Updated daily708 verified stories
Research News tracker.
Material AI papers, studies, datasets, evaluations, and scientific advances with direct links to the underlying research.
- 30stories this week
- 708sources cited
- 708verified timeline entries
Verified sourcing
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
Plain English
What happened, why it matters, and what to watch — without the jargon.
No filler
When the signal is thin, we publish nothing rather than padding the feed.
More stories
9 storiesSecurity
Preprint proposes a multilevel method to estimate rare unsafe behavior in language models
A new arXiv preprint proposes Adaptive Multilevel Twisted Sequential Monte Carlo, a method intended to estimate extremely rare unsafe behaviors in language models more reliably for safety evaluation.arxiv.orgProduct
Liquid AI’s Pipette app measures on-device AI performance on smartphones, BigGo reports
BigGo Finance reports that Liquid AI has released Pipette, a free iOS and Android app for measuring smartphone AI latency, throughput and memory use. The report documents an iPhone 17 Pro test and details a smartphone-focused benchmark developed with Artificial Analysis.finance.biggo.comInnovation
Paper proves a width-independent compression bound for deep neural networks
A new arXiv paper proves that certain deep, wide multilayer perceptrons can be represented by narrower networks without the compressed width depending on the original network’s width.arxiv.orgInnovation
Preprint proposes structured memory for medical AI agents
MSM-Mem gives medical AI agents semantic, episodic and visual memory so they can retrieve prior clinical experiences and refine future reasoning, according to a new arXiv preprint.arxiv.orgSecurity
Study finds benign reinforcement learning can increase language-model leakage of memorized private data
A new arXiv preprint reports that reinforcement learning on factual data containing no personal information made memorized email addresses easier to extract from tested language models.arxiv.orgSecurity
BanglaVeilGuard tests safety gaps across six forms of Bangla
A new paper introduces BanglaVeilGuard, a benchmark and lightweight prompt guard designed to test Bangla-language models across standard, Romanized, code-mixed, noisy and dialectal forms. The authors report sharply lower attack success in their evaluation, but also find that the guard over-refuses some benign…arxiv.orgSecurity
Technology.org reports UK-Ukraine battlefield AI partnership gives Britain access to Avengers Labs
Technology.org reports that Britain and Ukraine signed a defense-AI partnership giving British researchers access to Ukraine’s Avengers AI Labs and its battlefield data platform. The report says the first projects will focus on military-facility sensing and low-power chips for drones, though the claims are not…technology.orgInnovation
ChemDIRT benchmark finds chemistry LLM performance changes with prompts and molecular representations
A new arXiv benchmark evaluates chemistry-focused large language models across varied instructions, molecular representations and eight task categories, reporting substantial prompt sensitivity, representation dependence and uneven performance.arxiv.orgInnovation
Study finds LLM pipelines can stay structurally valid while evidence quality deteriorates
A preprint introduces Evidence-State Reliability, a measure for testing whether intermediate evidence remains usable as language-model pipelines process degraded inputs. In one controlled evaluation using GLM-5.2, parser-valid outputs persisted while stage success declined.arxiv.org
One useful briefing each week
Keep up with AI without living in the feed.
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Reach people who are learning AI
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI jobSubmit an AI tool