A Multi-Agent Framework with Automated Decision Rule Optimization for Cross-Domain Misinformation Detection

When OCR Learns to Rebuild: Unisound U1-OCR Converts Images and PDFs into Editable PowerPoint Files in One Click, Ushering in a New Era of Intelligent Document Reconstruction

Recently, Unisound upgraded its U1-OCR foundation model for document intelligence and officially launched the IMG2PPT intelligent conversion feature. Built on the U1-OCR document intelligence engine, IMG2PPT reconstructs content from static JPG, PNG, and PDF files as natively editable PowerPoint presentations. Text boxes, images, lists, shapes, tables, and other elements become native, editable PowerPoint objects. This marks OCR's progression from single-purpose content recognition to a new stage of intelligent understanding and structural reconstruction.

Unisound Officially Launches U2-RadiMed, Ushering Medical Imaging AI into the Era of

As the first domestic imaging foundation model to deeply integrate medical image understanding, report generation, clinical reasoning and decision‑making, and visual question answering in a full‑chain solution, U2‑RadiMed supports multimodal mixed input of single or multiple images and text. It completes a closed‑loop workflow of “view images → retrieve knowledge → reason → conclude,” providing clinicians with comprehensive AI assistance—from lesion screening and precise diagnosis to personalized treatment decisions—thus propelling imaging diagnosis to a higher level of intelligence.

Unisound U2-ASR and U2-TTS Get Comprehensive Multilingual Upgrades, Enabling Global Listening and Speaking with a Single Model

Unisound has recently completed a full‑scale upgrade to the multilingual capabilities of U2‑ASR and U2‑TTS. U2‑ASR adds recognition for 13 new international languages, while U2‑TTS introduces speech synthesis for 8 Southeast Asian languages. With this update, the U2 Speech Large Model now supports over 100 Chinese dialects and more than 15 international languages.

Unisound Launches U2‑Med Expert‑Level Tri‑Medical Large Model: Reshaping the Industry with

Guided by the original aspiration of “AI for Good”, U2‑Med strives to upgrade AI from a conversational tool into a “digital expert” embedded in front‑line business workflows. By breaking down data and business silos across medical care, medical insurance and pharmaceuticals, and reconstructing the industrial ecosystem, U2‑Med is not only scaling new technical heights but also pursuing better human health outcomes. It leverages AI to drive medical services toward greater efficiency, equity and intelligence, so that the warmth of technology can truly benefit public health and well‑being worldwide.

VAPO: End-to-end Slide-Enhanced Speech Recognition with Omni-modal Large Language Models

HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment

Reinforcement Learning with Verifiable Reward (RLVR) has proven effective for training reasoning-oriented large language models, but existing methods largely assume high-resource settings with abundant training data. In low-resource scenarios, RLVR is prone to more severe entropy collapse, which substantially limits exploration and degrades reasoning performance. To address this issue, we propose Hybrid-domain Entropy dynamics ALignment (HEAL), a framework tailored for few-shot RLVR. HEAL first selectively incorporates high-value general-domain data to promote more diverse exploration. Then, we introduce Entropy Dynamics Alignment (EDA), a reward mechanism that aligns trajectory-level entropy dynamics between the target and general domains, capturing both entropy magnitude and fine-grained variation. Through this alignment, EDA not only further mitigates entropy collapse but also encourages the policy to acquire more diverse exploration behaviors from the general domain. Experiments across multiple domains show that HEAL consistently improves few-shot RLVR performance. Notably, using only 32 target-domain samples, HEAL matches or even surpasses full-shot RLVR trained with 1K target-domain samples.

GuideTree: Guideline-Induced Review Trees for Long Medical Records

Reviewing medical records for clinical and insurance decisions must handle long, heterogeneous documents while producing consistent, traceable, guideline-compliant outcomes under strict latency and cost constraints. We propose GuideTree, which compiles textual guidelines into a fixed review tree of evidence-grounded verification primitives. GuideTree uses short per-document summaries only for routing each check to a minimal set of document types and candidates; final verification always reads full document text and returns structured evidence. The tree is induced offline via a cost-aware split-and-prune search and updated safely through regression-tested, versioned patches. Across 1,000 cases from four industrial review scenarios and four LLM backbones, GuideTree achieves 84.5-92.8 Macro-F1, outperforming the strongest non-expert baselines by 3.3-7.6 points and matching ExpertTree within 0.2-0.6 points (avg. 0.38). On chronic disease with Qwen3-235B-A22B-Instruct, GuideTree reduces average I/O volume to 74K input+output characters (-82% vs. long-context prompting) and average latency to 22s (-83% vs. long-context prompting), while reaching 99% decision consistency over K=5 reruns.

FocalOrder: Focal Preference Optimization for Reading Order Detection

Reading order detection is the foundation of document understanding. Most existing methods rely on uniform supervision, implicitly assuming a constant difficulty distribution across layout regions. In this work, we challenge this assumption by revealing a critical flaw: Positional Disparity, a phenomenon where models demonstrate mastery over the deterministic start and end regions but suffer a performance collapse in the complex intermediate sections. This degradation arises because standard training allows the massive volume of easy patterns to drown out the learning signals from difficult layouts. To address this, we propose FocalOrder, a framework driven by Focal Preference Optimization (FPO). Specifically, FocalOrder employs adaptive difficulty discovery with an exponential moving average mechanism to dynamically pinpoint hard-to-learn transitions, while introducing a difficulty-calibrated pairwise ranking objective to enforce global logical consistency. Extensive experiments demonstrate that FocalOrder establishes new state-of-the-art results on OmniDocBench v1.0 and Comp-HRDoc. Our compact model not only outperforms competitive specialized baselines but also significantly surpasses large-scale general VLMs. These results demonstrate that aligning the optimization with the intrinsic structural ambiguity of documents is critical for mastering complex document structures.