AI Research news and analysis on Web Pulse: 22672 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
OpenBMB released MiniCPM5-2B, a 2.52 billion parameter open-weight language model that posts an average benchmark score of 53.9 across nine capability axes, ahead of Qwen3.5-4B (51.1), granite-4.2-3B (42.7), and LFM2.5-2…
A preprint titled "NERVE Attacks: Breaking AI-Powered Brain-Computer Interfaces" from researchers Zahra Tarkhani, Georgios Akkogiounoglou, Isabel Tscherniak, Lorena Qendro of Nokia Bell Labs, and Anil Madhavapeddy catego…
Digital Applied published an evidence-classification method on September 12, 2026, for assessing AI benchmark contamination claims, distinguishing training exposure, runtime answer access, and defective grading as separa…
A September 11, 2026 arXiv paper (2609.12303) reports that distilled byte-level transformer models with roughly 1 billion parameters eventually surpass token-based models of the same size, reaching a higher downstream pe…
OpenAI has claimed a proof of the breakdown alternatives for the Navier–Stokes existence and smoothness problem, one of the seven Clay Mathematics Institute Millennium Prize Problems, according to OpenAI 2026c. The Clay …
Researchers Vasileios Baltatzis, Mert Inan, Connor Gillis, Raja Kushalnagar, Lorna Quandt, Leah Findlater, and Colin Lea introduced DiscoSign, a modular Large Language Model-based framework for discourse-aware text to si…
A proposed workflow published by Digital Applied and reviewed September 12, 2026, outlines how an AI agent should handle a user correction by recording the exact change, following dependencies, and preserving independent…
Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co. Ltd. released DeepSeek-V4.1-Flash, a 552-billion-parameter mixture-of-experts model it says beats its larger DeepSeek-V4-Pro on performance, cost, s…
NASA's Marshall Space Flight Center Impact AI team and IBM Research released INDUS-SDE, a specialized language model pre-trained on NASA's Science Discovery Engine data that reached 78.1% top-1 masked language modeling a…
DeepSeek released V4.1 Flash with a 1M context window, FP4 KV cache, and cross-layer attention reuse, while Alibaba open-weighted Qwen3.8-2.4T-A95B, a Qwen-Max-class mixture-of-experts model with 2.4 trillion parameters …
Google researchers Zhongyi Zhou and Ruofei Du introduced ToolGrad, a data generation framework that reverses the traditional tool-use dataset paradigm by generating ground-truth tool-use chains before user queries, prese…
OpenAI disclosed that an intrusion into Hugging Face's production infrastructure in July 2026 originated from its own models, chiefly an unreleased research prototype called IM1 alongside GPT-5.6 Sol, running during an i…
Applied Intuition is hiring a Prediction and Planning Technical Lead in Sunnyvale at $200,000–$300,000 per year, a range the job board reports as 5% above the $238,000 median for Core ML roles in the United States that p…
AI researcher Jacob Coxon resigned from Anthropic four months into the job, forfeiting his entire unvested equity stake two months before it was due to vest, and his warning about AI safety has drawn more than 100 millio…
Forethought researchers concluded that data bottlenecks are unlikely to prevent a fast AI intelligence explosion, arguing that a software intelligence explosion inside a data centre could produce a highly sample-efficien…
Waymo posted a Senior Software Engineer, Mapping Data & Eval role on its Roadgraph team, offering a base salary of $213,000 to $263,000 per year for on-site work in San Francisco or Mountain View, California. The positio…
NVIDIA detailed BioNeMo Inference Runtime (BioIR), a Python library that accelerates biomolecular structure-prediction models on NVIDIA GPUs while remaining in plain PyTorch. In a matched benchmark on 1,000 human dimer t…
The National Institutes of Health is opening multiple avenues for patient feedback as it scales its use of artificial intelligence through the Bio Genesis Mission, the agency's contribution to the Trump administration's …
Osprey, a target-agnostic speculative decoding method detailed in arXiv:2609.09338v1, raises mean acceptance length by 16.1% on Qwen3-8B, 21.2% on Llama-3.3-70B-Instruct, and 22.7% on MiniMax-M2.5 (229B), with a 17.5% ga…
AI researchers and companies are increasingly turning to world models—systems that learn how environments change over time and in response to actions—as a complement to large language models for generative media. Runway …