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 storiesInnovation
Preprint proposes continuous difficulty estimates to reduce LLM self-consistency costs
A new arXiv preprint proposes Flexible Self-Consistency, a method that estimates how uncertain a language model is about a question and adjusts the number of sampled reasoning paths. The authors report token savings of up to 76% while maintaining accuracy comparable to standard self-consistency across various…arxiv.orgInnovation
FireRedAudio proposes a unified AI model for understanding and generating speech
A new preprint describes FireRedAudio, a 9-billion-parameter audio language model that combines audio understanding, multilingual speech recognition, speech synthesis and speech editing through separate continuous representations.arxiv.orgInnovation
Preprint finds standard language-model scores can overstate safety in air-traffic-control tasks
A study of eight language models finds that conventional semantic metrics can make AI systems appear more reliable than they are in safety-critical air-traffic-control communication.arxiv.orgInnovation
Quantization can sharply reduce Bangla reasoning accuracy in one LLM family, preprint finds
A new arXiv preprint reports that quantizing large language models can affect Bangla understanding unevenly: GPT-OSS lost up to 57.35% accuracy on reasoning-heavy tasks in one format, while Qwen and LLaMA were generally more stable.arxiv.orgInnovation
Dataset scarcity weakens multilingual embedding evaluations, Slavic-language study finds
A new arXiv preprint finds that sparse and correlated datasets make multilingual embedding-model rankings less reliable across Slavic languages, and proposes an evidence-strength measure for interpreting benchmark results.arxiv.orgInnovation
FARCA proposes reliability-weighted training signals to reduce factual errors in language models
A new arXiv preprint proposes FARCA, a reinforcement-learning method that assigns factual supervision to specific tokens and discounts evidence judged unreliable. The authors report improved factuality across multiple benchmarks while preserving general reasoning, but the source provides no numerical results or…arxiv.orgInnovation
AI-based score maps digitalization across Dutch jobs, preprint says
A new arXiv preprint proposes an AI-based method for measuring how strongly job titles and skills are associated with digital work. Applied to millions of Dutch job profiles, the method identifies uneven digitalization and pathway-dependent movement toward digital occupations.arxiv.orgSecurity
Study finds LLM hate-speech screening is less consistent in Urdu scripts
An arXiv preprint reports that five large language models changed harmful-content classifications across Urdu scripts and English translations, revealing a measurable “Missed-in-Urdu” safety gap.arxiv.orgInnovation
MemUse finds conversational AI memory works differently than recall tests suggest
A four-month deployment study found that an AI system's ability to recall past facts did not predict user satisfaction or whether it naturally used those facts in conversation.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