1
我国《人工智能拟人化互动服务管理暂行办法》今起施行，明确服务提供者安全主体责任
AI拟人化互动服务新规今日施行，明确安全评估与主体责任，为行业划清红线。
IT之家 7 月 15 日消息，国家互联网信息办公室、国家发展改革委、工业和信息化部、公安部、国家市场监督管理总局于 4 月 10 日联合公布《 人工智能拟人化互动服务管理暂行办法 》（以下简称《办法》），该办法自今起施行。 IT之家注：拟人化互动服务是指企业利用人工智能技术，向境内公众提供模拟自然…
2
FairCoder: Probing LLM Bias in High-Stakes Decision Making via Coding Tasks
通过编码任务揭示大语言模型在高风险决策中的潜在偏见，创新评估方法值得关注
arXiv:2501.05396v3 Announce Type: replace Abstract: Large language models (LLMs) are increasingly used in high-stakes decisions such as hiring and col…
3
LLM Judges Can Be Too Generous When There Is No Reference Answer
研究揭示LLM评估者没有参考答案时评分偏高，需警惕开源模型自评偏差
arXiv:2607.12885v1 Announce Type: new Abstract: LLM judges are increasingly being used to evaluate open-ended model responses, often in no-reference s…
4
DeepMind CEO calls for an independent standards body to regulate frontier AI
DeepMind CEO呼吁设立独立标准机构，对前沿AI进行自愿审查并可能转为强制要求，引发行业监管新思考。
DeepMind CEO Demis Hassabis is proposing an AI "standards body" modeled after FINRA, to test frontier models and develop best practices for their rele…
5
When and Why Does Multi-Agent Debate Fail and Does It Really Underperform?
多智能体辩论为何失效？这篇论文系统分析失败条件，挑战你对AI协作的固有认知。
arXiv:2510.20963v2 Announce Type: replace Abstract: Multi-agent debate (MAD) was proposed as a promising approach for ensembling the wisdom of multipl…
6
内存占用 54 GB → 3.9GB！苹果正评估 PrismML 量化 AI 技术，让 iPhone 15 及后续机型运行 27B 模型
IT之家 7 月 15 日消息，昨日（7 月 14 日）在接受 CNBC 采访时，PrismML 首席执行官巴博克 · 哈西（Babak Hassibi）表示， 其公司确实和苹果公司接洽，并透露苹果公司正在评估其相关 AI 量化技术。 在采访中，哈西表示：“他们 [指苹果公司] 目前正在认真评估我们…
7
FIRE-Bench: Evaluating AI Agents on the Rediscovery of Scientific Insights
一个评估AI智能体能否“重新发现”科学洞察的创新基准，ICML 2026收录，值得研究者关注。
arXiv:2602.02905v2 Announce Type: replace Abstract: Autonomous agents powered by large language models (LLMs) promise to accelerate scientific discove…
8
I Ran 10 AI Coding Models Through 5 Tasks: A Data Scientist's Take
数据科学家亲测10款AI编码模型，5个任务评分+定价数据，帮你选最值工具。
I Ran 10 AI Coding Models Through 5 Tasks: A Data Scientist's Take I'll be honest — I went into this expecting a clear winner. I came out with a scatt…
9
Faithful by Design: Evaluating and Improving LLM-Generated Clinical Trial Summaries for Multi-Stakeholder Audiences
LLM 生成的临床试验摘要能否兼顾医生与患者需求？这项研究系统评估并改进了多受众场景下的忠实性问题，附有公开方法与分析。
arXiv:2607.09932v1 Announce Type: cross Abstract: Large language models are increasingly used to summarize clinical trial results for healthcare provi…
10
Imaging-101: Benchmarking LLM Coding Agents on Scientific Computational Imaging
新基准Imaging-101系统评估LLM编码代理在科学计算成像上的表现，揭示当前模型能力边界。
arXiv:2607.10789v1 Announce Type: new Abstract: Computational imaging, which recovers hidden signals from indirect, noisy measurements, underpins quan…
11
SpurLens: Automatic Detection of Spurious Cues in Multimodal LLMs
提出自动检测多模态大模型中虚假线索的SpurLens方法，助力提升模型可靠性与可解释性。
arXiv:2503.08884v3 Announce Type: replace-cross Abstract: Unimodal vision models are known to rely on spurious correlations, but it remains unclear to…
12
Capabilities of Claude Fable 5 on Biomedical Challenge Problems
首个公开评估Claude Fable 5在生物医学挑战问题的表现，直面基准饱和与评估方法可靠性问题。
arXiv:2607.10849v1 Announce Type: new Abstract: Frontier language models are increasingly evaluated on biomedical benchmarks, but two problems undermi…
13
Knowledge Distillation for Automated AI Tutor Evaluation
用知识蒸馏技术自动化评估AI导师的教学质量，FATE方法为教育大模型评价提供新思路
arXiv:2607.10647v1 Announce Type: new Abstract: The rapid integration of Large Language Models (LLMs) into K-12 and higher education has outpaced the …
14
Who&When Pro: Can LLMs Really Attribute Failures in AI Agents?
一篇来自arXiv的论文，系统评估大型语言模型在AI代理故障归因任务上的能力边界与不足。
arXiv:2607.09996v1 Announce Type: new Abstract: Automated failure attribution uses LLMs to identify where and why agentic systems fail. As agents beco…
15
Are LLMs ready for HardChoices?
大模型能否应对"HardChoices"？这篇Konvens 2026论文给出了前瞻性测评。
arXiv:2607.11471v1 Announce Type: new Abstract: A lot of research attention has been devoted to checking whether large language models (LLMs) are poli…
16
Eval-Pair Matrix: Answer-Paired Meta-Evaluation of LLM Judges for Grounded RAG
提出Answer-Paired元评估方法，用于衡量LLM评判者在RAG场景下的可靠性，为AI评估体系提供新思路。
arXiv:2607.10626v1 Announce Type: new Abstract: LLM-as-a-judge evaluation is widely used for retrieval-augmented generation (RAG), but reusing the sam…
17
RVN-Bench: A Benchmark for Reactive Visual Navigation
全新基准测试RVN-Bench，专为反应式视觉导航设计，推动机器人自主导航评估标准化
arXiv:2603.03953v2 Announce Type: replace-cross Abstract: Safe visual navigation is critical for indoor mobile robots operating in cluttered environme…
18
Characterising AI Models for Cataloguing
探讨如何通过特征工程和评估方法，系统性地刻画AI模型在目录编目任务中的表现与局限性。
arXiv:2607.11353v1 Announce Type: cross Abstract: The creation of digital collections involves not only the digitisation of content, but also the crea…
19
Turn the Linux GhostLock Report Into a Fleet Patch Plan
别再被安全报告吓到，教你将GhostLock漏洞报告转化为可执行的补丁计划，从系统信息检查开始理性评估风险。
GhostLock, reported as a long-lived Linux kernel stack use-after-free, is trending again today. Do not jump from a research headline directly to “all …
20