Kandinsky 6.0 Improves Its Predecessor’s Synchronized Video
A bidirectional dual-stream diffusion transformer couples a pretrained video model with a new audio stream to generate five-second clips with 44 kHz sound.
Oct 8, 20265 min2610.05608
Computer Vision
cs.CVImage processing, computer vision, pattern recognition, and scene understanding.
Sort:
HERO Puts Acquisition Robustness at the Center of Pathology Pretraining
A ViT-G/14 pathology encoder combines DINO and iBOT pretraining with high-resolution Gram anchoring, targeting reliable representations across center, scanner, and stain variation.
Oct 2, 20263 min2609.35943
PhysBrain 1.5 Unifies Perception, Action, and Prediction
An 8B autoregressive model turns language, end-effector motion, and visual targets into one token stream, reaching a 72.5 average across 28 embodied benchmarks.
Sep 27, 20264 min2609.14973
MinerU-Chem Reports 93.02% Exact Molecular Recognition
A chemistry-focused extension to MinerU converts document regions into molecule and reaction records, outperforming the evaluated GPT-5.6-Sol comparison on a SMILES benchmark subset.
Sep 14, 20262 min2608.03525
Verifiable Rewards Improve Native Visual Reasoning Training
A 300-task procedural suite replaces preference-only judging with task-specific scorers, raising matched reinforcement-learning performance from 0.509 to 0.548.
Sep 4, 20264 min2608.26105
Million-Clip Dataset Tests Video Reasoning Beyond Visual Quality
VBVR pairs 200 curated reasoning tasks with rule-based, human-aligned evaluation to study whether video models generalize across spatiotemporal reasoning problems.
Aug 30, 20264 min2602.20159
SkillNet Cuts Agent Steps Through Reusable Skills
An ontology, evaluation scheme, and 600,000-skill repository let agents retrieve and compose prior procedures, reporting 40% higher rewards with 30% fewer steps.
Aug 23, 20264 min2603.04448
FilmBench Reveals Cinematic Gaps in Video Generators
A director-designed taxonomy and evaluation agent track human model rankings at Spearman ρ=0.95–0.96 while exposing multi-shot and dynamic-aesthetics failures.
Aug 3, 20264 min2607.24241
Open Image Model Narrows Gap Under Tight Compute
Boogu-Image-0.1 combines a stronger multimodal encoder, agentic prompt rewriting, and curated data to train competitive generation and editing models for about $400K.
Jul 29, 20265 min2607.13125 Code available
Open Egocentric Data Lowers Barriers for Robot Learning
Smartphone capture and a modular processing toolchain turn 2,000 hours of human manipulation video into structured supervision for embodied models.
Jul 29, 20265 min2607.14183