Share story
Science AI

CheXPoT study improves AI radiology reports by removing unsupported findings

Researchers report in a preprint that CheXPoT improved the precision of AI-generated chest radiograph reports by checking findings against image evidence and removing unsupported statements. On a 4,145-image held-out split, adding it to MedGemma-4B raised micro-average precision from 19.2% to 46.8% and F1 from 23.0% to 49.4%. The framework breaks reports into individual assertions, checks them using image features and verification code, then rewrites the reports. Precision improved significantly on 8 of 14 CheXpert classes and F1 on 5, with no significant decreases in either metric.
Sources
cs.CL updates on arXiv.org
Published by Tech & Business, a media brand covering technology and business. This story was sourced from cs.CL updates on arXiv.org and reviewed by the T&B editorial agent team.
Back to Newswire
Keep reading
Full wire
Science
Science

MIT-led team receives up to $38 million for disease-monitoring patch

An MIT-led team has received funding from the U.S. Advanced Research Projects Agency for Health for a four-year project worth up to $38 million to develop a wearable patch that continuously monitors biomarkers of heart disease and sepsis.

MIT News
AI Science
AI Science

MEMENTO study improves book question answering with a 20K-token context

Researchers report in a preprint that MEMENTO raised Qwen3-8B's accuracy from 36.9% to 67.8% across four book question-answering benchmarks while holding no more than 20K tokens in context. The tests included books of up to 2.5M tokens.

cs.CL updates on arXiv.org
AI Robotics
AI Robotics

DreamX preprint reports 300× compression of video action representations

Researchers report in a DreamX preprint that they compressed each latent video frame from 1,200 tokens to 4 while retaining action information that transfers across human and robot embodiments and environments.

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30
Robotics Science
Robotics Science

RoboTryOn study reports robot task success rising from 2% to 60%

Researchers report in an arXiv preprint that robot policies trained with RoboTryOn synthetic demonstrations improved average success on novel tasks from 2% to 60%, without real robot demonstrations of those tasks.

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30
AI Capital
AI Capital

Fal raises $1.3 billion at $16 billion valuation

Fal raised $1.3 billion in a funding round led by Bessemer Venture Partners that values the AI inference startup at $16 billion, TechCrunch reported, citing Bloomberg. Nvidia, Sound Ventures and T. Rowe Price participated. The valuation doubles the $8 billion assigned in April.

Startups | TechCrunch