🤖 AI 资讯日报 — 2026-06-29
数据采集时间: 2026-06-29 08:32 UTC+8
共收录 90 条资讯
📡 T1 官方一手信源
Simon Willison
Apple
- Machine Learning ⭐0.98545
- At Apple, we believe privacy is a fundamental human right.
xAI
Hugging Face
OpenAI
NVIDIA
GitHub
OpenAI
- Codex blog posts ⭐0.9846
- Search the blog. Search docs. Suggested. responses create reasoning_effort realtime prompt caching.
Hugging Face
Apple
Simon Willison
- Simon Willison's Weblog ⭐0.98219
- This is a bad state of affairs. Consider, in particular, some industry dynamics: Frontier models are trained at an enormous cost.
GitHub
xAI
- Introducing /goal ⭐0.98196
- Use /goal for long-running autonomous task execution in Grok Build.
NVIDIA
Apple
GitHub
NVIDIA
Hugging Face
Simon Willison
xAI
OpenAI
Simon Willison
xAI
- Grok for Word ⭐0.97575
- Use the Grok add-in for Microsoft Word to turn notes into documents.
Hugging Face
Apple
GitHub
NVIDIA
OpenAI
Hugging Face
Apple
Simon Willison
xAI
OpenAI
GitHub
NVIDIA
Anthropic
📰 T1.5 媒体报道
HackerNews
ITHOME
THE DECODER
🔬 T2 社区与学术
HF Daily Papers
- The Verification Horizon: No Silver Bullet for Coding Agent Rewards ⭐41
- A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabil...
- JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting ⭐31
- Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft bu...
- GUI vs. CLI: Execution Bottlenecks in Screen-Only and Skill-Mediated Computer-Use Agents ⭐28
- Computer-use agents can execute software tasks through either graphical interfaces or programmatic command interfaces, but existing evaluations confound interaction modality with differences in tasks,...
- Running the Gauntlet: Re-evaluating the Capabilities of Agents Beyond Familiar Environments ⭐17
- As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically bu...
- Why Multi-Step Tool-Use Reinforcement Learning Collapses and How Supervisory Signals Fix It ⭐16
- Tool use enables large language models (LLMs) to perform complex tasks, and recent agentic reinforcement learning (RL) methods show promise for enhancing model capabilities. However, RL alone often le...
- LISA: Likelihood Score Alignment for Visual-condition Controllable Generation ⭐13
- The prevalent dual-branch paradigm, i.e., training a side network to encode visual conditions and fusing its intermediate-layer features to a frozen pretrained main network, has shown remarkable succe...
- Discretizing Reward Models ⭐10
- Despite their widespread use, the role of reward models in shaping reinforcement learning is poorly understood. Reward models offer a tempting promise: they automatically estimate response quality in...
- Information-Aware KV Cache Compression for Long Reasoning ⭐9
- Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression m...
- Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents ⭐8
- Process reward models enable fine-grained, step-level evaluation of LLMs, yet building them for agentic settings remains prohibitively difficult: long-horizon interactions, irreversible actions, and s...
- CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies ⭐8
- As LLM agents become capable of increasingly long-horizon tasks, evaluating their performance in economic systems is becoming increasingly important. Unlike existing benchmarks that primarily evaluate...
- Hallucination in World Models is Predictable and Preventable ⭐8
- Modern generative world models render increasingly realistic action-controllable futures, yet they frequently hallucinate: rollouts remain visually fluent while drifting from the ground-truth dynamics...
- ABACUS: Adapting Unified Foundation Model for Bridging Image Count Understanding and Generation ⭐4
- ABACUS is a unified vision-language model that handles object counting, crowd counting, referring-expression counting, and count-faithful image generation without any benchmark-specific training requi...
- When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models ⭐3
- Multi-model LLM systems such as routing, voting, cascades, fusion, and mixture-of-agents are used to beat single-model accuracy. We show that their gain is capped by a quantity the field rarely report...
- How Post-Training Shapes Biological Reasoning Models ⭐3
- Scientific reasoning models for biology combine language models with foundation models trained on multimodal biological data, including DNA, RNA, and proteins. These models are built through post-trai...
- EO-WM: A Physically Informed World Model for Probabilistic Earth Observation Forecasting ⭐2
- Earth Observation (EO) forecasting aims to predict future Earth surface dynamics from satellite observations under changing meteorological conditions. In this paper, we view this task as a partially o...
ArXiv
- [[2409.02668] Introduction to Machine Learning](https://arxiv.org/abs/2409.02668) ⭐0.98574
- This book introduces the mathematical foundations and techniques that lead to the development and analysis of many of the algorithms that are used in machine
- [[2604.15821] Breaking the Training Barrier of Billion-Parameter Universal ML Interatomic Potentials](https://arxiv.org/abs/2604.15821) ⭐0.98369
- Deployed across two Exascale supercomputers, our code attains a peak performance of 1.2/1.0 EFLOPS
- arXiv.org e-Print archive ⭐0.97881
- arXiv is a free distribution service and an open-access archive for nearly 2.4 million scholarly articles
- Unlimited OCR Works Welcome the Era of One-shot Long-horizon Parsing ⭐0.97236
- Recently, end-to-end OCR models, exemplified by DeepSeek OCR, have once again thrust OCR into the spotlight.
- [[2412.17643] Advances in Machine Learning Research Using Knowledge Graphs](https://ar5iv.labs.arxiv.org/html/2412.17643) ⭐0.96717
- The study uses CSSCI-indexed literature from the China National Knowledge Infrastructure (CNKI) database.
- [[2605.27923] Do We Really Need Quantum Machine Learning?](https://arxiv.org/abs/2605.27923) ⭐0.96649
- A feature count of 10 qubits and a sample size in the range of 200-500 emerge as practical operating points.
- [[2606.06473] MLEvolve: A Self-Evolving Framework for Automated ML Algorithm Discovery](https://arxiv.org/abs/2606.06473) ⭐0.96211
- Large language model (LLM) agents are used for automated machine learning algorithm discovery.
- Machine Learning in Biomechanics: Key Applications and Limitations ⭐0.95456
- This chapter provides an overview of recent and promising Machine Learning applications in pose estimation, feature estimation, event detection.
- Position: The AI and ML Community Should Adopt a More Transparent Peer Review Process ⭐0.94967
- This position paper advocates for a more transparent, open, and well-regulated peer review.
- [[2504.00709] Science Autonomy using Machine Learning for Astrobiology](https://arxiv.org/abs/2504.00709) ⭐0.94704
- In recent decades, artificial intelligence (AI) including machine learning has advanced astrobiology.
KOL Twitter
- INS贴文账号哪里有（购买网址CXzhan.com） ⭐0.48323447
- Real-time posts from X ・ meng@shao__meng 聚焦AI 工具普及与产品体验,内容偏实用、面向普通用户和工具使用者。thinkingjimmy@thinkingjimmy ・ 适合关注工具构
- 嘉兴高级外围上门资源 ⭐0.44869256
- 一句话介绍：一台放在家里，和你一起成长的个人AI 系统。Oki Home 插电、扫码即可使用。
- 武汉高级资源外围大学生上门 ⭐0.44637465
- 字节跳动Seed：字节AI 研究团队官方信息源，适合跟踪前沿模型方向。
- 大连高级资源外围大学生上门 ⭐0.42076674
- 它揭示了当前AI发展的两大阵营：一方是科技巨头主导的中心化模式，另一方是区块链赋能的去中心化革命。
- 德阳高级小姐 ⭐0.41563448
- Sahara AI 是一个为AI开发和AI数据/模型资产化提供基础设施的AI链。
- 南宁高级小姐 ⭐0.40500212
- Sahara AI 是一个为AI开发和AI数据/模型资产化提供基础设施的AI链。
- 海口高级资源 ⭐0.3983836
- 这是AI从聊天工具变成基础设施的关键。n8n：测试所有平台后最好的选择。
- 嘉兴高级资源 ⭐0.38131228
- Claude Code 是目前最强的AI 编程助手，但注册、订阅、使用的每一步都暗藏风控雷区。
- 海口资源（高级小姐） ⭐0.36772484
- Claude Code 是目前最强的AI 编程助手。
- 昆明资源（高级小姐） ⭐0.358174
- Claude Code 是目前最强的AI 编程助手。
由 Hermes Agent 自动生成 | 2026-06-29