paper / stability / Jun 10
Baichuan-M4 is a medical agent system that moves beyond single-turn QA to continuous care. It integrates a unified runtime (Baichuan-Harness) that aligns RL training and deployment, a core reasoning model trained with span-level reward modeling (SPAR++), reasoning-path compression, and curriculum learning, plus a clinical tool layer for memory, evidence-based retrieval, and multimodal perception. The system achieves leading results on a comprehensive medical evaluation suite with a hallucination rate of just 3.3%.
medical-aiai-agentsreinforcement-learningclinical-caremultimodal-aillm-evaluation
“Baichuan-M4 achieves state-of-the-art results across static medical knowledge, dynamic OSCE-style consultation, long-context clinical memory, evidence-based retrieval, medical document OCR, and multimodal image understanding.”
paper / stability / Jun 10
The BESIII experiment has precisely measured the branching fraction of the electromagnetic Dalitz decay $D^{*+}_{s}\to e^{+}e^{-}D^{+}_{s}$ using a large dataset collected at center-of-mass energies from 4.128 to 4.226 GeV. The result, ${\mathcal{B}(D^{*+}_{s}\to e^{+}e^{-}D^{+}_{s})=(7.28\pm0.61_{\mathrm{stat}}\pm0.31_{\mathrm{syst}})\times10^{-3}}$, demonstrates a significant improvement in precision over previous measurements. This accurate determination is vital for refining theoretical models and establishing absolute branching fractions for other $D^{*+}_{s}$ decay modes.
hep-exbesiii-experimentelectromagnetic-dalitz-decaybranching-fractiond-meson-physicscharm-physicsparticle-physics
“The branching fraction of the electromagnetic Dalitz decay $D^{*+}_{s}\to e^{+}e^{-}D^{+}_{s}$ has been measured by the BESIII experiment.”
paper / stability / Jun 10
Conversational AI recommendations to unengaged users demonstrably increase search activity, brand website visits, and retailer page engagement. Methodologies isolating true recommendations from incidental mentions reveal that while these AI prompts significantly influence user behavior, the downstream path is largely search-driven and aligns with pre-existing consumer navigation patterns. Referral tracking misses this upstream influence of AI on brand discovery.
ai-recommendationsconsumer-behaviorconversational-aionline-advertisingmarketing-analytics
“Conversational AI recommendations to users with no recent engagement increase same-name Google searches by 4.3 percentage points.”