Generative retrieval has emerged as a novel paradigm that leverages large language models (LLMs) to autoregressively generate document identifiers. Although promising, the mechanisms that underpin its performance and scalability remain largely unclear. We conduct a systematic investigation of training and inference scaling laws in generative retrieval, exploring how model size, training data scale, and inference-time compute jointly influence retrieval performance. To address the lack of suitable metrics, we propose a novel evaluation measure inspired by contrastive entropy and generation loss, providing a continuous performance signal that enables robust comparisons across diverse generative retrieval methods. Our experiments show that n-gram-based methods demonstrate strong alignment with both training and inference scaling laws, especially when paired with larger LLMs. Furthermore, increasing inference computation yields substantial performance gains, revealing that generative retrieval can significantly benefit from higher compute budgets at inference. Across these settings, LLaMA models consistently outperform T5 models, suggesting a particular advantage for larger decoder-only models in generative retrieval. Taken together, our findings underscore that model sizes, data availability, and inference computation interact to unlock the full potential of generative retrieval, offering new insights for designing and optimizing future systems.
In the Generative AI era, safeguarding 3D models has become increasingly urgent. While invisible watermarking is well-established for 2D images with encoder-decoder frameworks, generalizable and robust solutions for 3D remain elusive. The main difficulty arises from the renderer between the 3D encoder and 2D decoder, which disrupts direct gradient flow and complicates training. Existing 3D methods typically rely on per-scene iterative optimization, resulting in time inefficiency and limited generalization. In this work, we propose a single-pass watermarking approach for 3D Gaussian Splatting (3DGS), a well-known yet underexplored representation for watermarking. We identify two major challenges: (1) ensuring effective training generalized across diverse 3D models, and (2) reliably extracting watermarks from free-view renderings, even under distortions. Our framework, named GS-Marker, incorporates a 3D encoder to embed messages, distortion layers to enhance resilience against various distortions, and a 2D decoder to extract watermarks from renderings. A key innovation is the Adaptive Marker Control mechanism that adaptively perturbs the initially optimized 3DGS, escaping local minima and improving both training stability and convergence. Extensive experiments show that GS-Marker outperforms per-scene training approaches in terms of decoding accuracy and model fidelity, while also significantly reducing computation time.
State Space Models (SSMs) have recently demonstrated outstanding performance in long-sequence modeling, particularly in natural language processing. However, their direct application to medical image segmentation poses several challenges. SSMs, originally designed for 1D sequences, struggle with 3D spatial structures in medical images due to discontinuities introduced by flattening. Additionally, SSMs have difficulty fitting high-variance data, which is common in medical imaging. In this paper, we analyze the intrinsic limitations of SSMs in medical image segmentation and propose a unified U-shaped encoder-decoder architecture, Meta Mamba UNet (MM-UNet), designed to leverage the advantages of SSMs while mitigating their drawbacks. MM-UNet incorporates hybrid modules that integrate SSMs within residual connections, reducing variance and improving performance. Furthermore, we introduce a novel bi-directional scan order strategy to alleviate discontinuities when processing medical images. Extensive experiments on the AMOS2022 and Synapse datasets demonstrate the superiority of MM-UNet over state-of-the-art methods. MM-UNet achieves a Dice score of 91.0% on AMOS2022, surpassing nnUNet by 3.2%, and a Dice score of 87.1% on Synapse. These results confirm the effectiveness of integrating SSMs in medical image segmentation through architectural design optimizations.
In the evolving landscape of 6G networks, semantic communications are poised to revolutionize data transmission by prioritizing the transmission of semantic meaning over raw data accuracy. This paper presents a Vision Transformer (ViT)-based semantic communication framework that has been deliberately designed to achieve high semantic similarity during image transmission while simultaneously minimizing the demand for bandwidth. By equipping ViT as the encoder-decoder framework, the proposed architecture can proficiently encode images into a high semantic content at the transmitter and precisely reconstruct the images, considering real-world fading and noise consideration at the receiver. Building on the attention mechanisms inherent to ViTs, our model outperforms Convolution Neural Network (CNNs) and Generative Adversarial Networks (GANs) tailored for generating such images. The architecture based on the proposed ViT network achieves the Peak Signal-to-noise Ratio (PSNR) of 38 dB, which is higher than other Deep Learning (DL) approaches in maintaining semantic similarity across different communication environments. These findings establish our ViT-based approach as a significant breakthrough in semantic communications.
Consistency regularization has prevailed in semi-supervised semantic segmentation and achieved promising performance. However, existing methods typically concentrate on enhancing the Image-augmentation based Prediction consistency and optimizing the segmentation network as a whole, resulting in insufficient utilization of potential supervisory information. In this paper, we propose a Multi-Constraint Consistency Learning (MCCL) approach to facilitate the staged enhancement of the encoder and decoder. Specifically, we first design a feature knowledge alignment (FKA) strategy to promote the feature consistency learning of the encoder from image-augmentation. Our FKA encourages the encoder to derive consistent features for strongly and weakly augmented views from the perspectives of point-to-point alignment and prototype-based intra-class compactness. Moreover, we propose a self-adaptive intervention (SAI) module to increase the discrepancy of aligned intermediate feature representations, promoting Feature-perturbation based Prediction consistency learning. Self-adaptive feature masking and noise injection are designed in an instance-specific manner to perturb the features for robust learning of the decoder. Experimental results on Pascal VOC2012 and Cityscapes datasets demonstrate that our proposed MCCL achieves new state-of-the-art performance. The source code and models are made available at https://github.com/NUST-Machine-Intelligence-Laboratory/MCCL.
Recent efforts in neural compression have focused on the rate-distortion-perception (RDP) tradeoff, where the perception constraint ensures the source and reconstruction distributions are close in terms of a statistical divergence. Theoretical work on RDP describes properties of RDP-optimal compressors without providing constructive and low complexity solutions. While classical rate distortion theory shows that optimal compressors should efficiently pack space, RDP theory additionally shows that infinite randomness shared between the encoder and decoder may be necessary for RDP optimality. In this paper, we propose neural compressors that are low complexity and benefit from high packing efficiency through lattice coding and shared randomness through shared dithering over the lattice cells. For two important settings, namely infinite shared and zero shared randomness, we analyze the RDP tradeoff achieved by our proposed neural compressors and show optimality in both cases. Experimentally, we investigate the roles that these two components of our design, lattice coding and randomness, play in the performance of neural compressors on synthetic and real-world data. We observe that performance improves with more shared randomness and better lattice packing.
We implement a ML-based attention framework with component-specific decoders, improving optical power spectrum prediction in multi-span networks. By reducing the need for in-depth training on each component, the framework can be scaled to multi-span topologies with minimal data collection, making it suitable for brown-field scenarios.
Egocentric video-language pretraining has significantly advanced video representation learning. Humans perceive and interact with a fully 3D world, developing spatial awareness that extends beyond text-based understanding. However, most previous works learn from 1D text or 2D visual cues, such as bounding boxes, which inherently lack 3D understanding. To bridge this gap, we introduce EgoDTM, an Egocentric Depth- and Text-aware Model, jointly trained through large-scale 3D-aware video pretraining and video-text contrastive learning. EgoDTM incorporates a lightweight 3D-aware decoder to efficiently learn 3D-awareness from pseudo depth maps generated by depth estimation models. To further facilitate 3D-aware video pretraining, we enrich the original brief captions with hand-object visual cues by organically combining several foundation models. Extensive experiments demonstrate EgoDTM's superior performance across diverse downstream tasks, highlighting its superior 3D-aware visual understanding. Our code will be released at https://github.com/xuboshen/EgoDTM.
Video object segmentation is crucial for the efficient analysis of complex medical video data, yet it faces significant challenges in data availability and annotation. We introduce the task of one-shot medical video object segmentation, which requires separating foreground and background pixels throughout a video given only the mask annotation of the first frame. To address this problem, we propose a temporal contrastive memory network comprising image and mask encoders to learn feature representations, a temporal contrastive memory bank that aligns embeddings from adjacent frames while pushing apart distant ones to explicitly model inter-frame relationships and stores these features, and a decoder that fuses encoded image features and memory readouts for segmentation. We also collect a diverse, multi-source medical video dataset spanning various modalities and anatomies to benchmark this task. Extensive experiments demonstrate state-of-the-art performance in segmenting both seen and unseen structures from a single exemplar, showing ability to generalize from scarce labels. This highlights the potential to alleviate annotation burdens for medical video analysis. Code is available at https://github.com/MedAITech/TCMN.
Global effective receptive field plays a crucial role for image style transfer (ST) to obtain high-quality stylized results. However, existing ST backbones (e.g., CNNs and Transformers) suffer huge computational complexity to achieve global receptive fields. Recently, the State Space Model (SSM), especially the improved variant Mamba, has shown great potential for long-range dependency modeling with linear complexity, which offers a approach to resolve the above dilemma. In this paper, we develop a Mamba-based style transfer framework, termed SaMam. Specifically, a mamba encoder is designed to efficiently extract content and style information. In addition, a style-aware mamba decoder is developed to flexibly adapt to various styles. Moreover, to address the problems of local pixel forgetting, channel redundancy and spatial discontinuity of existing SSMs, we introduce both local enhancement and zigzag scan. Qualitative and quantitative results demonstrate that our SaMam outperforms state-of-the-art methods in terms of both accuracy and efficiency.
Scaling architectures have been proven effective for improving Scene Text Recognition (STR), but the individual contribution of vision encoder and text decoder scaling remain under-explored. In this work, we present an in-depth empirical analysis and demonstrate that, contrary to previous observations, scaling the decoder yields significant performance gains, always exceeding those achieved by encoder scaling alone. We also identify label noise as a key challenge in STR, particularly in real-world data, which can limit the effectiveness of STR models. To address this, we propose Cloze Self-Distillation (CSD), a method that mitigates label noise by distilling a student model from context-aware soft predictions and pseudolabels generated by a teacher model. Additionally, we enhance the decoder architecture by introducing differential cross-attention for STR. Our methodology achieves state-of-the-art performance on 10 out of 11 benchmarks using only real data, while significantly reducing the parameter size and computational costs.
This study investigates the problem of learning linear block codes optimized for Belief-Propagation decoders significantly improving performance compared to the state-of-the-art. Our previous research is extended with an enhanced system design that facilitates a more effective learning process for the parity check matrix. We simplify the input dataset, restrict the number of parameters to learn and improve the gradient back-propagation within the model. We also introduce novel optimizers specifically designed for discrete-valued weights. Based on conventional gradient computation, these optimizers provide discrete weights updates, enabling finer control and improving explainability of the learning process. Through these changes, we consistently achieve improved code performance, provided appropriately chosen hyper-parameters. To rigorously evaluate the performance of learned codes in the context of short to medium block lengths, we propose a comprehensive code performance assessment framework. This framework enables a fair comparison between our learning methodology and random search approaches, ensuring statistical significance in our results. The proposed model pave the way for a new approach to the efficient learning of linear block codes tailored to specific decoder structures.
Background: This study proposes a Vision-Language Model (VLM) leveraging the SIGLIP encoder and Gemma-3b transformer decoder to enhance automated chronic tuberculosis (TB) screening. By integrating chest X-ray images with clinical data, the model addresses the challenges of manual interpretation, improving diagnostic consistency and accessibility, particularly in resource-constrained settings. Methods: The VLM architecture combines a Vision Transformer (ViT) for visual encoding and a transformer-based text encoder to process clinical context, such as patient histories and treatment records. Cross-modal attention mechanisms align radiographic features with textual information, while the Gemma-3b decoder generates comprehensive diagnostic reports. The model was pre-trained on 5 million paired medical images and texts and fine-tuned using 100,000 chronic TB-specific chest X-rays. Results: The model demonstrated high precision (94 percent) and recall (94 percent) for detecting key chronic TB pathologies, including fibrosis, calcified granulomas, and bronchiectasis. Area Under the Curve (AUC) scores exceeded 0.93, and Intersection over Union (IoU) values were above 0.91, validating its effectiveness in detecting and localizing TB-related abnormalities. Conclusion: The VLM offers a robust and scalable solution for automated chronic TB diagnosis, integrating radiographic and clinical data to deliver actionable and context-aware insights. Future work will address subtle pathologies and dataset biases to enhance the model's generalizability, ensuring equitable performance across diverse populations and healthcare settings.
3D shape generation has greatly flourished through the development of so-called "native" 3D diffusion, particularly through the Vecset Diffusion Model (VDM). While recent advancements have shown promising results in generating high-resolution 3D shapes, VDM still struggles with high-speed generation. Challenges exist because of difficulties not only in accelerating diffusion sampling but also VAE decoding in VDM, areas under-explored in previous works. To address these challenges, we present FlashVDM, a systematic framework for accelerating both VAE and DiT in VDM. For DiT, FlashVDM enables flexible diffusion sampling with as few as 5 inference steps and comparable quality, which is made possible by stabilizing consistency distillation with our newly introduced Progressive Flow Distillation. For VAE, we introduce a lightning vecset decoder equipped with Adaptive KV Selection, Hierarchical Volume Decoding, and Efficient Network Design. By exploiting the locality of the vecset and the sparsity of shape surface in the volume, our decoder drastically lowers FLOPs, minimizing the overall decoding overhead. We apply FlashVDM to Hunyuan3D-2 to obtain Hunyuan3D-2 Turbo. Through systematic evaluation, we show that our model significantly outperforms existing fast 3D generation methods, achieving comparable performance to the state-of-the-art while reducing inference time by over 45x for reconstruction and 32x for generation. Code and models are available at https://github.com/Tencent/FlashVDM.
Modelling human variation in rating tasks is crucial for enabling AI systems for personalization, pluralistic model alignment, and computational social science. We propose representing individuals using value profiles -- natural language descriptions of underlying values compressed from in-context demonstrations -- along with a steerable decoder model to estimate ratings conditioned on a value profile or other rater information. To measure the predictive information in rater representations, we introduce an information-theoretic methodology. We find that demonstrations contain the most information, followed by value profiles and then demographics. However, value profiles offer advantages in terms of scrutability, interpretability, and steerability due to their compressed natural language format. Value profiles effectively compress the useful information from demonstrations (>70% information preservation). Furthermore, clustering value profiles to identify similarly behaving individuals better explains rater variation than the most predictive demographic groupings. Going beyond test set performance, we show that the decoder models interpretably change ratings according to semantic profile differences, are well-calibrated, and can help explain instance-level disagreement by simulating an annotator population. These results demonstrate that value profiles offer novel, predictive ways to describe individual variation beyond demographics or group information.
In this work, we propose an improved automorphism ensemble (AE) decoder for polar codes. With successive cancellation (SC) variant automorphisms, multiple decoding paths in the AE decoder produce their estimates of the transmitted codeword. In the proposed scheme, the decoding results from the multiple paths are cleverly utilized to generate a new channel output with which an additional decoding path is established. Performance evaluations clearly demonstrate that the proposed decoder achieves significantly improved error-rate performance as compared to the existing AE decoder for polar codes.
DETR-based methods, which use multi-layer transformer decoders to refine object queries iteratively, have shown promising performance in 3D indoor object detection. However, the scene point features in the transformer decoder remain fixed, leading to minimal contributions from later decoder layers, thereby limiting performance improvement. Recently, State Space Models (SSM) have shown efficient context modeling ability with linear complexity through iterative interactions between system states and inputs. Inspired by SSMs, we propose a new 3D object DEtection paradigm with an interactive STate space model (DEST). In the interactive SSM, we design a novel state-dependent SSM parameterization method that enables system states to effectively serve as queries in 3D indoor detection tasks. In addition, we introduce four key designs tailored to the characteristics of point cloud and SSM: The serialization and bidirectional scanning strategies enable bidirectional feature interaction among scene points within the SSM. The inter-state attention mechanism models the relationships between state points, while the gated feed-forward network enhances inter-channel correlations. To the best of our knowledge, this is the first method to model queries as system states and scene points as system inputs, which can simultaneously update scene point features and query features with linear complexity. Extensive experiments on two challenging datasets demonstrate the effectiveness of our DEST-based method. Our method improves the GroupFree baseline in terms of AP50 on ScanNet V2 (+5.3) and SUN RGB-D (+3.2) datasets. Based on the VDETR baseline, Our method sets a new SOTA on the ScanNetV2 and SUN RGB-D datasets.
Semi-supervised learning is of great significance in medical image segmentation by exploiting unlabeled data. Among its strategies, the co-training framework is prominent. However, previous co-training studies predominantly concentrate on network initialization variances and pseudo-label generation, while overlooking the equilibrium between information interchange and model diversity preservation. In this paper, we propose the Masked Image Consistency and Discrepancy Learning (MICD) framework with three key modules. The Masked Cross Pseudo Consistency (MCPC) module enriches context perception and small sample learning via pseudo-labeling across masked-input branches. The Cross Feature Consistency (CFC) module fortifies information exchange and model robustness by ensuring decoder feature consistency. The Cross Model Discrepancy (CMD) module utilizes EMA teacher networks to oversee outputs and preserve branch diversity. Together, these modules address existing limitations by focusing on fine-grained local information and maintaining diversity in a heterogeneous framework. Experiments on two public medical image datasets, AMOS and Synapse, demonstrate that our approach outperforms state-of-the-art methods.
Large foundation models trained on large-scale vision-language data can boost Open-Vocabulary Object Detection (OVD) via synthetic training data, yet the hand-crafted pipelines often introduce bias and overfit to specific prompts. We sidestep this issue by directly fusing hidden states from Large Language Models (LLMs) into detectors-an avenue surprisingly under-explored. This paper presents a systematic method to enhance visual grounding by utilizing decoder layers of the LLM of an MLLM. We introduce a zero-initialized cross-attention adapter to enable efficient knowledge fusion from LLMs to object detectors, a new approach called LED (LLM Enhanced Open-Vocabulary Object Detection). We find that intermediate LLM layers already encode rich spatial semantics; adapting only the early layers yields most of the gain. With Swin-T as the vision encoder, Qwen2-0.5B + LED lifts GroundingDINO by 3.82 % on OmniLabel at just 8.7 % extra GFLOPs, and a larger vision backbone pushes the improvement to 6.22 %. Extensive ablations on adapter variants, LLM scales and fusion depths further corroborate our design.
Hyperspectral Images (HSIs) are crucial across numerous fields but are hindered by the long acquisition times associated with traditional spectrometers. The Coded Aperture Snapshot Spectral Imaging (CASSI) system mitigates this issue through a compression technique that accelerates the acquisition process. However, reconstructing HSIs from compressed data presents challenges due to fixed spatial and spectral resolution constraints. This study introduces a novel method using implicit neural representation for continuous hyperspectral image reconstruction. We propose the Mixed Granularity Implicit Representation (MGIR) framework, which includes a Hierarchical Spectral-Spatial Implicit Encoder for efficient multi-scale implicit feature extraction. This is complemented by a Mixed-Granularity Local Feature Aggregator that adaptively integrates local features across scales, combined with a decoder that merges coordinate information for precise reconstruction. By leveraging implicit neural representations, the MGIR framework enables reconstruction at any desired spatial-spectral resolution, significantly enhancing the flexibility and adaptability of the CASSI system. Extensive experimental evaluations confirm that our model produces reconstructed images at arbitrary resolutions and matches state-of-the-art methods across varying spectral-spatial compression ratios. The code will be released at https://github.com/chh11/MGIR.