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
MARS paper reports higher competitive-programming pass rates by routing tasks among specialist LLMs
A new preprint describes MARS, a prompt-only system that assigns programming problems to topic-specialist language models and reports a 14.4-percentage-point gain over direct prompting on CodeContests.arxiv.orgInnovation
Benchmark finds vision-language models struggle to identify ambiguous targets through dialogue
A new benchmark reports that large vision-language models perform well below human baselines when they must identify visual targets through incomplete descriptions and interaction.arxiv.orgIndustry
The Daily Beacon reports UT Research Foundation sued Anthropic over neural-network patents
The University of Tennessee Research Foundation alleges Anthropic used technology covered by two university-owned patents in Claude Code and other software without authorization. The Daily Beacon reports that the case seeks damages, but the complaint’s technical evidence and Anthropic’s response remain unavailable…utdailybeacon.comInnovation
Paper proposes CellFill for reversible updates to quantized language models without changing stored bits
An arXiv paper presents CellFill, a method that adds knowledge to deployed 4-bit language models through a separate update file while preserving the original quantized weights exactly. The authors report high fact-injection rates and reversible updates, but the work remains an unpeer-reviewed preprint requiring…arxiv.orgInnovation
Prospective study finds LLM accuracy falls sharply on ambiguous clinical registry questions
A multi-site prospective study reports that a large language model matched 87% of human-consensus answers when extracting information from unprocessed medical records, but accuracy fell to 62% on questions requiring event timing and greater clinical reasoning.arxiv.orgInnovation
A decision-focused framework audits how 2D AI foundation models transfer into 3D perception
A new arXiv paper introduces Lift, Associate, and Fuse, a framework for examining how 2D foundation-model predictions become persistent 3D segmentation information across 161 systems.arxiv.orgInnovation
MV2GF uses a visual foundation model to improve pedestrian detection across camera layouts
A paper accepted by ECCV 2026 proposes MV2GF, a multi-view pedestrian-detection system designed to generalize better when camera arrangements differ from those seen during training. Its authors use a visual geometric foundation model to estimate 3D structure and place image features in more appropriate world-space…arxiv.orgInnovation
Preprint proposes a more efficient way to improve AI understanding of long videos
A new arXiv preprint introduces Segment-to-Video Supervision, a training method designed to help multimodal AI systems identify relevant details in long videos while reducing training and inference overhead.arxiv.orgInnovation
LoRC paper reports a cross-generator method for detecting AI-generated images
Researchers propose LoRC, a detector based on low-rank collapse in semantic residuals, reporting an average accuracy improvement of 7.0% across multiple benchmarks and 97.0% accuracy on images from 39 unseen generators.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