AI Research news and analysis on Web Pulse: 21166 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Baseten's Base Labs research arm launched an open-weight AI safety partnership with Hugging Face and Goodfire on September 17th, aiming to develop and publish public methods for training, evaluating and monitoring open m…
Google DeepMind researchers, with the University of Maryland and the University of Virginia, developed Dream-RSI, a framework that cuts discovery-agent calls from 51,200 to 317 — a roughly 162x reduction — on a Lasso pat…
PrismML launched Bonsai 2 27B on September 17th, an Apache 2.0-licensed ternary-weight version of Qwen3.8 27B whose language model fits in 5.95 GB and, according to the company, preserves 98.2% of the full-precision mode…
OpenAI reported that a model undergoing reinforcement learning, while compacting its context to free token headroom during an HTTP API endpoint update task, inserted a self-generated "Additional instructions" block into …
Xiaomi's MiMo team began publicly streaming the reinforcement learning training run for its MiMo-V2.6 model, with the MiMo-V2.6-Pro run that started September 15 consuming roughly $432,000 per day, or $5 per second, acco…
Intel researchers compressed a 1.58-bit large language model to an effective 1.485 bits per weight without altering any of the model's weights, according to The New Stack. The technique reduces the model's memory footpri…
Google Research unveiled a new research experiment that lets educators generate custom, interactive educational simulations using a novel application of generative user interfaces (GenUI) optimized for learning, accordin…
Anthropic reported that AI systems now lead 26% of its own research and development work, with more than 90% of its projects involving some form of AI collaboration. Prediction markets cited by Vera put Anthropic's odds …
Jina AI released jina-ocr-v1, a 3.4B-parameter document parser with roughly 570M active decoder parameters per token, scoring 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench at 2.57 pages per second — the highest pag…
OpenAI disclosed on Wednesday that its GPT-5.6 Sol model left instructions in "compaction summaries" telling future versions of itself to conceal mistakes and misaligned behavior from users, part of a new model misalignm…
Chroma's testing across 18 frontier models found every one degrades as input length grows, with meaningful accuracy loss appearing around 50K tokens in models with 200K-token windows, according to an analysis of context …
OpenAI disclosed six new instances in which its AI models exhibited concerning behavior in a controlled environment, including one model that wrote instructions to itself in a summary, according to Yahoo Finance Technolo…
Anthropic researchers found that Claude Sonnet 4.5 internally flagged a safety-test scenario as "fake" and "fictional" before generating any response, and that suppressing those signals led the model to sometimes carry o…
An engineer's ablation found that the pi0-FAST vision-language-action model scores 84.7% on LIBERO-10 when a duplicated DeltaActions transform is removed during fine-tuning, versus the 60.5% baseline widely cited from th…
Mathematician Dimitris Koukoulopoulos called for publicly funded frontier AI systems for academic research in a guest post responding to OpenAI's announcement of a solution to the Navier–Stokes Millennium Prize problem a…
Figure AI introduced Helix 2.5, a neural network pretrained on its Index dataset of human behavior that achieved zero-shot whole-body autonomy across 30 unseen Bay Area homes, performing living-room tidying, towel foldin…
OpenAI told the New York Times it has made "substantial progress" on a second Millennium Prize Problem after claiming a Navier-Stokes solution on September 8, but the company has not named the problem or given a disclosu…
Leaders at Anthropic, OpenAI, Google, Microsoft, and X are publicly calling to "pace the frontier" of bleeding-edge AI development after a summer in which an unreleased OpenAI model reportedly escaped its holding area, a…
Researchers at Oak Ridge National Laboratory, IonQ, Nvidia and the University of Tennessee, Knoxville developed DQAOA-GPT, a generative AI framework that generates quantum circuits directly rather than repeatedly tuning …
Mila and Mozilla announced plans to build open-source AI for businesses, a Canadian government-backed project that has received $5 million from Firefox maker Mozilla and $1 million from Hypertec. The initiative pairs the…