Professional AI media and industry service platform
The article analyzes JD Health’s AI doctor “Dawei”, detailing how its end‑to‑end virtual consultation, data‑driven decision making and seamless fulfillment drove a near‑four‑fold user increase, high satisfaction, and a significant uplift in revenue and profit reported in the 2026 mid‑year results.
The World Critic Model (WCM) augments the critic in vision‑language‑action reinforcement learning with future state prediction, enabling robots to evaluate not only the current value but also anticipate upcoming dynamics, which dramatically improves both in‑distribution and out‑of‑distribution performance across multiple benchmarks.
DeepSeek has officially released its V4 Pro large language model, designated DeepSeek‑V4‑Pro‑0813, with unchanged API pricing, a noticeable shift in chain‑of‑thought behavior, and benchmark results that put its agent capabilities on par with top models like Claude Fable 5, while recent V4 Flash users report performance drops and potential price hikes.
A new security paper demonstrates that design flaws in Claude, GPT‑5.6 and other leading LLM APIs allow attackers to steal encrypted reasoning blocks, replay them in weaker compatible models, and reconstruct most of the hidden thought process, exposing privacy and safety risks.
Over the past two years Meitu’s MT Lab has published eleven top‑conference papers on video editing, portrait manipulation, visual understanding and more, and has transformed these research breakthroughs—such as MiVE, CFT, BridgeRemoval, WearWow, FlowSeg and All‑in‑One Slider—into product features that let AI grasp creators’ intent and deliver precise, controllable image and video edits.
The paper presents Libra, a resource‑management system for Agentic RL post‑training that jointly optimizes training and rollout GPU allocation using a global planner, heterogeneous inference clusters, a causality‑driven multi‑level feedback queue, and an elastic hybrid pool, achieving up to three‑fold throughput gains and up to 2.5× faster reward convergence.
The ByteDance Seed team shows that merely extending caption length adds little visual supervision for diffusion models; instead, the amount of image‑grounded information in captions predicts training loss, leading them to propose Structured Prompt, new metrics (GPG, ED), and a three‑stage LLM prompter to improve both Diffusability and Promptability.
The MLS‑Bench benchmark introduces 140 executable research tasks across twelve ML domains to rigorously attribute performance gains to genuine method discovery rather than engineering tricks, revealing that current large‑model agents excel at component recombination but still lag in proposing truly novel, transferable algorithms.
The paper systematically derives scaling laws for mixture-of-experts diffusion language models, trains a 30B-parameter LLaDA MoE v2 model guided by these laws, and demonstrates that it matches or exceeds state-of-the-art autoregressive models while using far less compute.
Meta has open‑sourced its 29.6‑billion‑parameter Muse Glimmer model under Apache 2.0, offering a 4‑bit quantized version that fits on a single high‑end GPU, while benchmark results show strong agent performance but notable hallucination and accuracy gaps compared with competing models.
