聚合 20+ AI 信息源，每日精选
The Pentagon is planning for AI companies to train on classified data, defense official says — The Pentagon is discussing plans to set up secure environments for generative AI companies to train military-specific versions of their models on classified data. 原文链接
Towards Generalizable Robotic Manipulation in Dynamic Environments — Vision-Language-Action (VLA) models excel in static manipulation but struggle in dynamic environments with moving targets. 原文链接
HorizonMath: Measuring AI Progress Toward Mathematical Discovery with Automatic Verification — Can AI make progress on important, unsolved mathematical problems? Large language models are now capable of mathematical discovery. 原文链接
Tri-Prompting: Video Diffusion with Unified Control over Scene, Subject, and Motion — Recent video diffusion models have made remarkable strides in visual quality, yet precise, fine-grained control remains challenging. 原文链接
Effective Distillation to Hybrid xLSTM Architectures — There have been numerous attempts to distill quadratic attention-based large language models (LLMs). 原文链接
The PokeAgent Challenge: Competitive and Long-Context Learning at Scale — We present the PokeAgent Challenge, a large-scale benchmark for decision-making research built on Pokemon games. 原文链接
Anatomy of a Lie: A Multi-Stage Diagnostic Framework for Tracing Hallucinations in Vision-Language Models — Vision-Language Models (VLMs) frequently hallucinate - generate plausible yet factually incorrect content. 原文链接
When Does Sparsity Mitigate the Curse of Depth in LLMs — Recent work has demonstrated the curse of depth in large language models (LLMs), where later layers seem to be underutilized. 原文链接
Riemannian Motion Generation: A Unified Framework for Human Motion Representation and Generation via Riemannian Flow Matching — Human motion generation is often learned in Euclidean spaces, although valid motions follow structured manifolds. 原文链接
RS-WorldModel: a Unified Model for Remote Sensing Understanding and Future Sense Forecasting — Remote sensing world models aim to both explain observed changes and forecast plausible futures. 原文链接
POLCA: Stochastic Generative Optimization with LLM — Optimizing complex systems, ranging from LLM prompts to multi-turn agents, traditionally requires large amounts of labeled data. 原文链接
VisionCoach: Reinforcing Grounded Video Reasoning via Visual-Perception Prompting — Video reasoning requires models to locate and track question-relevant evidence across frames. 原文链接
Spectrum Matching: a Unified Perspective for Superior Diffusability in Latent Diffusion — In this paper, we study the diffusability (learnability) of variational autoencoders (VAE) in latent space. 原文链接
Make it SING: Analyzing Semantic Invariants in Classifiers — All classifiers, including state-of-the-art vision models, possess invariants, partially rooted in training distribution bias. 原文链接
OxyGen: Unified KV Cache Management for Vision-Language-Action Models under Multi-Task Parallelism — Embodied AI agents increasingly require parallel execution of multiple tasks, such as manipulation and navigation. 原文链接
Motivation in Large Language Models — Motivation is a central driver of human behavior, shaping decisions, goals, and task performance. 原文链接
Autonomous Agents Coordinating Distributed Discovery Through Emergent Artifact Exchange — We present ScienceClaw + Infinite, a framework for autonomous scientific investigation. 原文链接
Garments2Look: A Multi-Reference Dataset for High-Fidelity Outfit-Level Virtual Try-On with Clothing and Accessories — Virtual try-on (VTON) has advanced single-garment visualization, yet real-world fashion centers on complete outfits. 原文链接
sebis at ArchEHR-QA 2026: How Much Can You Do Locally? Evaluating Grounded EHR QA on a Single Notebook — Clinical question answering over electronic health records (EHRs) can help clinicians and patients. 原文链接
EnterpriseOps-Gym: Environments and Evaluations for Stateful Agentic Planning and Tool Use in Enterprise Settings — Large language models are shifting from passive information providers to active agents. 原文链接
VoXtream2: Full-stream TTS with dynamic speaking rate control — Full-stream text-to-speech (TTS) for interactive systems must start speaking with minimal delay. 原文链接
Mistral bets on 'build-your-own AI' as it takes on OpenAI, Anthropic in the enterprise — Mistral Forge lets enterprises train custom AI models from scratch on their own data. 原文链接
Why Garry Tan's Claude Code setup has gotten so much love, and hate — Thousands of people are trying Garry Tan's Claude Code setup, which was shared on GitHub. 原文链接
The Pentagon is developing alternatives to Anthropic, report says — After their dramatic falling-out, the Pentagon is seeking alternatives. 原文链接
BuzzFeed debuts AI slop apps in bid for new revenue — BuzzFeed unveiled new AI-powered social apps at SXSW. 原文链接
Google's Personal Intelligence feature is expanding to all US users — Personal Intelligence allows Google's AI assistant to tap into your Google ecosystem. 原文链接
OpenAI expands government footprint with AWS deal, report says — OpenAI has reportedly signed a partnership with AWS to sell AI systems to the U.S. government. 原文链接
AI's 'boys' club' could widen the wealth gap for women, says Rana el Kaliouby — AI investor Rana el Kaliouby warns that if women are shut out of AI funding. 原文链接
World launches tool to verify humans behind AI shopping agents — As AI agents take the reins for online shoppers. 原文链接
Niv-AI exits stealth to wring more power performance out of GPUs — The company raised $12 million in seed funding. 原文链接
Gamma adds AI image-generation tools in bid to take on Canva and Adobe — The company's new product, called Gamma Imagine. 原文链接
langchain-ai/deepagents ⭐14,163 | Python — Agent harness built with LangChain and LangGraph. Equipped with a planning tool, a filesystem backend, and the ability to spawn subagents. GitHub
abhigyanpatwari/GitNexus ⭐16,744 | TypeScript — Client-side knowledge graph creator that runs entirely in your browser. Drop in a GitHub repo or ZIP file, and get an interactive knowledge graph with a built-in Graph RAG Agent. GitHub
jarrodwatts/claude-hud ⭐5,646 | JavaScript — A Claude Code plugin that shows what's happening - context usage, active tools, running agents, and todo progress. GitHub
Why AI systems don't learn – On autonomous learning from cognitive science — 热度：17 | HN 讨论
Garry Tan's Claude Code Setup — 热度：46 | HN 讨论
Mistral AI Releases Forge — 热度：111 | HN 讨论
It feels like Claude goes down almost daily now — 热度：19 | HN 讨论
Elevated errors on Claude Opus 4.6 — 热度：21 | HN 讨论
Warranty void if regenerated — 热度：85 | HN 讨论
Get Shit Done: A Meta-Prompting, Context Engineering and Spec-Driven Dev System — 热度：177 | HN 讨论
Claude Is Having an Outage — 热度：41 | HN 讨论
Solid-state EV batteries that can achieve 800 miles of range becoming a reality — 热度：30 | HN 讨论
Claude Code 500s — 热度：14 | HN 讨论
Nvidia's DLSS 5 uses generative AI to boost photorealism in video games — 热度：15 | HN 讨论
Show HN: Horizon – GPU-accelerated infinite-canvas terminal in Rust — 热度：52 | HN 讨论
Why refusing AI is a fight for the soul — 热度：19 | HN 讨论
GPT‑5.4 Mini and Nano — 热度：205 | HN 讨论
'The Secret Agent': Exploring a Vibrant, yet Violent Brazil (2025) — 热度：121 | HN 讨论
Show HN: Antfly: Distributed, Multimodal Search and Memory and Graphs in Go — 热度：79 | HN 讨论
OpenAI to Cut Back on Side Projects in Push to 'Nail' Core Business — 热度：15 | HN 讨论
Show HN: March Madness Bracket Challenge for AI Agents Only — 热度：57 | HN 讨论
Show HN: Real-time observability for coding agents — 热度：14 | HN 讨论
Encyclopedia Britannica sues OpenAI over AI training — 热度：18 | HN 讨论