每日更新1797 經過驗證的故事
非營利教育團隊以簡單易懂的英語解釋了產品發布、政策轉變、安全研究和行業動向的經過來源檢查的人工智慧報道。
每個故事都連結到最有力的可用證據：可用的原始來源，否則明確歸因的報告。
發生了什麼、為什麼重要以及值得關注的內容——無需行話。
當訊號很弱時，我們不會發布任何內容，而是填充提要。
經過來源檢查的人工智慧故事，最新的優先，適合那些需要了解人工智慧而不追逐炒作的人。
India Today reports that Nvidia’s agreement with AI startup Poolside would give Nvidia access to Poolside’s technology and bring more than 100 employees into its Nemotron open-weight model effort. The report also says Nvidia will invest $1 billion in Poolside at a $12 billion pre-money valuation, while rising…
An arXiv preprint introduces PUMA, a 900-task benchmark for evaluating multimodal AI in Polish cultural and linguistic contexts, reporting strong visual question-answering performance but substantial weaknesses in complex audio and document understanding.
A preprint introduces HiDiffTIR, a reinforcement-learning method that gives different weights to tool-use trajectories and reasoning steps based on their difficulty.
AI News reports that XPENG’s physical AI unit has secured more than $900 million in funding at a $6.3 billion valuation to develop and commercialize its IRON humanoid robot.
A new arXiv paper presents LURE, a zero-data self-play method in which one language model sets task difficulty while another solves verifiable reasoning challenges. The authors report stronger out-of-distribution zero-shot accuracy than trained baselines across nine held-out benchmarks, but the abstract does not…
A paper accepted to EMNLP 2026 introduces EDGE, a training framework that helps language-model agents reuse useful experience instead of relying on external retrieval at inference time. The authors report higher success rates on ALFWorld and WebShop and say the method retained most of its gains after removing the…
Alibaba has officially rolled out Wan3.0, which the company says can generate 30-second videos from documents, spreadsheets, slides and web pages. Reuters reported the release alongside Alibaba’s $10 billion share sale and increased AI spending.
An arXiv preprint introduces NC-GRPO, a reinforcement-learning method that perturbs a vision-language model’s hidden representation rather than its input image. The authors report improved out-of-domain mathematical reasoning and hallucination robustness on Qwen2.5-VL-7B, while noting tradeoffs between…
A new arXiv preprint describes DamageScope, a retrieval-augmented system that combines satellite imagery with vision-language and language models to support natural-language queries about property damage. The authors report up to 14-fold faster indexing and roughly threefold reductions in operational cost and…
Rappler, republishing a Reuters report, says the United Kingdom and Ukraine signed a partnership to develop AI tools for defense and security, including British access to Ukraine’s Avengers AI Labs battlefield-data platform.
An arXiv preprint presents LiteEvent-AE, a compact autoencoder for event-based vision that the authors say cuts model size and energy use while maintaining recognition performance on two constrained devices.
A new arXiv preprint introduces GapSight, a method that lets vision-language models selectively revisit image regions when a low-resolution global view loses important detail. The authors report higher average scores across six benchmarks, while noting that the results come from their own experiments and remain…
每週一次有用的簡報
取得本週經過驗證的人工智慧新聞、原始數據、有用的工具、學習精選和新鮮的人工智慧工作。