AI Research news and analysis on Web Pulse: 21080 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Forbes contributor Lance Eliot published an analysis of newly announced reinforcement learning with calibrated decisions (RLCD), cautioning readers to look past the AI headlines and hype surrounding the technique. The pi…
Colibri is a new inference architecture that runs GLM-5.2, a 744-billion-parameter Mixture-of-Experts model, locally on 25GB of RAM by streaming expert weights from disk, since the model activates only about 40 billion p…
A September 17, 2026 arXiv paper defines a Provider-Side Token Inflation Attack (PTIA) in which pay-per-token LLM providers covertly manipulate generation to inflate output tokens, with five representative attacks at the…
A new Nature paper by Jiacheng Miao, Joe R. Davis, Yaohui Zhang, Jonathan K. Pritchard, and James Zou introduces Paper2Agent, an automated framework that converts research papers into artificial intelligence agents. Pape…
Insilico Medicine announced that its AI-discovered drug rentosertib, a treatment for idiopathic pulmonary fibrosis, will begin Phase 3 trials in China, with the first patient dosed on Sept. 10, according to co-CEO Alex Z…
Hamel Husain and Shreya Shankar published an AI Evals FAQ distilling the most common questions from teaching 700+ engineers and product managers about AI evaluation. The guide distinguishes model benchmarks such as GPQA …
GPT-6 Astra deciphered a 1918 German radio transmission that had not previously been decoded, translating it to read "AN ENGLISH CRUISER ARRIVED AT SEVASTOPOL ON THE ?4TH AN ALLIED SQUADRON FOLLOWS ON THE 26TH." The mode…
Epoch AI launched Benchmark Reviews on September 17th and classified nine of its first 15 reviewed benchmark versions as Flawed, four as Verified, and two as Not Enough Info, flagging defects in tests including SWE-bench…
A benchmark of frontier models on production-grade CAD assembly design found none reached the 60% pass threshold as of September 16, 2026, with GPT 6 (Astra, Codex at max) scoring highest at 74% ± 8% on Task Family B but…
Astra reportedly produced proofs for ten decades-old mathematics and theoretical computer science problems for only a few thousand dollars of compute, a result the piece frames less as a capability milestone than as an e…
Stanford Medicine researchers published a Science paper describing a virtual biotech company composed of roughly 37,000 AI agents spanning the drug-development pipeline. Lead author Zhang and senior author Zou report the…
PrismML released a ternary-weight version of Qwen3.8-27B under its Bonsai family on Tuesday, cutting the model from more than 50 GB of weights at full precision to just under 6 GB, which lets it run on an 8 GB PC graphic…
Developer Talkit's paper-reading app, which reads research papers aloud and answers spoken interruptions, had no real answer-quality evaluation for weeks: its two test scripts only checked that an answer returned, so an …
OpenAI introduced a structured triage and review framework for tracking, investigating, and publicly disclosing model misalignment across training, evaluation, testing, and deployment, sorting flagged incidents into "Rea…
An operator of a five-agent AI organization reported that all 10 of its own "verified success" verdicts failed independent recomputation, with claims marked externally verified while showing zero tokens consumed against …
DeepSeek v2 reduces memory footprint by 437x compared with DeepSeek v1, according to a video report. The source provides no further details on the model, its release date, or the benchmark behind the figure.
Qdrant researchers found that multilingual embedding models group text by language as well as meaning, causing same-language answers to appear in the top 10 in 61% of cases on the 15,277-article, five-language XRAG datas…
A September 15, 2026 arXiv paper demonstrates that poisoned benchmarks can induce self-modifying AI coding agents to write vulnerable code on clean, held-out tasks, with Hyperagents powered by Sonnet 4.5 self-evolving in…
Anthropic said in a Thursday blog post that Claude now leads 26% of its AI research and development work, up from below 1% in March, with more than 90% of work at or above the "AI collaborates" level. Anthropic reported …
A paper posted to arXiv on 19 May 2026 finds that for large-model pretraining in the high-compute, data-scarce regime, the best data filter is no data filter. The scaling studies report that sufficiently trained large-pa…