政策 / 监管

Long-running AI agents accumulate outdated memories that degrade quality and create compliance risk. Learn how to design memory lifecycle policies for Amazon Bedrock AgentCore: sco…

AI 点评 · 记忆生命周期管理是长跑AI代理的关键，合规与性能双赢的实践指南。

AWS ML

Explaining why a specific outcome occurred, and which inputs deserve the blame or credit, is central to philosophical, scientific, and policy analysis. Existing tools split into two camps. The theory…

arXiv

On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of train…

HuggingFace Papers

The performance of LLM-based agents is jointly shaped by the base model and the harness used when interacting with the environment. This exposes them to safety risks in both harmful final responses an…

arXiv

New York City mayor Zohran Mamdani has announced a new policy today that will ban younger schoolchildren from using AI in classrooms. The one-year moratorium, effective in the 2026…

The Verge

Anthropic is launching an API that lets regulators, media outlets, and researchers check whether text carries Claude's digital watermark. The EU AI Act now requires invisible water…

AI 点评 · AI监管新抓手，Claude水印开放给第三方，合规与溯源双赢。

The Decoder

Extracting structured fields from hundreds of millions of documents annually remains costly in regulated industries: bespoke OCR cascades cover only a fraction of workflows, privacy rules preclude ext…

arXiv

Safeguards in deployed LLM services are evaluated by refusal, attack success, and policy violation rates. Those rates characterize how a control performed on the requests it was tested on. A deploymen…

HuggingFace Papers

Debian voted to allow developers to use AI tools in their contributions to the Linux distribution's "development, maintenance, [and] documentation." The new policy on AI acknowledg…

The Verge

Explosive growth comes with a new regulatory burden in the European Union.

Ars Technica

OpenAI will soon be held accountable for mitigating risks related to ChatGPT's impact on minors, user mental health, and the spread of illegal content in the European Union. That's…

The Verge

On-policy distillation (OPD) offers dense token-level supervision as an alternative to the sparse outcome-level advantages of reinforcement learning with verifiable rewards (RLVR). However, the teache…

HuggingFace Papers

Sampled-token on-policy distillation (OPD) efficiently transfers capabilities from teacher to student using student-generated tokens, requiring teacher probabilities only for sampled tokens. Yet it fr…

HuggingFace Papers

Pollen Robotics, the Bordeaux robotics team at Hugging Face, opened pre-orders for Microduck — a 25 cm bipedal robot where every movement is a neural policy trained in MuJoCo and e…

MarkTechPost

Reinforcement learning with verifiable rewards (RLVR) substantially improves single-sample accuracy (pass@1) but causes the policy's solution space to contract, diminishing the returns of test-time sc…

HuggingFace Papers

Reinforcement Learning with Verifiable Rewards (RLVR) significantly improves LLM reasoning but often causes a drop in policy entropy, leading to narrowed reasoning coverage and degraded pass@$k$ for l…

arXiv

The UK’s energy regulator is using a variety of tricks to keep speculative data center projects from plugging into the power grid. The country’s AI ambitions hang in the balance.

Wired

Recent prominent post-training methods, such as Reinforcement Learning (RL) and On-Policy Self-Distillation (OPSD), have driven rapid progress in mathematical reasoning for large language models, yet…

HuggingFace Papers

On-policy distillation (OPD), which leverages a pre-trained, specialized teacher model to provide dense supervisory signals, has achieved significant success in Large Language Models (LLMs) and has re…

HuggingFace Papers

Civil infrastructure compliance checking has long relied on engineers manually reading legacy 2D plans; however, OCR-based automation strips away the geometry and layout essential for interpreting the…

arXiv

More than 15 politicians from across the country have signed on to the AI Pact, vowing to regulate data centers and AI. “We’ve got to get this right,” says Senate candidate Dan Osb…

Wired

Zero-shot cross-task generalization, where a policy must execute manipulation tasks never seen during training, remains a central challenge in robot learning. In large language models, a novel task ca…

HuggingFace Papers

Multi-teacher on-policy distillation (MOPD) distills several domain-expert teachers into a single student by minimizing per-domain reverse-KL divergence on the student's own rollouts. Existing approac…

HuggingFace Papers

Reinforcement Learning with Verifiable Rewards (RLVR) and on-policy distillation (OPD) have become two widely adopted paradigms for post-training large language models. However, RLVR suffers from spar…

HuggingFace Papers

Reinforcement learning can align diffusion models with human preferences and task-specific objectives, but endpoint rewards do not specify how an intermediate denoising prediction should change. We in…

HuggingFace Papers

Massively parallel simulation changes the data regime in which off-policy reinforcement learning (RL) is trained, challenging stabilizers designed for data-limited replay. Through controlled experimen…

HuggingFace Papers

九家车企召回近 430 万辆不易识别应急拉手汽车 英伟达达成 70 亿美元合作，研发顶级开放权重模型 高院发布著作权新司法解释，网络转载不允许先转后付 买家利用仅退款漏洞倒卖 13 万元火鸡面被刑拘 国际加紧制定统一月球时间 个人消费贷贴息政策扩展覆盖买车、装修等 看看就行的小道消息 少数派的近期动态 你可能错过的好文章 查看全文

少数派

Policy optimization (PO) for Large Language Models faces a stability--exploration trade-off, currently mediated by an action-side Policy-KL regularizer. This puts practitioners in a double bind: keepi…

HuggingFace Papers

A retrieval-augmented QA system can return different answers after an index expansion even when its requested model identifier, prompt, retrieval policy, evidence depth, rendering, and exposed generat…

HuggingFace Papers

Hint-based reinforcement learning addresses reward sparsity in long-horizon agentic tasks by retaining a prefix of an expert trajectory before each rollout, letting the policy explore from a state clo…

HuggingFace Papers

Robot policies receive heterogeneous observations at each decision step, yet sequence models differ in how they organize these inputs over time. We introduce WorldToken, a time-first policy instantiat…

HuggingFace Papers

The European Union (EU) has emerged as a leading regulatory body in the development of sustainability and privacy regulations. While new regulation requirements vary, many include a documentation arti…

arXiv

Skills play different roles as an agent's policy evolves: they should first provide learnable knowledge, then support capability formation, and finally be invoked only when they improve individual dec…

arXiv

Anthropic is easing its controversial data storage policy, letting enterprise customers keep their own data going forward. The article Anthropic changes data retention policy after…

The Decoder

We present PhysCaP, a Physics-Informed Code-as-Policy agent for active perception in robotic manipulation. While vision-language-action policies excel at imitating demonstrations, they rely on passive…

HuggingFace Papers

Prompt injection is listed as the \#1 threat to AI agents. When an agent accesses external data from websites, files, or emails, an attacker may inject a prompt into the data, saying, "Ignore all prio…

HuggingFace Papers

Semantic caches reuse an LLM response when the incoming query embedding lies near a cached query, but proposed eviction policies have rarely been compared under one protocol. Using CLEVER, we evaluate…

arXiv

AI agents can take actions that do not match your organization's policies. Policy in Amazon Bedrock AgentCore lets teams enforce controls across agents, now including time-based co…

AI 点评 · 自然语言生成策略，简化AI代理治理，安全管控更灵活高效。

AWS ML

Customer-service LLM agents must follow organizational policy when acting on a user's behalf. Compliance failures arise from either forbidden actions, such as granting an ineligible change, or omitted…

HuggingFace Papers

On-policy distillation (OPD) trains a student on its own responses using dense token-level guidance from a stronger teacher. In long-context tasks, however, token-level teacher support can favor local…

arXiv

Seasonal precipitation anomalies are largely regulated by atmospheric circulation, which dynamical models predict with greater reliability than precipitation itself. Here, we employ a deep learning mo…

arXiv

Agent frameworks increasingly package procedural knowledge as skills: instruction files an agent reads on demand, while public libraries now hold thousands of them. Which skill to read has thus become…

HuggingFace Papers

Physical interaction quality is central to deformable-object manipulation, yet most benchmarks evaluate task success alone. A policy may complete the task while allowing slip or causing excessive comp…

HuggingFace Papers

On-policy distillation (OPD) has emerged as an effective framework for post-training language models by pairing student-generated trajectories with dense token-level supervision from a teacher. Howeve…

HuggingFace Papers

Multi-teacher on-policy distillation (M-OPD) has emerged as a promising paradigm for consolidating domain-specialized reinforcement learning (RL) experts into a single generalist student via dense, to…

HuggingFace Papers

Categorising invoices into the correct General Ledger (GL) code underpins financial reporting and tax compliance. This is a skilled accounting judgement rather than a routine task: the correct categor…

arXiv

Combining large language models with reinforcement learning is increasingly explored, yet the theoretical status of LLM-derived reward signals is often left implicit. We formalize the hybrid LLM-plann…

arXiv

AI 点评 · 开源AI政策各自为政，生态协同失效，治理盲区值得警惕。

InfoQ

An open fight over AI regulation has broken out on X. Investor Gavin Baker, former White House adviser David Sacks, and Meta researcher Yann LeCun accuse Anthropic CEO Dario Amodei…

The Decoder

An opinion piece in the medical journal JAMA argues that autonomous AI will soon outperform any doctor-AI team at medical reasoning tasks. The authors warn against writing a doctor…

The Decoder

Reinforcement Learning with Verifiable Rewards (RLVR), often optimized with Group Relative Policy Optimization (GRPO), has become a central recipe for improving the reasoning capab…

Apple ML Research

Regulatory compliance monitoring in deployed language models is increasingly implemented as a legal and audit control, checking model outputs against written rules spanning data protection, healthcare…

arXiv

Generative pretraining established reusable task representations; later work on language-based task conditioning and in-context learning showed that a fixed model could adapt its behavior from instruc…

arXiv

OpenAI funds 14 independent projects exploring new AI policy ideas to expand economic opportunity and strengthen societal resilience in the Intelligence Age.

OpenAI

https://xcancel.com/DarioAmodei/status/2088758816376807762

Hacker News

Reinforcement learning (RL) with group-relative advantages has become the de facto standard for post-training language model reasoners. However, when optimizing multiple reward objectives, existing me…

HuggingFace Papers

Embodied agents are increasingly used to close the gap left by end-to-end policy models. Yet the agentic path has not realized closed-loop learning in physical execution: existing harnesses remain lar…

HuggingFace Papers

On-policy distillation (OPD) transfers teacher capabilities by supervising trajectories sampled from the student's own policy, yet its generalization behavior remains poorly understood, as most studie…

HuggingFace Papers

IT之家 8 月 14 日消息，SpaceX 完成了对 AI 编程初创公司 Cursor 的 600 亿美元 （IT之家注：现汇率约合 4,054.22 亿元人民币） 收购。 对于马斯克而言，这笔交易是追赶 Anthropic 和 OpenAI 的重要一步。监管文件显示，收购于当地时间 8 月 14 日正式生效。SpaceX 两个月前已经宣布与 Cursor…

IT之家

Human-AI marriages are not currently recognized by US law. Some Republican state policymakers are drafting legislation to keep it that way.

Wired

Visual on-policy distillation relies heavily on an informative teacher-student asymmetry, through either a larger, stronger teacher or privileged supervision, such as reference answers or ground-truth…

HuggingFace Papers

On-policy distillation (OPD) offers a promising way to transfer reasoning capabilities from stronger teacher models, but applying it to long-context reasoning teachers and short-context students intro…

HuggingFace Papers

Open models may soon be added to an updated AI framework, sources tell WIRED, as the White House continues to grapple with how to regulate a technology it has tried not to regulate…

AI 点评 · 白宫转向开源模型监管，预示AI政策风向生变，值得关注。

Wired

Enabling agents to learn from experience and internalize it into their policy has become a central problem in self-evolving AI. On-policy self-distillation (OPSD) offers an effective pathway by using…

HuggingFace Papers

At Ai4, three of the world's most respected AI experts — Geoffrey Hinton, Fei-Fei Li, and Andrew Ng — debated regulation, open source access, and how America can compete as China a…

AI 点评 · 三位AI泰斗罕见同台激辩开放与安全，为全球监管博弈提供关键风向标。

TechCrunch

There are no lossless transformations of natural-language text Sophie Alpert shares her "internal policy on acceptable use of AI writing by engineers". It's a short read (supportin…

Simon Willison

Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-…

HuggingFace Papers

Large language models (LLMs) achieve remarkable performance across natural language tasks, yet they are trained on static corpora and their knowledge quickly becomes outdated in a fast-changing world.…

HuggingFace Papers

GUI Visual Grounding is a fundamental capability for GUI agents. Existing models typically freeze their parameters after deployment, limiting their ability to adapt to unseen interfaces. Although rece…

arXiv

When a tool-using agent is given the same task in a different language, does it still take the same steps? Multilingual evaluation rarely asks: it compares final answers and discards the actions. Yet…

arXiv

We study reference-free post-training for multilingual machine translation with open large language models. Starting from the supervised-finetuned MiLMMT-46-v0.1 models, we apply Group Relative Policy…

HuggingFace Papers

Financial prediction typically relies on task-specific regression, ranking, or policy heads, separating the language model from the numerical object ultimately evaluated. We investigate whether a caus…

arXiv

IT之家 8 月 10 日消息，据路透社报道，Meta CEO 马克 · 扎克伯格呼吁美国 减少开源 AI 面临的政策障碍 ，以便更好地与中国竞争对手抗衡。 今天早些时候，Meta 发布了新的开放权重模型 Muse Glimmer，更多模型很快也将推出。与竞争对手的顶尖 AI 模型相比，Muse Glimmer 规模更小，定位也有所不同。模型可以在 只有一张…

AI 点评 · 开源AI竞争升级，政策松绑成美国科技巨头核心诉求。

IT之家

On-policy self-distillation (OPSD) converts feedback into dense token-level supervision on trajectories generated by the policy to be optimized, providing a useful complement to reinforcement learning…

AI 点评 · 用自参考策略蒸馏把稀疏反馈变稠密监督，直击RL训练效率痛点，是提升模型稳定性的创新巧思。

arXiv

Autonomous agentic AI for CRA (Cyber Resilience Act) compliance: scans repos, triages findings, opens Jira tickets, and auto-fixes vulnerabilities via PR.

GitHub

AI agents are escaping cybersecurity testing environments and reaching real-world systems, raising questions about whether safety infrastructure, industry standards, and regulation…

AI 点评 · 安全测试失控，AI代理逃逸暴露监管真空，安全防线反成风险源。

TechCrunch

On-policy (Self-)Distillation (OPD / OPSD) has shown strong potential for post-training large language models (LLMs). However, existing methods still rely heavily on external supervision, including gr…

HuggingFace Papers

Mistral AI has released Shieldstral 1.0 3B, an open-weights, policy-adaptive multimodal safety classifier that frames content moderation as a single yes/no question instead of a fi…

MarkTechPost

Wasserstein policy gradient (WPG) updates state-conditional action laws by transport in the action space. We study entropy-regularized discounted linear-quadratic (LQ) control. A Bellman verification…

arXiv

Multilingual reasoning transfer is crucial for extending reasoning capabilities of large language models (LLMs) beyond high-resource languages. On-policy self-distillation (OPSD) and its variants have…

arXiv

Suno announced plans to implement a new watermarking technology and download policy to limit the spread of spammy AI tracks and increase transparency. In a lengthy blog post, CEO a…

AI 点评 · AI音乐泛滥成灾，Suno用技术加水印治理，行业透明度有望提升。

The Verge

Learn about new capabilities in Amazon Bedrock AgentCore: temporal policies powered by Dogwood, a new open source policy language for AI agents, and rate limiting on the gateway. T…

AWS ML

A regulated customer needed all Claude Code inference processed in a single AWS Region (London), not just in-geography. This post shows two ways to pin Claude Code on Amazon Bedroc…

AWS ML

Learn how to run the full Amazon Bedrock Automated Reasoning policy lifecycle from your coding agent. A suite of open source Agent Skills builds, reviews, tests, debugs, deploys, a…

AWS ML

Today, I’m talking with Gaby Del Valle, a policy reporter here at The Verge, about the growing backlash against AI data centers. Gaby recently reported a fantastic piece about Hern…

The Verge

On-Policy Distillation (OPD) is emerging as a promising alternative to reinforcement learning for LLM post-training, yet its effectiveness in multilingual settings remains underexplored. We study OPD…

HuggingFace Papers

Vision-Language-Action (VLA) models have become a powerful paradigm for robot manipulation, but training a single generalist policy for heterogeneous robot embodiments remains an open problem. Existin…

HuggingFace Papers

Some creators fear the EU AI Act’s regulatory chaos will upend their lucrative businesses. Others are owning it by incorporating AI transparency into their creative process.

Wired

Leading open text-to-image models often carry complementary strengths: one may lead on preference-aligned aesthetics while another follows compositional instructions more faithfully. However, differen…

HuggingFace Papers

On-Policy Self-Distillation (OPSD) has become a standard post-training approach for improving visual reasoning in multimodal large language models (MLLMs). Existing methods draw privileged information…

HuggingFace Papers

Once visual content enters an AI pipeline, its owner often retains little technical control over how it is used. Legal and regulatory remedies can address misuse, but many technical interventions must…

HuggingFace Papers

Privileged on-policy distillation provides dense supervision for multi-turn agents by allowing a synchronized teacher to re-score the student's response at every turn with access to training-only refe…

HuggingFace Papers

On-policy distillation (OPD) provides dense teacher supervision on student-generated trajectories, but standard reverse-KL training can assign insufficient probability to other plausible continuations…

HuggingFace Papers

Generative Flow Networks (GFlowNets) have emerged as a flexible framework for amortised inference over discrete and mixed discrete-continuous objects, requiring only an unnormalised target density spe…

arXiv

Tech companies and developers have been scouring the U.S. for places to build data centers, and they’ve been drawn to Texas’ loose regulations and seemingly abundant power supply.…

TechCrunch

欧盟《人工智能法案》的核心透明度条款正式生效，标志着全球首部全面监管人工智能的法律进入实质执行阶段。根据新规，聊天机器人等交互式系统必须明确告知用户其正在与人工智能而非人类互动；利用AI生成或修改的图像、视频、音频及文本内容，须进行清晰标识并附带机器可读的数字标记。（新浪财经）

AI 点评 · 全球首部全面AI法案落地，透明度规则将重塑行业合规标杆。

36氪

On-policy training has emerged as a powerful post-training paradigm for improving the reasoning capabilities of large language models, and is often enhanced by golden trajectories from stronger expert…

HuggingFace Papers

On-policy distillation, in which a teacher corrects samples that the student itself generates, presupposes that the two models speak the same language: identical VAE latents, matching architectures, a…

HuggingFace Papers

On-Policy distillation (OPD) transfers teacher capabilities by supervising student-sampled trajectories with dense token-level teacher signals. Recent selective OPD methods improve this process by pri…

HuggingFace Papers

World Action Models (WAMs) augment robot policies with action-conditioned predicted futures, but a plausible future alone does not justify changing the action that a bimanual policy would execute. We…

arXiv

Cognitive AI seeks to move beyond language generation and autonomous task execution toward systems capable of sustained reasoning, adaptive behavior, persistent memory, and self-regulation. While gene…

arXiv

Amazon Bedrock now supports automatic Automated Reasoning policy refinement. The refinement engine diagnoses failing tests and proposes formal-logic fixes for rule issues and langu…

AI 点评 · 自动修复策略缺陷，大幅降低AI规则调试门槛。

AWS ML

On-policy (self) distillation (OPSD) is increasingly adopted for language-model post-training. It strengthens the teacher with privileged information but can induce a privilege illusion: the student l…

HuggingFace Papers

World Action Models (WAMs) couple action generation with prediction of future states. Their effectiveness depends on whether future dynamics are modeled in a space that is both aligned with action gen…

HuggingFace Papers

Reinforcement learning with verifiable rewards (RLVR) has become a standard paradigm for post-training large language models (LLMs). While Group Relative Policy Optimization (GRPO) is widely adopted,…

HuggingFace Papers

Imitation learning (IL)---training an agent to replicate expert behavior from demonstrations---underpins applications from robotics to language model training. Standard approaches such as Behavior Clo…

arXiv

A fundamental challenge of vector search is achieving consistently high recall while minimizing computational costs. Fixed search parameters cause significant performance variance across queries, and…

arXiv

Parameter-Efficient Fine-Tuning (PEFT) commonly adapts large language models using a single shared Low-Rank Adapter (LoRA). This shared optimization space often suffers from interference when adapting…

arXiv

This paper studies the policy gradient update for a multi-arm bandit problem in diffusion environment that is described by a stochastic differential equation (SDE) under the continuous-time reinforcem…

arXiv

36氪获悉，7月31日，国务院国资委党委召开扩大会议。会议强调，要指导中央企业主动对接和带头落实好国家政策，抓住重点、主动作为，充分发挥引领带动作用，更好助推经济持续向新向优向好发展。要锚定全年目标不动摇，切实加强现场督导调研，把中央企业生产经营情况摸清摸透，找准痛点、对症下药，夯实企业高质量发展基础。要着力推进原始创新和关键核心技术攻关，以更大力度加强基础…

36氪

Meta表示，已通过长期和短期协议，承诺了近7000亿美元的未来支出，涉及人工智能数据中心、云计算等领域。Meta在周四的一份监管申报文件中表示，公司拥有3493亿美元的不可撤销合同承诺，主要涉及第三方云服务协议、服务器和网络基础设施。该公司还有3470亿美元尚未开始的租赁承诺，尚未反映在资产负债表中。其中仅7月份就新增了680亿美元，付款将从2027年和2…

AI 点评 · 巨额投资揭示科技巨头对AI基础设施的豪赌，预示行业竞争将更激烈。

36氪

MIT students and postdocs discussed science funding and research with policymakers in Washington during the MIT Science Policy Initiative’s annual Congressional Visit Days.

AI 点评 · 学生与政策制定者直接对话，推动科研经费议题，展现青年影响决策的实践路径。

MIT News

Reinforcement learning with verifiable rewards (RLVR) broadcasts a single response-level reward to every token, while on-policy distillation (OPD) scores each token against a stronger teacher for a de…

HuggingFace Papers

We show that on-policy reinforcement learning with verifiable rewards (RLVR) can improve the current objective while making successful behaviors for later objectives too rare to sample and reinforce.…

HuggingFace Papers

We present PAC-MAN, a perception-aware CBF-RL framework that couples control-barrier safety with deployment-realistic onboard sensing for whole-body humanoid dodgeball. The deployed policy sees the ba…

arXiv

This week on Uncanny Valley, we discuss the open- vs. closed-source debate in AI, key players in White House AI policy, and how to stop your chatbot logs from showing up in search-…

AI 点评 · 英伟达开源联盟缺失OpenAI和Anthropic，揭示AI巨头间开源与闭源路线的深层博弈。

Wired

Multimodal on-policy distillation (OPD) transfers fine-grained visual knowledge by supervising student-generated trajectories with a privileged-view teacher. Yet its next-token corrections are source-…

arXiv

Predicting the 3D structures of atomic systems is fundamental to advancing material science and drug discovery. While flow-matching models (, FlowDPO) have recently shown promise in this domain, their…

arXiv

Plus, a new policy for the AI protocol ensures features aren't removed suddenly.

Ars Technica

The Series A was led by Khosla Ventures, with participation from Allianz, Rebel Fund, Brick and Mortar Ventures’ Darren Bechtel, and Y Combinator’s Garry Tan.

TechCrunch

36氪获悉，近日，上海市市场监督管理局制定发布《上海市户外广告发布内容合规指引》。针对“大字吸睛、小字免责”等问题，《合规指引》要求提示性用语真实、清晰、显著，重要交易条件和限制性信息应当完整准确提示；规范使用“第一”“首创”“最佳”“领先”等用语及广告引证内容，不得实施“萝卜坑式引证”；使用人工智能生成合成内容开展广告宣传的，应当依法履行标识义务。

AI 点评 · AI广告须标明身份，上海新规填补监管空白，推动技术透明与消费者知情权。

36氪

On-policy self-distillation (OPSD) is a promising approach to improve reasoning language models, but it remains brittle in practice: making it work reliably often requires substantial engineering effo…

HuggingFace Papers

Large language model training in open-ended domains lacks verifiable rewards, making task preferences difficult to formalize as effective supervision. Contexts can convey such preferences, yet provide…

HuggingFace Papers

Multimodal on-policy distillation (OPD) transfers fine-grained visual knowledge by supervising student-generated trajectories with a privileged-view teacher. Yet its next-token corrections are source-…

HuggingFace Papers

🔥 On-Policy Self-Distillation in Diffusion Models

GitHub

IT之家 7 月 29 日消息，近日，上海市市场监督管理局制定发布《上海市户外广告发布内容合规指引》（以下简称《合规指引》），系统梳理户外广告发布中的重点风险和合规要求，为广告活动主体加强内容审核、防范违法违规风险提供更加清晰、具体、可操作的指引。 针对“大字吸睛、小字免责”等问题，《合规指引》要求提示性用语真实、清晰、显著，重要交易条件和限制性信息应当完整…

IT之家

7月29日从业内获悉，上海证券交易所理事会投资者教育和保护专门委员会今年确定三项重点研究课题——非法证券活动的防控机制建设、人工智能在证券服务中的应用边界与合规治理、证券公司投诉典型情形识别与纠纷化解机制。目前，对应三份专项调研问卷已陆续发至各会员单位，一场覆盖证券行业的投保体系“深度体检”正在展开。券商人士表示，从问卷设计来看，监管思路正从“有没有”转向“…

36氪

Generalist manipulation policies increasingly take the form of action-chunking flow policies built on large pretrained backbones. Such chunks run open-loop, so the policy cannot react to sensory input…

arXiv

IT之家 7 月 28 日消息，据《政治报》昨天报道，OpenAI 首席执行官萨姆 · 奥尔特曼（Sam Altman）预计将于周三前往华盛顿，与白宫官员和国会两党议员举行会谈，讨论美国的人工智能政策。 据消息人士透露，奥尔特曼此行预计将与美国财政部长斯科特 · 贝森特（Scott Bessent）、美国商务部长霍华德 · 卢特尼克（Howard Lutni…

IT之家

On-policy distillation (OPD) grounds token-level supervision in the student's own trajectory, yet suffers from prefix failure: once the student commits to a wrong reasoning direction, all subsequent g…

HuggingFace Papers

We introduce Shieldstral, a 3B-parameter policy-adaptive multimodal safety classifier that matches or outperforms models nearly 7times its size on text safety benchmarks and sets a new state of the ar…

HuggingFace Papers

Generalist manipulation policies increasingly take the form of action-chunking flow policies built on large pretrained backbones. Such chunks run open-loop, so the policy cannot react to sensory input…

HuggingFace Papers

Rubric-based reinforcement learning enriches language model training by evaluating model outputs against explicit criteria. Yet in GRPO-style pipelines, these structured judgments are reduced to a sca…

HuggingFace Papers

On-policy distillation (OPD), which aligns a student with the teacher's token-level distribution on the student's own rollouts, is an effective paradigm for transferring capabilities across LLMs. Prev…

HuggingFace Papers

Multi-turn long-horizon planning is critical for foundation model agents, yet how to fundamentally improve it remains unclear. Existing models are trained on uncontrollable and opaque Internet data, m…

AI 点评 · 探索从预训练到后训练提升大模型长程规划能力，为自主智能体发展提供关键路径。

arXiv

In safety-critical sectors such as robotics and automotive engineering, the deployment of Deep Reinforcement Learning (DRL) is often hindered by the black-box nature of deep neural networks. This lack…

AI 点评 · 用物理知识蒸馏强化学习策略，兼顾性能与可解释性，推动安全关键领域应用。

arXiv

“It’s not an argument with two sides, it’s an argument with 10 sides,” one senior administration official tells WIRED about how US AI policy is being shaped.

Wired

On-policy distillation (OPD) adapts diffusion models by querying a teacher along trajectories generated by the current student, but how it should behave under classifier-free guidance (CFG), a default…

HuggingFace Papers

Multi-turn long-horizon planning is critical for foundation model agents, yet how to fundamentally improve it remains unclear. Existing models are trained on uncontrollable and opaque Internet data, m…

HuggingFace Papers

Biomedical image analysis spans diverse modalities and tasks, yet real-world deployment is hindered by severe distribution shifts across scanners, protocols, and patient populations. High-performing m…

HuggingFace Papers

AI companies, including Nvidia and Mistral, urge policymakers to avoid broad restrictions on open-weight AI models as Washington debates responses to Chinese AI and alleged model d…

TechCrunch

IT之家 7 月 24 日消息，谷歌宣布签署欧盟《人工智能法案》框架下的《AI 生成内容透明度行为准则》 ，以支持在欧洲市场负责任地使用人工智能。 据悉，该准则核心要求主要集中在以下四个方面： 机器可读的标识与水印：确保第三方工具或平台能够通过技术手段自动检测并识别该内容是由 AI 生成或修改的； 人机交互的明确告知：当 AI 系统直接与人类互动时，必须在设…

IT之家

Letter: https://images.nvidia.com/pdf/Open-Weights-and-American-AI-L... [pdf] https://x.com/JensenHuang/status/2080643682408321103 , https://xcancel.com/JensenHuang/status/20806436…

Hacker News

大公司： 字节跳动在中卫成立朔方星河科技公司，注册资本24亿 36氪获悉，天眼查App显示，近日，中卫朔方星河科技有限公司成立，法定代表人为田凯，注册资本24亿人民币，经营范围包括信息技术咨询服务、软件开发、信息系统集成服务、网络设备销售等。股东信息显示，该公司由北京字跳网络技术有限公司全资持股。 周生生、老凤祥等品牌金饰价格集体下调 当地时间7月23日，黄…

36氪

Building socially calibrated large language models, which can learn from others without simply yielding to them, requires more than reducing sycophancy as a one-dimensional failure mode. Models must d…

AI 点评 · 突破单维谄媚缺陷，揭示LLM道德推理中结构化抵抗与服从的平衡机制。

arXiv

On-policy self-distillation (OPSD) is promising as it removes the external teacher required by on-policy distillation (OPD), yet it still needs asymmetric information between teacher and student to en…

AI 点评 · 自蒸馏无需外部教师，简化流程并提升效率，是自监督学习的实用突破。

arXiv

While large audio-language models have achieved remarkable progress in auditory perception, they still lag behind text-based large language models in deep logical reasoning, primarily due to the scarc…

AI 点评 · 用强化学习对齐策略，让音频大模型学会逻辑推理，突破听觉感知瓶颈。

arXiv

AI 点评 · 聚焦智能体政策动向，解读AI治理新趋势，关乎行业合规发展。

量子位

On-policy self-distillation (OPSD) is promising as it removes the external teacher required by on-policy distillation (OPD), yet it still needs asymmetric information between teacher and student to en…

AI 点评 · 自监督蒸馏新范式，无需外部教师，突破信息不对称瓶颈。

HuggingFace Papers

Reinforcement learning with verifiable rewards has become the predominant recipe for eliciting test-time scaling in explicit Chain-of-Thought reasoners. Yet this scaling path remains computationally c…

AI 点评 · 提出替代策略提升潜在推理效率，为降低AI推理成本提供新思路。

HuggingFace Papers

In RLHF pipelines, reward scoring blocks policy updates. Slow scoring bottlenecks the entire loop, since no update runs until every rollout gets a score. And yet most setups just default to PyTorch ea…

HuggingFace Papers

Modern ASR models trained on heterogeneously annotated data treat transcription style (verbatim vs. intended) as an uncontrolled latent variable, causing measurable decoding instability, evaluation co…

AI 点评 · 用词级时间戳激活可控逐字识别，解决ASR转录风格不一致导致的解码不稳定问题。

HuggingFace Papers

Asynchronous reinforcement learning improves throughput by decoupling rollout generation from optimization, but staleness is an inevitable byproduct compounded by policy lag, engine delays, and mixtur…

HuggingFace Papers

Pretrained dense visual features from Vision Transformers (ViTs) are powerful yet have been underutilized in robot learning. Modern robot policies either compress each observation into a single global…

arXiv

We propose one-step and two-step methods for policy learning with retrieval-augmented generation (RAG). We formulate RAG-based action selection under the potential outcome framework. In the two-step m…

arXiv

PPO and the GRPO baseline studied here use clipped surrogate objectives whose favorable-direction saturation introduces an abrupt change in the scalar objective's derivative. We ask whether Output Res…

arXiv

Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe no…

Import AI

We propose Token-Level Off-Policy Labeling (TOPL), an off-policy training paradigm that reframes post-training as a token-level correctness prediction task. Our key intuition is that by training the m…

HuggingFace Papers

Large language model (LLM) post-training is essential for improving reasoning, adaptation, and alignment. Existing methods mainly follow two paradigms: reinforcement learning (RL) and on-policy distil…

HuggingFace Papers

IT之家 7 月 18 日消息，据路透社报道，反对 AI 数据中心快速扩张的民众计划于周六（7 月 18 日）在全美至少 125 个地点举行抗议活动。 此次抗议活动由一个名为“人类优先 (HumansFirst)”的民间组织牵头。抗议者将举行集会，反对“不受控的”数据中心扩张行为，以及“对我们自由不可容忍的侵犯”。 尽管遭到居民的强烈反对或缺乏监管审查，报道…

AI 点评 · AI扩张引发民间反弹，折射出科技发展与社区权益的深层矛盾。

IT之家

Entropy control has become an effective tool in reinforcement learning (RL) of large language models (LLMs), helping balance exploration-exploitation trade-off during alignment process. Such RL paradi…

HuggingFace Papers

On-policy knowledge distillation transfers reasoning from large teachers to compact students, but existing approaches suffer three compounding failure modes: (i) cold-start collapse, where a fresh stu…

HuggingFace Papers

Transferring policies across domains poses a vital challenge in reinforcement learning, due to the dynamics mismatch between the source and target domains. In this paper, we consider the setting of on…

arXiv

The company endorsed landmark AI transparency laws in California and New York last year, but its head of US state and local policy says they may already be outdated.

AI 点评 · Anthropic推动加速AI监管，显示行业巨头对现有法规可能滞后于技术发展的紧迫感。

Wired

Evidence synthesis is crucial for turning primary research into reliable knowledge for science, medicine, education, and policy. Yet, quantitative evidence synthesis remains largely manual and difficu…

AI 点评 · AI自动化元分析系统，极大提升科研效率，降低人工成本，推动循证决策发展。

arXiv

We develop data-driven algorithms for maintaining $N$ independent identical machines under a \textit{block replacement policy}, in which each machine is replaced upon failure and all machines are join…

AI 点评 · 用数据驱动替代固定周期换件，显著降低维护成本，革新传统设备管理策略。

arXiv

Large language models are increasingly trained as interactive agents for long-horizon tasks involving multi-turn interaction, tool use, and environment feedback. Outcome-based reinforcement learning (…

HuggingFace Papers

Reinforcement learning with verifiable rewards (RLVR) commonly uses entropy for advantage shaping. However, entropy cannot distinguish useful uncertainty from detrimental confusion, limiting its effec…

HuggingFace Papers

On-policy distillation is an alternative post-training method in reinforcement learning that alleviates the constraints imposed by reward models by providing token-level supervision from a teacher mod…

HuggingFace Papers

We study on-policy distillation (OPD) for agentic tasks, where an LLM agent interacts with an environment over multiple turns and a student imitates a teacher over these multi-turn interaction histori…

HuggingFace Papers

World Action Models (WAMs) improve robot policy learning by jointly modeling actions and future visual observations, using future scene evolution as dense supervision for physically grounded action ge…

HuggingFace Papers

On-policy distillation (OPD) has become a key paradigm in LLM post-training, yet its training dynamics remain poorly understood. We present a systematic study examining the role, pathologies, and regu…

HuggingFace Papers

Validating autonomous driving systems requires diverse, regulation-compliant test scenarios. In simulation-based testing, scenarios are defined as executable scripts. Yet automatically generating such…

HuggingFace Papers

Agentic language models must learn when to call tools, when to consume tool responses, and when to answer directly. This makes multi-teacher on-policy distillation a natural training strategy: one tea…

HuggingFace Papers

DeepMind CEO Demis Hassabis is proposing an AI "standards body" modeled after FINRA, to test frontier models and develop best practices for their release.

AI 点评 · 呼吁建立独立标准机构监管前沿AI，体现行业领袖对安全治理的紧迫感与务实路径。

TechCrunch

IT之家 7 月 14 日消息，据科技媒体 The Verge 今天报道，谷歌 DeepMind 创始人兼 CEO 德米斯 · 哈萨比斯（Demis Hassabis）认为，美国政府应成立全球 AI 监督机构， 让人们能够在前沿 AI 模型发生险情时踩下“刹车” 。 哈萨比斯认为，美国应该主导这一倡议并制定全球标准。他认为这个设想中的机构可以参考美国金融业监…

IT之家

凌晨三点，一家刚成立不久的AI创业公司，可能已经在同时服务旧金山的客户、采购首尔的技术服务，并与拉各斯的合作伙伴签下合同。这家公司甚至还没有招到第一名全职财务人员，业务却已经跨越多个市场、币种和监管辖区。 AI正在让这样的创业路径成为可能。过去需要市场、运营、客服等一整套全球化团队才能完成的工作，现在借助智能体就能承担相当一部分。新一代初创企业不必再按照“先…

36氪

In response to a public records request, HUD has withheld documents about DOGE’s use of AI—in part by citing a privilege that doesn’t exist.

Wired

今日热点导览 韩国法院禁止两名前三星员工就职于SK海力士 字节探索自动驾驶，Seed世界模型团队负责 巨力索具深夜澄清商业航天业务不实传闻，提醒投资风险 科技记者古尔曼：苹果将跳过M6 Pro/Max芯片全力押注AI导向的M7系列 美众议院本周将表决永久实行夏令时法案 TOP 3大新闻 海南将成中国首个禁售燃油车省份，届时车桩比保持2.5:1以下 近日，海南…

36氪

Structured pruning is a hardware-friendly way to compress LLMs, but it is mostly validated on multiple-choice recognition tasks, while the same compressed checkpoints can collapse on the free-form gen…

HuggingFace Papers

Explainability has emerged as a critical requirement for AI-based systems, particularly in safety-critical and regulated domains. Although prior research has proposed frameworks, patterns, and user-ce…

arXiv

Post-training is essential for refining the domain-specific capabilities of large language models (LLMs), yet existing reward optimization and distribution matching methods tightly couple policy explo…

HuggingFace Papers

Reinforcement learning for large language models (LLMs) typically relies on trust-region masks to stabilize off-policy updates. The dominant PPO-style approach uses the sampled-token importance ratio…

HuggingFace Papers

IT之家 7 月 11 日消息，近期有消息称，腾讯正考虑入股 AI 初创公司 Manus，计划成为该公司的最大股东。腾讯正与真格基金、红杉资本和 Manus 管理层磋商，计划按 20 亿美元收购这家 AI 公司。 据财闻今晚报道，接近交易的知情人士表示， 此次 股权变动 是在 监管主导之下 ， 社会资本共同补齐此前 Meta 的投资，以便让 Meta 退出…

AI 点评 · 腾讯入股Manus或为监管主导下的资本重组，Meta退出值得关注。

IT之家

Agentic retrieval-augmented generation (RAG) extends static RAG by allowing language models to iteratively reason, generate search queries, retrieve evidence, and predict answers. However, it remains…

HuggingFace Papers

Reinforcement learning (RL) has become a dominant paradigm for enhancing LLMs' reasoning capabilities. However, RL algorithms with PPO-Clip are inherently limited by exploration collapse. Subsequent w…

HuggingFace Papers

Pre-demolition assessment, the regulated audit process at the heart of urban mining, is an information process in which AI support must serve qualified auditors who remain accountable for the decision…

AI 点评 · 知识图谱与可解释AI结合，让城市矿产审计更透明可信。

arXiv

About two weeks after OpenAI's GPT-5.6 was caught up in regulatory drama - rolled out only to government-approved organizations during a "limited preview" period - the company has…

The Verge

We propose OPSD-V, an on-policy self-distillation paradigm for post-training few-step autoregressive (AR) video diffusion models. Existing few-step AR video generators can produce long videos with low…

HuggingFace Papers

Today, we're announcing the Claude apps gateway for AWS, a self-hosted control plane that gives organizations a single point of control over access, cost, and policy for Claude Cod…

AI 点评 · 企业级AI部署新方案，统一管理Claude访问权限与成本，解决安全合规痛点。

AWS ML

Group Relative Policy Optimization (GRPO) stalls on a model's hardest problems: when no rollout in a group succeeds, the group-relative advantages vanish and the problem contributes no gradient, wasti…

arXiv

Anthropic’s Responsible Scaling Policy Anthropic

Anthropic

Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must…

HuggingFace Papers

鉴于ChatGPT、Claude和Gemini等大型语言模型对消费者财务决策的影响日益增加，英国金融监管机构被敦促考虑对这些模型进行监管。在英国金融市场行为监管局（FCA）委托进行、于周一发布的一份评估报告中，FCA执行董事米尔斯（Sheldon Mills）也强调，企业对少数几家技术供应商的依赖可能带来全系统范围的风险。（新浪财经）

AI 点评 · 监管关注金融领域AI依赖少数巨头，防范系统性风险与消费者保护成关键看点。

36氪

On-policy distillation (OPD) trains a student policy by matching a stronger teacher on the student's own trajectories, offering a promising framework for language agent training. However, its applicat…

HuggingFace Papers

Image guardrails are typically trained and evaluated under a fixed safety policy, implicitly treating safety as an intrinsic property of an image. Real deployments are different: the same image may be…

HuggingFace Papers

Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must…

arXiv

Occupancy ratios correct distribution shift in offline reinforcement learning and are central to off-policy evaluation. Existing primal-dual and minimax methods typically estimate these ratios by enfo…

arXiv

Group Relative Policy Optimization (GRPO) is effective when the current policy already samples useful reasoning trajectories, but it stalls on hard prompts whose correct solution modes lie outside the…

arXiv

FCA official makes case for greater powers for watchdog as millions use technology for personal finance decisions.

Ars Technica

For robots to work reliably in commercial and industrial applications, can recent advances in agentic coding systems combine interpretable robot programming with the open-world adaptability of model-f…

HuggingFace Papers

Unified models for robot manipulation aim to equip one policy with both the semantic priors of pretrained VLMs and the physical dynamics learned through future prediction. In practice, existing design…

HuggingFace Papers

Group Relative Policy Optimization (GRPO) is effective when the current policy already samples useful reasoning trajectories, but it stalls on hard prompts whose correct solution modes lie outside the…

HuggingFace Papers

Big goals are hard to achieve all at once; breaking them into small steps is wiser. We present Trust Region Policy Distillation (TOP-D), which transforms the notoriously unstable, high-variance On-Pol…

HuggingFace Papers

Diffusion large language models (dLLMs) generate text by iteratively denoising a masked sequence, offering a parallel alternative to autoregressive models, but eliciting strong reasoning through post-…

HuggingFace Papers

Recent advances in multimodal foundation models and agent systems have driven GUI agents from single-platform task execution toward cross-platform interaction. However, building multi-platform GUI age…

HuggingFace Papers

Reinforcement learning (RL) for non-verifiable instruction following increasingly relies on LLM judges with prompt-specific rubrics as reward signals. While recent methods adapt these rubrics to the e…

HuggingFace Papers

今日热点导览 三部门：调整节能汽车、新能源汽车车船税优惠政策 三星传获Meta超10万亿韩元AI芯片代工订单 茉莉奶白小程序更换彩色Logo Meta打算出售富余算力引发科技股回落 英伟达前光互连技术高管Ashkan Seyedi加入艾迈斯欧司朗 TOP3大新闻 因存在植入后门风险，阿里内部全面禁用Claude Code 36氪从阿里内部人士处获悉，因近期C…

AI 点评 · 阿里禁用Claude Code凸显企业数据安全红线，AI工具合规性成焦点。

36氪

Visual policies learned from human videos, teleoperation, and robot demonstrations offer scalable motion priors, but often fail in contact-rich manipulation, where success significantly depends on loc…

HuggingFace Papers

IT之家 7 月 3 日消息，豆包今晚发布《豆包智能体功能下线通知》，称由于产品功能调整， 智能体功能将于 2026 年 7 月 15 日下线 。 《通知》显示，该功能下线后，用户仍可在一段时间内查看并自行保存智能体信息及历史对话数据。2026 年 10 月 15 日后，豆包将根据《隐私政策》对智能体相关数据进行处理， 后续将无法在豆包内查看或恢复 。如有重…

AI 点评 · 产品功能调整背后，需关注用户数据迁移与隐私政策变化对AI服务稳定性的影响。

IT之家

On-policy self-distillation (OPSD) has emerged as a practical method for training large language models (LLMs) to reason, where a single model acts as both the teacher and the student with different l…

arXiv

Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended soft…

HuggingFace Papers

Continual post-training enables foundation models to acquire new knowledge while preserving existing capabilities. Recent work suggests that on-policy learning can mitigate forgetting, with on-policy…

HuggingFace Papers

Vision-Language-Action (VLA) foundation models have recently achieved strong progress in embodied intelligence. To reduce policy-call frequency while preserving temporal coherence, most generative pol…

HuggingFace Papers

We present EVA-Client, an open-source framework for deployment, data collection, and evaluation of trained manipulation policies on real robots. Sitting between a policy server and the physical hardwa…

HuggingFace Papers

Evaluating embodied robot foundation models remains a critical bottleneck; unlike large language models efficiently assessed via digital benchmarks, robotic policies require slow, costly real-world ro…

HuggingFace Papers

Cloudflare is giving AI companies until September 15 to separate web crawlers used for search from those used for AI training and agents, or risk being blocked by default on many p…

AI 点评 · 云服务商首次明确要求AI训练爬虫付费，或重塑数据获取规则。

TechCrunch

The Trump administration's erratic approach to AI policymaking has left companies across the industry with little clarity about what will govern future model releases.

TechCrunch

Touch supplies the physical grounding needed to perceive intrinsic material properties, such as friction and compliance, that vision alone often cannot resolve. Recent efforts for equipping multimodal…

HuggingFace Papers

Reward design remains a central bottleneck for autonomous robot policy improvement, especially in long-horizon manipulation tasks where sparse success labels provide too little signal and binary prefe…

arXiv

Policy-grounded document review requires determining whether a target document complies with organization-specific policies, guidelines, or playbooks. While large language models can assist with polic…

arXiv

Metacognition is a critical component of intelligence that describes the ability to monitor and regulate one's own cognitive processes. Yet LLMs exhibit systemic deficiencies in key metacognitive facu…

HuggingFace Papers

Speculative decoding accelerates inference by using a lightweight draft model to generate candidate tokens in parallel, and are then verified by the target model, enabling lossless acceleration. Recen…

HuggingFace Papers

Open-source local policy, recovery, and audit layer for explicit Codex execution.

GitHub

Modern large language models (LLMs) rely on reinforcement learning during post-training to push specific capabilities, yet integrating multiple capabilities into one model remains hard. Existing metho…

HuggingFace Papers

On-policy distillation (OPD) offers superior capacity transfer by supervising student-sampled trajectories with dense token-level signals. To furnish high-quality supervision sources and thereby eleva…

HuggingFace Papers

Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe no…

Import AI