TOPIC · AI4SCIENCE · 19 PAPERS

AI for Science

大语言模型推理对齐智能体多模态视觉生成视觉理解与 3D 机器人与具身高效训练与推理安全与可解释 AI for Science 代码与软件工程评测与数据强化学习

近 30 天 19

#63 ▲155 ★19 ❝0 2026-10-13

Into the Unknown: Benchmarking AI agents on open scientific research problems

Into the Unknown 让 Agent 针对美国国家实验室科学家提供的未发表方向写续作论文，评测开放科研能力。

Carnegie Mellon University · Seungone Kim, Jiayi Geng, Young-Jun Lee +25

智能体 AI for Science 评测与数据 arXiv 代码 HF

#69 ▲54 ★809 ❝2 2026-10-09

RSI-Anything: Unlocking Recursive Self-Improvement with World Models

RSI-Anything 借助预测式世界模型选择行动和训练数据，实现跨领域的递归自我改进。

PhAI Labs · Taoyong Cui, Pinxian Zeng, Wanghan Xu +8

智能体 AI for Science arXiv 代码 HF

#100 ▲64 ★574 ❝0 2026-10-20

Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition

Sage-A1.5 是 35B MoE 智能体模型，新增子智能体并行协作、过程自检和开放式科学探索能力。

Intern Science · Wenjie Lou, Tianshuo Peng, Shiyang Feng +96

智能体 AI for Science arXiv 代码 HF

#129 ▲237 ❝0 2026-10-19

Scaling Parameter-Efficient Fine-Tuning for Brain Encoding

用参数高效微调把最大 104B 的多模态模型适配到脑编码任务，更准确地预测大脑对刺激的反应。

Mind Lab · Zhuofan Josh Ying, Guanglong Sun, Nolan Ho +6

AI for Science 高效训练与推理 arXiv HF

#141 ▲112 ★7 ❝1 2026-09-30

EvoDuet: Bilevel Co-Evolution of Web Searching and Task Solving for Scientific Discovery

EvoDuet 让解法和搜索查询双层协同进化，帮助大模型进化搜索在需要外部知识时不再停滞。

Minnesota NLP · Young-Jun Lee, Jinheon Baek, Soyeong Jeong +5

AI for Science 智能体 arXiv 代码 HF

#144 ▲33 ★141 ❝4 2026-09-21

From Pattern Recognizers to Personalized Companions: A Survey of Large Language Models in Mental Health

大模型在心理健康领域应用的综述，梳理从模式识别到个性化陪伴的演进脉络。

机构待核实 · He Hu, Yucheng Zhou, Qianning Wang +8

大语言模型 AI for Science arXiv 代码 HF

#160 ▲68 ★22 ❝0 2026-09-30

Scoring Higher, Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics

指出基于评分细则的 RL 用加权求和会被“刷分”（以临床问诊为例），并提出协议级细则加以缓解。

ByteDance · Maoqi Liu, Junwei He, Bowen Zhang +5

强化学习对齐 AI for Science arXiv 代码 HF

#180 ▲51 ★112 ❝0 2026-09-30

Does Learning Protein Folding Generalize to Broader Reasoning?

研究让大模型学习蛋白质折叠能否迁移成通用的空间与结构推理能力。

Shanghai JiaoTong University · Yong Liu, Zhanpeng Shi, Yizhou Dang +4

AI for Science 推理 arXiv 代码 HF

#193 ▲47 ★164 ❝0 2026-10-01

Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems

SciUtopia 用大模型 Agent 闭环模拟学术生态：选题、合作、投稿、评审、引用和资助。

Georgia Institute of Technology · Yiqiao Jin, Yiyang Wang, Lucheng Fu +8

智能体 AI for Science arXiv 代码 HF

#202 ▲63 ★14 ❝1 2026-09-30

Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers

SciSlopBench 收录 390 篇 AI 生成论文，检测看似合理但推理断裂的“科研垃圾”，并探索缓解办法。

Seoul National University · Yerim Oh, Young-Jun Lee, Jaewoo Ahn +2

评测与数据 AI for Science arXiv 代码 HF

#208 ▲103 ❝0 2026-09-29

Predictive Credit: Measuring What Scientific Explanations Add to Experimental Forecasts

提出“预测贡献”指标，衡量科研 Agent 对实验的解释是否真能改善结果预测；预注册检验结论为不确定。

Carnegie Mellon University · Jingjie Ning, Xueqi Li, Yibo Kong +1

AI for Science 评测与数据 arXiv HF

#237 ▲31 ★26 ❝1 2026-09-24

ExplorationBench: Measuring AI Systems' Exploration in Verifiable Alien Worlds

腾讯混元 ExplorationBench 在可验证的虚构世界中测 AI 的科学探索能力，排除预训练知识的干扰。

Tencent Hunyuan · Ming Zhang, Zhenghao Xiang, Peizhong Gao +17

AI for Science 评测与数据智能体 arXiv 代码 HF

#243 ▲53 ❝2 2026-09-29

AIM: Agentic Idea Management for Automated Research

谷歌 AIM 借鉴贝叶斯优化，自动管理、筛选和推进自动化科研中的研究想法。

Google · Hyeong Kyu Choi, Bhavana Dalvi Mishra, Jiefeng Chen +7

智能体 AI for Science arXiv HF

#254 ▲53 ★4 ❝0 2026-10-15

From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research

PhysClosureBench 用 50 个专家验证的物理研究任务，评估 Agent 能否组织研究并得出有证据支撑的结论。

Shanghai AI Laboratory · Yijia Mao, Song Dai, Tao Chen +14

AI for Science 智能体评测与数据 arXiv 代码 HF

#257 ▲65 ★12 ❝0 2026-09-28

PDE-JEPA: Predictive Representation Learning of Latent Dynamics Modeling for Parametric PDEs

PDE-JEPA 用预测式表示预训练学习参数化偏微分方程的潜在动力学，替代重建式目标。

Zhejiang University · Zhentao Tan, Jianrong Zhang, Ruijie Quan +1

AI for Science 高效训练与推理 arXiv 代码 HF

#262 ▲79 ❝0 2026-10-12

X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning

X-Ragent 用工具增强强化学习训练胸片推理，学会何时调用辅助证据、过滤不适用的量化证据。

Carnegie Mellon University · Yi Wang, Zhenlong Yuan, Yuchan Guo +9

强化学习推理 AI for Science arXiv HF

#263 ▲30 ★11 ❝1 2026-10-06

Stateless Language Agents: Scaling Long-Horizon Automated Research

斯坦福“无状态语言智能体”把研究状态放到外部，解决长程自动化科研中重复劳动和停滞的问题。

Stanford University · Qizheng Zhang, Changxiu Ji, Isaac Sun +15

智能体 AI for Science arXiv 代码 HF

#299 ▲24 ★4 ❝0 2026-10-05

Optimizing the Optimizer: Language Models Discover Faster Molecular Relaxation Algorithms

让语言模型 Agent 改写分子几何优化器，找到比现有最快开源方法更省力的算法。

机构待核实 · Artem Tsypin, Vladimir Deshchenya, Kuzma Khrabrov +4

AI for Science 智能体 arXiv 代码 HF

#306 ▲23 ❝0 2026-09-24

Where Does Retrieval-Based Open-Ended Evaluation Fail? Automatic Taxonomy Induction from Long-Form Medical Answer Factuality Verification

为医疗长答案的事实核查建立错误分类体系，找出基于检索的评测在哪里失效。

Center for Language and Speech Processing @ JHU · Heyuan Huang, Jirui Dai, Alexandra DeLucia +5

AI for Science 评测与数据 arXiv HF

在归档中检索该主题 →