Brief

Researchers have developed a suite of Foundation Inference Models (FIMs) designed to rapidly estimate parameters for various differential equations from time-series data. These models, including FIM-SDE for stochastic differential equations, FIM-PP for temporal point processes, and FIM-ODE for ordinary differential equations, are pretrained on broad distributions of synthetic data. This pretraining allows them to perform in-context (zero-shot) inference or be quickly fine-tuned to specific datasets, often outperforming traditional methods and specialized models that require extensive training. AI

Two new research papers introduce advancements in hypergraph neural networks (HNNs). One paper proposes HADES, a method for knowledge distillation that adapts to node heterophily, improving student model performance and inference speed. The other paper introduces Hypergraph U-Nets, a novel architecture that addresses the challenge of pooling and unpooling operations in HNNs, demonstrating superior performance in reconstruction, classification, and anomaly detection tasks. AI

Researchers have developed new frameworks for camouflaged object detection (COD) that address the issue of over-detection. One approach, CFCamo, uses a counterfactual benchmark to train agents to both detect camouflaged objects and abstain when no object is present, improving performance on existing datasets and achieving high pair accuracy on the new CF-COD benchmark. Another method, CamoSAM2, refines prompts for the Segment Anything Model 2 (SAM2) by integrating motion and appearance cues to enhance automatic detection and segmentation of camouflaged objects in videos, outperforming current state-of-the-art methods in mean intersection over union (mIoU) and inference speed. AI

Researchers have introduced TABVERSE, a new benchmark designed to evaluate how well Large Language Models (LLMs) and Vision-Language Models (VLMs) understand tables across different formats. The benchmark standardizes table content while varying its representation, such as HTML, Markdown, LaTeX, and rendered images. Initial findings indicate that model performance is significantly influenced by the table's format, with structured text generally outperforming images, though specific tasks and formats present unique challenges. AI

Researchers have developed a novel method for molecular design using large language models (LLMs) that moves beyond simple trial-and-error. By feeding detailed physicochemical rationales, such as orbital energies and atomic charges, back into the LLM instead of just numerical scores, the system acts as a causal reasoner. This self-reflective approach achieved a 100% success rate on moderate tasks for targeting HOMO-LUMO gaps and proved effective for dipole-moment design across multiple LLM backbones. AI

Researchers have developed a SpeechLLM designed for assessing L2 speech proficiency across multiple granularities and providing natural language rationales. This model, trained using a hybrid approach of supervised fine-tuning and Bounded Direct Preference Optimization, can predict sentence-level labels for accuracy, fluency, and prosody, as well as word/phoneme-level accuracy. While the model demonstrates strong performance and plausible sentence-level rationales, its faithfulness degrades at the word/phoneme level due to sparse and weakly aligned references. AI

Researchers have developed DecSelfMask, a novel method to improve classification performance in decoder-only language models using unlabeled data. This approach employs a relevance-guided masking strategy, identifying crucial text segments and training the model to reconstruct them. DecSelfMask demonstrated significant gains, outperforming standard supervised fine-tuning by nearly 20 points in Macro F1 on a dataset of 1.9 million clinical notes. AI

Researchers have developed the Graph Mamba Operator (GraMO), a novel approach for simulating interacting particle systems. GraMO integrates state-space models with graph-based learning to simultaneously handle spatial interactions and long-range temporal dependencies. This method aims to overcome limitations of existing models that often separate these dynamics, leading to error accumulation over extended prediction horizons. AI

Researchers have developed vesselFM-CT, a novel model designed to segment all blood vessels within CT images. This advancement aims to overcome the limitations of previous studies that focused on isolated vascular segments, enabling a more comprehensive analysis of the entire cardiovascular system. The model utilizes an iterative training process and a new TubeLoss function to handle the diverse structural variations of blood vessels, from large arteries to minuscule mesenteric vessels. AI

Chinese AI startup Stepfun has released its Step 3.7 Flash model, which reportedly achieves 97% of Claude's coding capabilities at one-ninth the cost. This new model offers a speed of 416 tokens per second, making it a compelling option for high-volume API users and real-time applications prioritizing cost-effectiveness and speed over absolute peak performance. AI

Researchers have developed new frameworks to automate the creation and management of software repositories, addressing a key bottleneck in automated software engineering. One system, RepoLaunch, successfully builds and tests code across various languages and platforms with a 78% success rate. Another effort introduces DeNovoSWE, a large dataset of 4,818 instances for training code agents to generate entire repositories from documentation, significantly improving performance on complex tasks. AI

Researchers have developed CapRL++, a novel framework for training image and video captioning models using reinforcement learning with verifiable rewards. This approach moves beyond traditional supervised fine-tuning by using a vision-free language model to assess caption quality based on its ability to answer questions about the visual content. Evaluations across numerous benchmarks demonstrate that CapRL++ enhances caption quality and pretraining, leading to significant downstream performance gains and enabling smaller models to match the capabilities of much larger ones. AI

Researchers have developed Echo-DM, a novel framework for removing artificial markers from clinical ultrasound images. This method utilizes a conditional latent diffusion model combined with region-aware fusion to restore images without relying on masks, preserving anatomical details. Experiments on the Echo-PAIR dataset show Echo-DM outperforms existing methods in marker removal and anatomical fidelity, offering efficient deployment options. AI

Researchers have introduced ExDet, a novel framework designed to improve open-domain open-vocabulary detection (ODOVD) capabilities. This lightweight system enhances the generalization of existing detectors to new categories and unseen domains without requiring training from scratch. ExDet utilizes text-guided extrapolation to infer visual prototypes and a detector-compatible rectification module to adjust representations, achieving state-of-the-art results on several benchmark datasets. AI

Researchers have investigated the temporal stability of machine learning models used to emulate satellite-based greenhouse gas retrievals. Their study, using data from the Greenhouse Gases Observing SATellite (GOSAT), found that prediction accuracy degrades over time when models are tested on data outside their training period. Incorporating time as a feature significantly improved methane predictions, with a simple Lasso model outperforming more complex neural networks and demonstrating greater stability. AI

Anthropic's AI model, Fable, has been programmed with overly conservative safeguards that prevent it from answering basic biology questions. This decision was made to mitigate the risk of the AI being used to develop bioweapons. As a result, Fable refuses to engage with most queries related to biological work. AI

Researchers have developed a novel two-stage framework called Rea2Seg for image segmentation tasks that leverage multimodal large language models (MLLMs). This approach first identifies candidate masks from an MLLM's attention maps and then uses the MLLM to reason over these candidates and select the most accurate one. To further evaluate and advance these capabilities, a new benchmark, ReasonSeg-SGDR, has been introduced to assess perception, grounding, and reasoning abilities across various dimensions. AI

Researchers have introduced PRISM, a novel framework for federated graph learning that addresses the challenge of modality deficiency across different clients. PRISM enables collaborative learning from decentralized graphs containing text and images, even when individual clients lack complete multimodal data. The framework proactively retrieves and imputes missing modality semantics from the federation, integrating them into local graph propagation with topology-aware control. Experiments demonstrate PRISM's effectiveness, showing an average improvement of 4.48% over state-of-the-art baselines on six multimodal graph datasets. AI

Researchers have developed a new method called Saturating Additive Rewards (SAR) to improve the precision of large language models in geometric tasks. This approach addresses a failure mode known as Outlier Gradient Masking, where a single constraint violation can hinder learning across all constraints. SAR decomposes rewards into bounded per-constraint terms, preserving partial progress and ensuring consistent gradients. An 8B parameter model using SAR achieved a 2.3x improvement in solving complex geometric problems compared to standard MSE-based rewards. AI

Researchers from XInsight Lab have developed a novel ensemble framework for micro-gesture recognition, achieving a new state-of-the-art result in the 4th MiGA Challenge at IJCAI 2026. Their approach integrates a self-supervised RGB model, pre-trained on a large unlabeled video dataset, with existing supervised models. This self-supervised component significantly improved performance, reaching 74.419% top-1 accuracy and outperforming previous benchmarks by over 1.2 percentage points. AI

Researchers have developed LiteVSR, a new framework for adapting pre-trained diffusion transformers for video super-resolution tasks. This approach uses a lightweight State-Aware Adapter that requires significantly fewer trainable parameters and less training time compared to existing methods. LiteVSR leverages flow matching to efficiently adapt the frozen transformer, enabling competitive restoration quality with minimal computational resources. AI

Researchers have developed a new framework called CREDiT to improve the reliability of video question-answering systems. This framework uses counterfactual reasoning and structural causal models to disentangle causal evidence from spurious correlations in video data. By decomposing representations into causal and non-causal components and employing feature-level causal interventions, CREDiT aims to create more trustworthy AI systems that can accurately localize evidence. AI

Researchers have introduced OmniGen-AR, a novel autoregressive framework designed for versatile image generation. This unified model can synthesize images from various inputs, including text, segmentation maps, depth information, and even existing images for editing or video prediction. To prevent condition tokens from influencing content tokens, the framework employs Disentangled Causal Attention (DCA), a technique that separates attention mechanisms during training. OmniGen-AR has demonstrated state-of-the-art performance on benchmarks like GenEval and VBench. AI

Researchers have introduced Ultra Flash, a novel cascaded streaming framework designed to generate high-resolution video in real-time. This system overcomes the limitations of previous models that were restricted to lower resolutions. Ultra Flash achieves impressive frame rates at 1K and 2K resolutions on a single GPU by employing a unique super-resolution training paradigm and a causal streaming latent upsampler. AI

Researchers have developed several new methods to combat hallucinations in video large multimodal models (VLMMs). One approach, MultiToP, refines unreliable visual tokens before language generation by selectively substituting them with a global patch token. Another method, ViSSRes, enhances video representations using a lightweight network to improve spatiotemporal and semantic consistency. A third technique focuses on refining textual embeddings to encourage better integration of visual information and reduce over-reliance on language priors. These methods have shown significant improvements in reducing hallucination rates and enhancing video understanding capabilities across various benchmarks. AI

Researchers have developed SwiftVR, a novel framework for real-time generative video restoration that addresses key bottlenecks in existing diffusion-based models. By employing mask-free shifted-window self-attention and a lightweight autoencoder, SwiftVR achieves high frame rates at resolutions up to 4K on powerful hardware and real-time 1080p streaming on consumer-grade GPUs. This advancement makes high-quality video restoration more accessible and practical for live streaming applications. AI