覆盖北京时间日期：2026-07-07、2026-07-08。聚焦 Agent、LLM reasoning/planning/tool use/memory、RAG、coding agents、evaluation/benchmark 与训练/推理基础设施。
Top Picks
#1 · Total 24/25
Wei Wu, Fangjing Wang, Fan Lu, He Sun, Shi Liu, Yunnan Wang · 2026-07-07 · arXiv export API
一句话结论：Despite recent progress of VLA foundation models, the disparity between laboratory conditions and real-world applications continues to impede their practical implementation. To bridge this g…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Despite recent progress of VLA foundation models, the disparity between laboratory conditions and real-world applications continues to impede their practical implementation. To bridge this gap, we present LingBot-VLA 2.0, which advances LingBot-VLA through imp…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#2 · Total 24/25
He Liu, Changtao Miao, Xinjie Yang, Tianle Song, Yin Wu, Junchi Chen · 2026-07-07 · arXiv export API
一句话结论：Large language models deployed in open-world applications require safety guardrails that are both robust to complex risks and efficient enough for low-latency runtime moderation. Existing gu…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Large language models deployed in open-world applications require safety guardrails that are both robust to complex risks and efficient enough for low-latency runtime moderation. Existing guardrails face a practical trade-off between lightweight classification…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#3 · Total 24/25
Grace Man Chen, Litao Guo, Yifan Wu, Yiyu Chen, Yenchi Tseng, Sicheng Liu · 2026-07-07 · arXiv export API
一句话结论：Large language models (LLMs) have demonstrated growing competence in web page generation. However, existing text-driven approaches rely on complex prompts that impose substantial demands on…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Large language models (LLMs) have demonstrated growing competence in web page generation. However, existing text-driven approaches rely on complex prompts that impose substantial demands on users and offer limited expressivity for page layout and cross-page vi…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#4 · Total 24/25
Yijun Zhang, Fan Xu, Jiaxin Ding, Yule Xie, Shiqing Gao, Xin Ding · 2026-07-07 · arXiv export API
一句话结论：Reinforcement learning has become a promising paradigm for improving large language model (LLM) agents on long-horizon search tasks, where the agent must make a sequence of intermediate deci…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Reinforcement learning has become a promising paradigm for improving large language model (LLM) agents on long-horizon search tasks, where the agent must make a sequence of intermediate decisions before receiving a final outcome. However, existing methods stil…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#5 · Total 24/25
Chenxu Wang, Yongkun Yang, Boyuan Du, Shiwei Lin, Huaping Liu · 2026-07-07 · arXiv export API
一句话结论：Deliberation plays a crucial role in collaboration; when humans work together, they naturally engage in communication to align information and reach an agreement. In this paper, we investiga…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Deliberation plays a crucial role in collaboration; when humans work together, they naturally engage in communication to align information and reach an agreement. In this paper, we investigate deliberative large language model (LLM) agents under partially obse…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#6 · Total 24/25
Wei Dong, Tianyu Fu, Zhe Yu, Hanning Wang, Anyang Su, Zhizhou Fang · 2026-07-07 · arXiv export API
一句话结论：As web agents increasingly demonstrate capabilities in automated task execution, the development of robust evaluation frameworks for assessing their navigation and task completion performanc…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：As web agents increasingly demonstrate capabilities in automated task execution, the development of robust evaluation frameworks for assessing their navigation and task completion performance has emerged as a critical research priority. However, existing bench…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#7 · Total 24/25
Hongliang Li, Yijin Liu, Zhiwei Zhang, Zihe Liu, Xinyue Lou, Jinan Xu · 2026-07-07 · arXiv export API
一句话结论：Large language model (LLM) agents have shown strong performance in long-horizon tasks that require planning, tool use, and interaction with external environments. However, most existing benc…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Large language model (LLM) agents have shown strong performance in long-horizon tasks that require planning, tool use, and interaction with external environments. However, most existing benchmarks implicitly assume a monolingual setting, where the entire execu…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#8 · Total 24/25
Adam Faci, Alessio Miaschi, Anne Combe, Pascal Cuxac, Francesca Frontini, Nicolas Larrousse · 2026-07-07 · arXiv export API
一句话结论：The integration of Large Language Models (LLMs) into scientific research workflows, particularly for bibliographic discovery and literature synthesis, raises significant methodological, epis…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：The integration of Large Language Models (LLMs) into scientific research workflows, particularly for bibliographic discovery and literature synthesis, raises significant methodological, epistemic and regulatory challenges for the Social Sciences and Humanities…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#9 · Total 24/25
Tianyuan Zhang, Xianglong Liu, Aishan Liu, Lu Wang, Yitong Zhang, Peng Yue · 2026-07-07 · arXiv export API
一句话结论：Environmental illusions (eg., shadows, reflections, and tire marks) are naturally existing yet overlooked phenomena in real-world driving environments. They can disturb visual perception, le…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Environmental illusions (eg., shadows, reflections, and tire marks) are naturally existing yet overlooked phenomena in real-world driving environments. They can disturb visual perception, leading to misinterpretation of the scene and posing serious safety risk…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#10 · Total 24/25
Huakun Liu, Qing Yu, Kent Fujiwara, Hideaki Uchiyama, Kiyoshi Kiyokawa · 2026-07-07 · arXiv export API
一句话结论：Generating temporally continuous and socially coherent human motion from text remains a fundamental challenge, particularly in realistic streams where people act alone, enter interactions, a…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Generating temporally continuous and socially coherent human motion from text remains a fundamental challenge, particularly in realistic streams where people act alone, enter interactions, and later disengage. Most existing methods generate fixed-length motion…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#11 · Total 24/25
Chang Nie, Jiaju Wei, Junlan Feng, Chaoyou Fu, Caifeng Shan · 2026-07-07 · arXiv export API
一句话结论：Agentic video understanding equips models with long-term memory to autonomously process and respond to continuous, long-horizon multimodal streams. However, advanced video agents often rely…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Agentic video understanding equips models with long-term memory to autonomously process and respond to continuous, long-horizon multimodal streams. However, advanced video agents often rely on ``detective-style'' iterative reasoning for action control (e.g., $…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#12 · Total 24/25
Runwei Guan, Shaofeng Liang, Jiacheng Weng, Xiaoyi Gu, Jia Weng, Daizong Liu · 2026-07-07 · arXiv export API
一句话结论：Male infertility contributes substantially to the global infertility burden, and sperm analysis remains central to diagnosis, treatment planning, and assisted reproductive technology. Conven…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Male infertility contributes substantially to the global infertility burden, and sperm analysis remains central to diagnosis, treatment planning, and assisted reproductive technology. Conventional semen evaluation, however, is labor-intensive, operator-depende…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 5Actionability 5
arXiv · PDF
#13 · Total 23/25
Tianxing Chen, Yue Chen, Zixuan Li, Junyuan Tang, Kailun Su, Haoran Lu · 2026-07-06 · arXiv export API
一句话结论：Generalist robot manipulation policies have advanced rapidly, yet existing benchmarks remain limited in systematically evaluating their capabilities. Many rely on simple, short-horizon, or s…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Generalist robot manipulation policies have advanced rapidly, yet existing benchmarks remain limited in systematically evaluating their capabilities. Many rely on simple, short-horizon, or skill-narrow tasks with limited capability coverage, and are often cond…
实验/证据：Evidence 4/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 4Actionability 5
arXiv · PDF
#14 · Total 23/25
Zhuofan Zhang, Tianxu Wang, Guoxi Zhang, Yixiong Lin, Xilin Wang, Hongming Xu · 2026-07-08 · arXiv export API
一句话结论：Mobile manipulation requires a robot to navigate to a target object or receptacle and then perform intended manipulation. However, reaching the vicinity of the target does not guarantee a ma…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Mobile manipulation requires a robot to navigate to a target object or receptacle and then perform intended manipulation. However, reaching the vicinity of the target does not guarantee a manipulation-ready base pose, a problem known as last-mile navigation. P…
实验/证据：Evidence 4/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 4Substance 5Evidence 4Actionability 5
arXiv · PDF
#15 · Total 23/25
So Hasegawa, Shailaja Keyur Sampat, Lei Liu, Wei-Peng Chen · 2026-07-08 · arXiv export API
一句话结论：Current benchmarks for evaluating Large Language Models (LLMs) in data analysis often fail to reflect real-world settings. They typically focus on fact retrieval from small tables and overlo…
为什么重要：贴近 Agent / LLM reasoning / coding / evaluation / personalization 研究线，适合快速转成复现实验或产品验证。
方法要点：Current benchmarks for evaluating Large Language Models (LLMs) in data analysis often fail to reflect real-world settings. They typically focus on fact retrieval from small tables and overlook the challenges of large multi-tabular datasets, external knowledge…
实验/证据：Evidence 5/5；需全文核验 benchmark、baseline、ablation 与代码可得性。
局限/风险：快筛基于官方元数据/摘要，结论强度以论文全文为准。
Lucian 下一步：抽取任务定义、指标与 baseline，加入 Auto Research 阅读/复现实验队列。
Relevance 5Novelty 3Substance 5Evidence 5Actionability 5
arXiv · PDF