arXiv:2604.28031v2 Announce Type: cross Abstract: When researchers iteratively refine ideas with large language models, do the models preserve fidelity to the original objective? We introduce DriftBench, a benchmark for evaluating constraint adherence in multi-turn LLM-assisted scientific ideation. Across 2,146 scored benchmark runs spanning seven models from five providers (including two open-weight), four interaction conditions, and […]
UCB bets $2 billion on Candid’s T cell engager ambitions
Ken Song has done it again. The biopharma veteran’s all-out effort to prove T cell engagers’ potential in autoimmune diseases is getting picked up by one of Europe’s oldest pharma companies. UCB is paying $2 …
Effective monitoring of online AI decision-making algorithms in just-in-time adaptive interventions
npj Digital Medicine, Published online: 04 May 2026; doi:10.1038/s41746-026-02669-4 Effective monitoring of online AI decision-making algorithms in just-in-time adaptive interventions
Multimodal deep learning model for multiclass classification of renal tumors
npj Digital Medicine, Published online: 04 May 2026; doi:10.1038/s41746-026-02697-0 Multimodal deep learning model for multiclass classification of renal tumors
Subspecialty-specific foundation model for intelligent gastrointestinal pathology
npj Digital Medicine, Published online: 04 May 2026; doi:10.1038/s41746-026-02684-5 Subspecialty-specific foundation model for intelligent gastrointestinal pathology
Noninvasive molecular subtyping of breast cancer using multimodal ultrasound spatiotemporal transformer
npj Digital Medicine, Published online: 04 May 2026; doi:10.1038/s41746-026-02699-y Noninvasive molecular subtyping of breast cancer using multimodal ultrasound spatiotemporal transformer
The absence of full lifecycle risk management for AI-based medical devices in radiology
npj Digital Medicine, Published online: 04 May 2026; doi:10.1038/s41746-026-02712-4 The absence of full lifecycle risk management for AI-based medical devices in radiology