PaddleOCR‑VL‑1.6, a compact 0.9B visual‑language model, diagnoses three types of weak regions, enriches targeted data, and applies a three‑stage CPT‑SFT‑RL training pipeline to reach a 96.33% overall score on OmniDocBench v1.6, surpassing much larger models across all document‑parsing tasks.
The weekly highlights introduce five cutting‑edge AI models—MiniCPM5-1B, HiDream-O1-Image, X2SAM, LocateAnything-3B and Granite 4.1 8B—detailing their architectures, novel training methods, performance claims such as SOTA on tool‑calling and code synthesis, and providing online demo links and a compute‑gift offering for developers.
DeepScientist V1.5, an open‑source AI system from Westlake University, claims to automate the full research pipeline—from literature review and code debugging to experiment analysis and paper writing—delivering SOTA results in weeks and offering one‑click deployment on Windows, Linux and macOS.
A joint team from Nvidia, Oxford University and the Quebec AI Institute presents Complexa, an atom‑level protein binder generation framework that unifies generative and refinement steps, achieves state‑of‑the‑art in‑silico success rates, and scales efficiently with test‑time compute.
Ant Group's open‑source native GUI agent UI‑Venus leverages multimodal large‑model and reinforcement‑learning techniques to outperform prior models on grounding and navigation benchmarks, while using a high‑quality data pipeline and a self‑evolving alignment mechanism to push the limits of GUI automation.
DeepKD introduces a double‑layer decoupling framework and a dynamic top‑K mask that adaptively denoises low‑confidence logits, addressing conflicts between target and non‑target knowledge flows; extensive experiments on CIFAR‑100, ImageNet‑1K, and MS‑COCO demonstrate consistent accuracy gains and state‑of‑the‑art performance.
PixelHacker introduces a latent class guidance (LCG) paradigm that injects foreground and background embeddings into a diffusion model, training on 14 million image‑mask pairs and achieving state‑of‑the‑art structural and semantic consistency across Places2, CelebA‑HQ and FFHQ benchmarks.
VideoPainter introduces a plug‑and‑play dual‑branch framework with a lightweight context encoder and ID‑resampling adapter, built on the massive VPData/VPBench dataset, and demonstrates state‑of‑the‑art performance across eight video restoration and editing metrics, while supporting flexible model integration and long‑video consistency.
The article analyzes D2LLM’s teacher‑student bi‑encoder architecture and Codefuse‑CGE’s PMA‑enhanced code embedding, showing how both models surpass BERT dual encoders and LLM cross‑encoders in accuracy, efficiency, and storage cost across semantic and code search benchmarks.
The article explains how the Anomaly Transformer leverages prior‑ and series‑association discrepancies, a learnable Gaussian kernel, and a Minimax training strategy to distinguish normal from abnormal points in time‑series data, achieving state‑of‑the‑art results on five benchmark datasets.
By constructing virtual mirror samples that occupy identical positions across source and target domains, the authors eliminate covariate shift while preserving distribution structure, enabling superior unsupervised domain adaptation that achieves state‑of‑the‑art performance on Office and VisDA benchmarks and improves real‑world lighting and gender‑recognition tasks.
