外观
Long-horizon, semi-verifiable agentic tasks often rely on rubric-based evaluation, but the evidence needed to assess these tasks can be spread across the environment state and the execution trace. A judge may overlook this evidence or acce…
Humanoid loco-manipulation requires coordinated whole-body interaction with the physical world, yet collecting robot demonstrations at scale remains costly. Human demonstrations offer a scalable alternative, but existing egocentric human s…
Cao 等人在 Nature Machine Intelligence 发表 AMIFound 基础模型，将跨七种医学影像模态的修复任务统一到单一模型中，能够在多种退化条件下实现稳健增强，并保留对诊断有用的临床信息。
Long-horizon agents often struggle to recover from failures whose causes lie several steps earlier, beyond what local reflection or backtracking can resolve. We introduce BELIEFWEAVER, a framework that combines persistent evidence across a…
研究者与US DOE国家实验室科学家合作，推出基准Into the Unknown：给AI Agent一篇科学论文和后续研究方向，要求写出续写论文，并用科学家给出的投稿条件形成评分标准。9个前沿Agent在量子计算、等离子体物理、计算化学、材料科学和科学机器学习领域各写15篇续写，共135篇，全部未满足所有条件，其中114篇未通过诚信检查，常见问题是声称未做的工作或报告与记录证据矛盾的数据。
Agent skills are Markdown documents that a tool-using agent reads as instructions and then acts on with the privileges of the session that loaded them. Installing one grants authority (the standing to direct the agent, not new privilege),…
3D part assembly aims to reconstruct a complete shape from a set of unassembled parts by predicting a rigid SE(3) transformation for each part. Because the placement of a part is rarely determined by its own geometry, the task demands reas…
研究者发布 Game2Real-Bench 基准，覆盖细粒度感知、空间推理、约束推理与序列规划四类能力，包含 1000 个可执行验证的程序化游戏样本和 400 个按推理需求对齐的真实样本。评测 25 个 VLM 后发现游戏与真实任务表现高度一致（相关系数 0.935）；one-shot 游戏中，部分模型的约束推理提升，但细粒度感知下降。团队还构建了含约 24K 样本、18 类任务、65 个 QA 家族的 Game2Real-Dataset。
本期头条是微软发布第三代自研 AI 加速器 Maia 300，称每兆瓦性能提升 40%，计划 2027 年起大规模部署；Meta 发布首批前沿级模型 Muse Wonder 与 Wonder Full，强化编码、基准表现和记忆用户信息能力。深度内容涉及 AI 生成软件的责任归属、Anthropic 测试模型自主复制能力、以及昂贵闭环任务为何难以训练。研究方面介绍了以不到 1% 预训练成本改造 Transformer 架构的 Priming 技术，以及 GLM-Legal 在 Harvey LAB 上 19.2% 全通过率。快讯还提到 Claude Opus 5.5 灰度测试、智谱 GLM-5.3-Highspeed 灰度上线约 490 tokens/s 等。
At AI Engineer New York, Anthropic drew on lessons from Unix and Linux to explain how agents can tackle bigger projects without overwhelming developers.
本期介绍 TasteVal 对 AI 科研能力的评测、HSS 视觉推理基准，以及 Reflection 和 Mistral 的模型进展；作者进一步讨论科研自动化、政府跨实验室训练与 AI 太空工业化的可能性。
Relational learning typically predicts from a database state, but decision making asks a different question: how will that state change under an action? In operational systems, the answer depends on system-specific dynamics that may not be…
Evaluation has been central to advances in AI, and it is equally critical to autonomous discovery. Yet scaling autonomous discovery to real-world engineering challenges faces a fundamental bottleneck: the cost of evaluating progress. In re…
Training LLM agents to interact with users over multiple turns is challenging because the learning signal depends not only on the agent, but also on the user simulator and environment, which jointly shape the agent's experience. Simulated…
Web application development often proceeds through successive requests that extend an existing implementation. Yet web-generation benchmarks primarily evaluate development toward fixed specifications. We introduce EvoWeb, a novel benchmark…
Recent advances in self-improving LLM agents have demonstrated the potential of AI systems that drive their own intelligence. However, many domain-specific applications rely on specialized foundation models that possess strong domain exper…
Reliable deployment of LLM-based multi-agent systems calls for runtime auditors that predict terminal outcomes from observed execution prefixes. These auditors typically learn from a limited corpus of trajectories with verified outcomes. E…
Inside Standard Bots’ AI stack, pretrained models learn factory tasks from demonstrations and improve through corrections from real deployments.
