budget yields little improvement.
  • Forgetting: The method periodically discards every agent’s session and the candidate history, then restarts the team from the best candidate found so far. The authors implement this in a team of agents that communicate.
  • Results: Across four tasks in heuristic programming, kernel optimization, and model training, forgetting gives up to 15% over parallel agents and up to 39% over a single agent with the same budget. On Anthropic’s VLIW kernel take-home, the team with forgetting reaches 1,077 cycles, against 1,172 for four parallel agents and 1,262 for one agent.
  • Why it matters: Most research-agent designs try to keep as much history as possible. This paper suggests that, for open-ended discovery, what an agent is allowed to remember matters as much as how much compute it gets.
  • 瓶颈期:让一个代理迭代更长时间、并行运行代理以及让并行代理共享候选方案,都会随着历史记录的积累而出现性能瓶颈。每个代理都锚定于其过去的尝试,并进行更少、更小的编辑,因此额外的预算带来的改进微乎其微。
  • 遗忘:该方法定期丢弃每个代理的会话和候选历史记录,然后从迄今为止找到的最佳候选方案重新开始整个团队。作者在一组通信代理中实现了这一机制。
  • 结果:在启发式编程、内核优化和模型训练的四个任务中,遗忘机制相比并行智能体最高提升15%,相比同等预算下的单个智能体最高提升39%。在Anthropic的VLIW内核家庭作业测试中,采用遗忘机制的团队达到1,077周期,而四个并行智能体为1,172周期,单个智能体为1,262周期。
  • 意义所在:大多数研究智能体设计试图保留尽可能多的历史记录。本文表明,对于开放式探索,智能体被允许记住的内容与其获得的计算资源同样重要。

Paper | Tweet

论文 | 推文

6. ReHarness

更进一步:量化金融体系

看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力

进入量化体系 →

相似阅读

关联信息,但可能不是同一事件