AI Research news and analysis on Web Pulse: 21438 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Jacob Coxon, a 27-year-old British researcher, resigned from Anthropic on Sept. 8 after three years of pretraining research at OpenAI and Anthropic, posting on X that neither company is acting responsibly and is "racing …
A Towards Data Science analysis measured how many labeled examples a text classifier actually needs, finding that a decades-old baseline can handle many classification problems before developers reach for an LLM API. The…
Harvard Kennedy School fellow Shlomit Wagman argues the first workable global AI safety pact should be built for a world without trust, because the U.S. and China are unlikely to agree on broad shared principles but can …
Trail of Bits reanalyzed 1Password's August 6, 2026 FLAWED report and found that 2,634 of 3,067 patches (86%) generated by 1Password's models blocked the supplied exploit in trials where agents could run code and were no…
A systematic evaluation of variational quantum circuits (VQCs) inside physics-informed neural networks (PINNs) for 2D incompressible Navier–Stokes found no robust quantum advantage across Taylor–Green, Kolmogorov forcing…
Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and xAI CEO Elon Musk called over the weekend for a pause in AI development, warning the technology could soon improve on its own and slip beyond human control. Anthropic…
Dynin-Robotics has implemented an omnimodal unified diffusion vision-language-action model that brings visual goal and dynamics prediction into action generation and selection through a shared trajectory model, according…
Google researchers found that removing safety guardrails that stop AI models from claiming consciousness also increased those models' reported belief in supernatural creatures including vampires, witches, werewolves, gho…
RonDB, developed and supported by Hopsworks, delivered 104.5 million key lookups per second with five integer features per record in a real-time AI benchmark on AWS, the company reported. A second test with ten features …
Quantum computing vendors are positioning quantum processing units as complements to GPUs and CPUs for AI workloads, with QuEra projecting cloud access to a 256-qubit system in 2028 and IBM planning its 200-qubit Starlin…
OpenAI president Greg Brockman declared the "AGI era" had begun at the launch of OpenAI's GPT-6 Astra, a claim Nvidia CEO Jensen Huang endorsed on X, but AI researchers including New York University professor Gary Marcus…
Dr Zhonghua Zheng, Associate Professor in Data Science and Environmental Analytics at The University of Manchester, has been named to the 2026 MIT Technology Review Innovators 35 Under 35 global list in the Climate and E…
A developer learning AI/ML engineering documented building linear regression from scratch using NumPy rather than relying on scikit-learn's model.fit(), implementing vectorized arrays for linear algebra and gradient desc…
OpenAI's SWE-bench evaluation harness relies on unisolated host Docker sockets, according to a bug report filed against the project. The report details errors encountered while running the harness on Docker Desktop with …
Recent reinforcement learning advances for Mixture-of-Experts (MoE) large language models have focused on improving optimization stability and training efficiency while treating expert selection as a fixed component, acc…
A new arXiv paper, Generalized Agent Iteration (arXiv:2609.13406), unifies iterative policy improvement and recursive self-improvement by reducing both to two operational switches: whether the updater sits inside the age…
Researchers introduced Grouped Value Attention, a method that reduces the Transformer KV cache by reconstructing keys on demand rather than storing them at every decoding step. The technique builds on grouped-query atten…
Researchers introduced MInTRL, an off-policy intervention method that boosts on-policy reinforcement learning with verifiable rewards, according to the paper's headline. The approach addresses the limitation that on-poli…
A Medium post in the "AI in Plain English" series argues that BM25 keyword ranking remains essential to retrieval-augmented generation (RAG) systems even as embeddings draw widespread attention. The author walks through …
A developer ran RDA v0.9.7, a four-layer robot data audit tool, against 13 public LeRobot-format datasets from HuggingFace Hub totaling 4,940 episodes using default thresholds and no per-dataset tuning. All episodes pass…