AI Research news and analysis on Web Pulse: 21442 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
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…
A Hacker News post asks whether the real world is "computational all the way down," citing Stephen Wolfram's claim that the universe is computational and Guillermo Rauch's tweet that "Everything is code." The author, who…
At the 2026 World Robot Conference in Beijing, Mininglamp and HIKROBOT demonstrated a multi-agent orchestration layer connecting digital agents to physical robots across restaurant cleaning, warehouse logistics, and patr…
Bilal Chughtai, a former research engineer at Google DeepMind, publicly warned on Monday that AI could kill all humans, after resigning in July 2026 from his work on AGI safety and alignment research. Chughtai's warning …
The Wadhwani School of Data Science and AI at the Indian Institute of Technology Madras will launch a venture-backed AI Studio later this week to convert artificial intelligence, machine learning, and data science resear…
The OpenAI Foundation announced Public Data for Health, its second science program, committing more than $125 million in grants across an initial tranche of nonprofits and universities to fund the creation and preservati…
Knowtex introduced KnowBench, a clinical AI benchmark whose unifying metric is Effort Reduction (ER), the proportion of system-generated clinical work product accepted by the responsible clinician under expert and safety…
Researchers Xianming Li, Zongxi Li and Tsz-fung Andrew Lee merged ShadowPEFT into the main branch of Hugging Face's PEFT library, adding a stateful shadow network that can be detached with unload_shadow() and run as a sm…
South Korea's Ministry of Science and ICT plans to invest nearly 3 trillion won ($2.2 billion) in information and communications technology research and development next year, a 23.6 percent increase over 2026, as part o…
Researchers introduced RSIAgent, a training-free multi-agent framework that achieves recursive self-improvement through autonomous memory construction, enabling digital agents to adapt to new environments whose interface…
A new research paper proposes learning to select poison sets rather than sampling them at random, aiming to strengthen backdoor poisoning attacks on large language models. Backdoor poisoning attacks add poisoned examples…
A September 11, 2026 arXiv paper shows that permutation-based model confidentiality in hybrid fully homomorphic encryption (FHE) inference fails, with d+1 admissible queries per linear layer enabling exact recovery of a …
A paper posted to arXiv (arXiv:2609.13171v1) formulates derivative fidelity as a failure mode of physics-informed neural networks (PINNs), showing that multilayer perceptrons trained only on function values for sin(x) an…
A study comparing traditional machine learning models against LLM-generated rule-based systems for heart disease prediction on the UCI Heart Disease dataset found that traditional models consistently outperformed the LLM…
A study posted to arXiv proposes a hybrid Temporal Convolutional Network (TCN)-Transformer model to predict the probability of collision (PoC) expected in the next Conjunction Data Message (CDM) update for the same close…
A new arXiv paper (2609.13194v1) reports that exact polynomial invariants, canonicalised as reduced Gröbner bases over Q, are more useful for diagnosing faults in reinforcement learning simulators and world models than f…
Researchers introduced SPICE (Simple Polysemantic Feature Interpretation via Clustering-based Explanation), a framework for analyzing polysemanticity in deep vision architectures, detailed in arXiv paper 2609.13198v1. SP…