Updated daily396 verified stories
Material AI papers, studies, datasets, evaluations, and scientific advances with direct links to the underlying research.
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
What happened, why it matters, and what to watch — no jargon tax.
When the signal is thin, we publish nothing rather than padding the feed.
A growing stream of verified perspectives for people who need to understand AI without chasing hype.
A survey accepted to Findings of EMNLP 2026 catalogs 80 methods for adapting foundation models without human labels, preference data, stronger teachers or executable verifiers. It warns that internal learning signals can either improve a model or recursively amplify its errors.
A new preprint reports that language models encode dialectal English unequally during tokenization, pre-training, post-training and inference, even when paired texts have the same meaning.
A new preprint proposes triggering bias corrections only when evidence of stereotyping accumulates during an LLM’s reasoning, reporting fewer interventions than fixed-interval methods but important accuracy tradeoffs across model types.
A paper accepted to Findings of EMNLP 2026 introduces OmniPhys, a benchmark of 15,246 physics questions and 19,850 images drawn from Chinese educational corpora. It evaluates multimodal AI systems on both physics reasoning and the generation of structured physics diagrams.
A new arXiv preprint introduces DCGC, a masked-diffusion framework that uses an upstream draft to revise reasoning globally instead of allowing early mistakes to propagate.
A new arXiv paper introduces VietAIDetector, an open-source zero-shot tool designed to identify AI-generated Vietnamese text without domain-specific training data. The authors say it supports long and scanned documents and outperforms methods developed primarily for English on out-of-domain evaluations.
An arXiv preprint introduces ReliableRAG, a retrieval-augmented generation framework that scores individual evidence triples for relevance and credibility before constructing multi-step answers.
A new arXiv paper introduces TOPAS, a scheduler that jointly manages cached LLM prefixes and request execution in multi-stage agent workflows. The authors report lower job-completion times than their tested baselines on synthetic workflows and two MetaGPT software-development workloads.
A new arXiv paper describes EmoVec, an inference-time method that injects emotion-specific vectors into language models to adjust emotional intensity while preserving the response’s meaning, fluency and coherence, according to experiments reported by the authors.
A preprint reports that targeting 32 attention heads in LLaVA-1.5-7B reduced hallucinated objects in captions on 400 held-out COCO images.
A new arXiv preprint presents JIT-Agent, a model designed to generate, repair and improve the software harnesses that guide AI agents. The authors report sizable benchmark gains across several language-model families, but the work remains an unreviewed preprint without independent validation in the source.
An arXiv preprint introduces $R^3$, a post-training method that uses free-form language reasoning to guide robotic manipulation policies. The authors report gains on two controlled benchmarks, while leaving real-world performance and the size of those gains unspecified.
One useful briefing each week
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
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 job Submit an AI tool