Neural transducers (NT) provide an effective framework for speech streaming, demonstrating strong performance in automatic speech recognition (ASR). However, the application of NT to speech translation (ST) remains challenging, as existing approaches struggle with word reordering and performance degradation when jointly modeling ASR and ST, resulting in a gap with attention-based encoder-decoder (AED) models. Existing NT-based ST approaches also suffer from high computational training costs. To address these issues, we propose HENT-SRT (Hierarchical Efficient Neural Transducer for Speech Recognition and Translation), a novel framework that factorizes ASR and translation tasks to better handle reordering. To ensure robust ST while preserving ASR performance, we use self-distillation with CTC consistency regularization. Moreover, we improve computational efficiency by incorporating best practices from ASR transducers, including a down-sampled hierarchical encoder, a stateless predictor, and a pruned transducer loss to reduce training complexity. Finally, we introduce a blank penalty during decoding, reducing deletions and improving translation quality. Our approach is evaluated on three conversational datasets Arabic, Spanish, and Mandarin achieving new state-of-the-art performance among NT models and substantially narrowing the gap with AED-based systems.
This paper reports on the development of a large-scale speech recognition model, Whale. Similar to models such as Whisper and OWSM, Whale leverages both a large model size and a diverse, extensive dataset. Whale's architecture integrates w2v-BERT self-supervised model, an encoder-decoder backbone built on E-Branchformer, and a joint CTC-attention decoding strategy. The training corpus comprises varied speech data, of not only public corpora but also in-house data, thereby enhancing the model's robustness to different speaking styles and acoustic conditions. Through evaluations on multiple benchmarks, Whale achieved comparable performance to existing models. In particular, it achieves a word error rate of 2.4% on the Librispeech test-clean set and a character error rate of 3.4% on the CSJ eval3 set, outperforming Whisper large-v3 and OWSM v3.1.
In the context of inverse problems $\bf y = Ax$, sparse recovery offers a powerful paradigm shift by enabling the stable solution of ill-posed or underdetermined systems through the exploitation of structure, particularly sparsity. Sparse regularization techniques via $\ell_0$- or $\ell_1$-norm minimization encourage solutions $\bf x$ that are both consistent with observations $\bf y$ and parsimonious in representation, often yielding physically meaningful interpretations. In this work, we address the classical inverse problem under the challenging condition where the sensing operator $\bf A$ is unknown and only a limited set of observation-target pairs $\{ \bf x,\bf y \}$ is available. We propose a novel neural architecture, TRUST, that integrates the attention mechanism of Transformers with the decoder pathway of a UNet to simultaneously learn the sensing operator and reconstruct the sparse signal. The TRUST model incorporates a Transformer-based encoding branch to capture long-range dependencies and estimate sparse support, which then guides a U-Net-style decoder to refine reconstruction through multiscale feature integration. The skip connections between the transformer stages and the decoder not only enhance image quality but also enable the decoder to access image features at different levels of abstraction. This hybrid architecture enables more accurate and robust recovery by combining global context with local details. Experimental results demonstrate that TRUST significantly outperforms traditional sparse recovery methods and standalone U-Net models, achieving superior performance in SSIM and PSNR metrics while effectively suppressing hallucination artifacts that commonly plague deep learning-based inverse solvers.
This paper proposes a novel Mixture of Prompt-Experts based Speaker Adaptation approach (MOPSA) for elderly speech recognition. It allows zero-shot, real-time adaptation to unseen speakers, and leverages domain knowledge tailored to elderly speakers. Top-K most distinctive speaker prompt clusters derived using K-means serve as experts. A router network is trained to dynamically combine clustered prompt-experts. Acoustic and language level variability among elderly speakers are modelled using separate encoder and decoder prompts for Whisper. Experiments on the English DementiaBank Pitt and Cantonese JCCOCC MoCA elderly speech datasets suggest that online MOPSA adaptation outperforms the speaker-independent (SI) model by statistically significant word error rate (WER) or character error rate (CER) reductions of 0.86% and 1.47% absolute (4.21% and 5.40% relative). Real-time factor (RTF) speed-up ratios of up to 16.12 times are obtained over offline batch-mode adaptation.
Understanding fine-grained object affordances is imperative for robots to manipulate objects in unstructured environments given open-ended task instructions. However, existing methods of visual affordance predictions often rely on manually annotated data or conditions only on a predefined set of tasks. We introduce UAD (Unsupervised Affordance Distillation), a method for distilling affordance knowledge from foundation models into a task-conditioned affordance model without any manual annotations. By leveraging the complementary strengths of large vision models and vision-language models, UAD automatically annotates a large-scale dataset with detailed $<$instruction, visual affordance$>$ pairs. Training only a lightweight task-conditioned decoder atop frozen features, UAD exhibits notable generalization to in-the-wild robotic scenes and to various human activities, despite only being trained on rendered objects in simulation. Using affordance provided by UAD as the observation space, we show an imitation learning policy that demonstrates promising generalization to unseen object instances, object categories, and even variations in task instructions after training on as few as 10 demonstrations. Project website: https://unsup-affordance.github.io/
In recent years, end-to-end approaches have made notable progress in addressing the challenge of speaker diarization, which involves segmenting and identifying speakers in multi-talker recordings. One such approach, Encoder-Decoder Attractors (EDA), has been proposed to handle variable speaker counts as well as better guide the network during training. In this study, we extend the attractor paradigm by moving beyond direct speaker modeling and instead focus on representing more detailed `speaker attributes' through a multi-stage process of intermediate representations. Additionally, we enhance the architecture by replacing transformers with conformers, a convolution-augmented transformer, to model local dependencies. Experiments demonstrate improved diarization performance on the CALLHOME dataset.
In this paper, we fine-tune encoder-only and decoder-only large language models (LLMs) to detect pathologies in IEEE 802.11 networks, commonly known as WiFi. Our approach involves manually crafting prompts followed by fine-tuning. Evaluations show that the sequential model achieves high detection accuracy using labeled data, while the causal model performs equally well for unlabeled data.
We develop a communication-theoretic framework for privacy-aware and resilient decision making in cyber-physical systems under misaligned objectives between the encoder and the decoder. The encoder observes two correlated signals ($X$,$\theta$) and transmits a finite-rate message $Z$ to aid a legitimate controller (the decoder) in estimating $X+\theta$, while an eavesdropper intercepts $Z$ to infer the private parameter $\theta$. Unlike conventional setups where encoder and decoder share a common MSE objective, here the encoder minimizes a Lagrangian that balances legitimate control fidelity and the privacy leakage about $\theta$. In contrast, the decoder's goal is purely to minimize its own estimation error without regard for privacy. We analyze fully, partially, and non-revealing strategies that arise from this conflict, and characterize optimal linear encoders when the rate constraints are lifted. For finite-rate channels, we employ gradient-based methods to compute the optimal controllers. Numerical experiments illustrate how tuning the privacy parameter shapes the trade-off between control performance and resilience against unauthorized inferences.
Transformer-based models have achieved strong performance in remote sensing image captioning by capturing long-range dependencies and contextual information. However, their practical deployment is hindered by high computational costs, especially in multi-modal frameworks that employ separate transformer-based encoders and decoders. In addition, existing remote sensing image captioning models primarily focus on high-level semantic extraction while often overlooking fine-grained structural features such as edges, contours, and object boundaries. To address these challenges, a lightweight transformer architecture is proposed by reducing the dimensionality of the encoder layers and employing a distilled version of GPT-2 as the decoder. A knowledge distillation strategy is used to transfer knowledge from a more complex teacher model to improve the performance of the lightweight network. Furthermore, an edge-aware enhancement strategy is incorporated to enhance image representation and object boundary understanding, enabling the model to capture fine-grained spatial details in remote sensing images. Experimental results demonstrate that the proposed approach significantly improves caption quality compared to state-of-the-art methods.
Camouflaged Object Detection (COD) presents inherent challenges due to the subtle visual differences between targets and their backgrounds. While existing methods have made notable progress, there remains significant potential for post-processing refinement that has yet to be fully explored. To address this limitation, we propose the Uncertainty-Masked Bernoulli Diffusion (UMBD) model, the first generative refinement framework specifically designed for COD. UMBD introduces an uncertainty-guided masking mechanism that selectively applies Bernoulli diffusion to residual regions with poor segmentation quality, enabling targeted refinement while preserving correctly segmented areas. To support this process, we design the Hybrid Uncertainty Quantification Network (HUQNet), which employs a multi-branch architecture and fuses uncertainty from multiple sources to improve estimation accuracy. This enables adaptive guidance during the generative sampling process. The proposed UMBD framework can be seamlessly integrated with a wide range of existing Encoder-Decoder-based COD models, combining their discriminative capabilities with the generative advantages of diffusion-based refinement. Extensive experiments across multiple COD benchmarks demonstrate consistent performance improvements, achieving average gains of 5.5% in MAE and 3.2% in weighted F-measure with only modest computational overhead. Code will be released.
Beyond traditional hybrid-based video codec, generative video codec could achieve promising compression performance by evolving high-dimensional signals into compact feature representations for bitstream compactness at the encoder side and developing explicit motion fields as intermediate supervision for high-quality reconstruction at the decoder side. This paradigm has achieved significant success in face video compression. However, compared to facial videos, human body videos pose greater challenges due to their more complex and diverse motion patterns, i.e., when using explicit motion guidance for Generative Human Video Coding (GHVC), the reconstruction results could suffer severe distortions and inaccurate motion. As such, this paper highlights the limitations of explicit motion-based approaches for human body video compression and investigates the GHVC performance improvement with the aid of Implicit Motion Transformation, namely IMT. In particular, we propose to characterize complex human body signal into compact visual features and transform these features into implicit motion guidance for signal reconstruction. Experimental results demonstrate the effectiveness of the proposed IMT paradigm, which can facilitate GHVC to achieve high-efficiency compression and high-fidelity synthesis.
Deterministic embeddings learned by contrastive learning (CL) methods such as SimCLR and SupCon achieve state-of-the-art performance but lack a principled mechanism for uncertainty quantification. We propose Variational Contrastive Learning (VCL), a decoder-free framework that maximizes the evidence lower bound (ELBO) by interpreting the InfoNCE loss as a surrogate reconstruction term and adding a KL divergence regularizer to a uniform prior on the unit hypersphere. We model the approximate posterior $q_\theta(z|x)$ as a projected normal distribution, enabling the sampling of probabilistic embeddings. Our two instantiations--VSimCLR and VSupCon--replace deterministic embeddings with samples from $q_\theta(z|x)$ and incorporate a normalized KL term into the loss. Experiments on multiple benchmarks demonstrate that VCL mitigates dimensional collapse, enhances mutual information with class labels, and matches or outperforms deterministic baselines in classification accuracy, all the while providing meaningful uncertainty estimates through the posterior model. VCL thus equips contrastive learning with a probabilistic foundation, serving as a new basis for contrastive approaches.
Encoder-based transformer models are central to biomedical and clinical Natural Language Processing (NLP), as their bidirectional self-attention makes them well-suited for efficiently extracting structured information from unstructured text through discriminative tasks. However, encoders have seen slower development compared to decoder models, leading to limited domain adaptation in biomedical and clinical settings. We introduce BioClinical ModernBERT, a domain-adapted encoder that builds on the recent ModernBERT release, incorporating long-context processing and substantial improvements in speed and performance for biomedical and clinical NLP. BioClinical ModernBERT is developed through continued pretraining on the largest biomedical and clinical corpus to date, with over 53.5 billion tokens, and addresses a key limitation of prior clinical encoders by leveraging 20 datasets from diverse institutions, domains, and geographic regions, rather than relying on data from a single source. It outperforms existing biomedical and clinical encoders on four downstream tasks spanning a broad range of use cases. We release both base (150M parameters) and large (396M parameters) versions of BioClinical ModernBERT, along with training checkpoints to support further research.
Lung cancer remains one of the most prevalent and fatal diseases worldwide, demanding accurate and timely diagnosis and treatment. Recent advancements in large AI models have significantly enhanced medical image understanding and clinical decision-making. This review systematically surveys the state-of-the-art in applying large AI models to lung cancer screening, diagnosis, prognosis, and treatment. We categorize existing models into modality-specific encoders, encoder-decoder frameworks, and joint encoder architectures, highlighting key examples such as CLIP, BLIP, Flamingo, BioViL-T, and GLoRIA. We further examine their performance in multimodal learning tasks using benchmark datasets like LIDC-IDRI, NLST, and MIMIC-CXR. Applications span pulmonary nodule detection, gene mutation prediction, multi-omics integration, and personalized treatment planning, with emerging evidence of clinical deployment and validation. Finally, we discuss current limitations in generalizability, interpretability, and regulatory compliance, proposing future directions for building scalable, explainable, and clinically integrated AI systems. Our review underscores the transformative potential of large AI models to personalize and optimize lung cancer care.
This work addresses a fundamental barrier in recommender systems: the inability to generalize across domains without extensive retraining. Traditional ID-based approaches fail entirely in cold-start and cross-domain scenarios where new users or items lack sufficient interaction history. Inspired by foundation models' cross-domain success, we develop a foundation model for sequential recommendation that achieves genuine zero-shot generalization capabilities. Our approach fundamentally departs from existing ID-based methods by deriving item representations exclusively from textual features. This enables immediate embedding of any new item without model retraining. We introduce unified item tokenization with Finite Scalar Quantization that transforms heterogeneous textual descriptions into standardized discrete tokens. This eliminates domain barriers that plague existing systems. Additionally, the framework features hybrid bidirectional-causal attention that captures both intra-item token coherence and inter-item sequential dependencies. An efficient catalog-aware beam search decoder enables real-time token-to-item mapping. Unlike conventional approaches confined to their training domains, RecGPT naturally bridges diverse recommendation contexts through its domain-invariant tokenization mechanism. Comprehensive evaluations across six datasets and industrial scenarios demonstrate consistent performance advantages.
To tackle the threat of fake news, the task of detecting and grounding multi-modal media manipulation DGM4 has received increasing attention. However, most state-of-the-art methods fail to explore the fine-grained consistency within local content, usually resulting in an inadequate perception of detailed forgery and unreliable results. In this paper, we propose a novel approach named Contextual-Semantic Consistency Learning (CSCL) to enhance the fine-grained perception ability of forgery for DGM4. Two branches for image and text modalities are established, each of which contains two cascaded decoders, i.e., Contextual Consistency Decoder (CCD) and Semantic Consistency Decoder (SCD), to capture within-modality contextual consistency and across-modality semantic consistency, respectively. Both CCD and SCD adhere to the same criteria for capturing fine-grained forgery details. To be specific, each module first constructs consistency features by leveraging additional supervision from the heterogeneous information of each token pair. Then, the forgery-aware reasoning or aggregating is adopted to deeply seek forgery cues based on the consistency features. Extensive experiments on DGM4 datasets prove that CSCL achieves new state-of-the-art performance, especially for the results of grounding manipulated content. Codes and weights are avaliable at https://github.com/liyih/CSCL.
Pretrained models have demonstrated impressive success in many modalities such as language and vision. Recent works facilitate the pretraining paradigm in imaging research. Transients are a novel modality, which are captured for an object as photon counts versus arrival times using a precisely time-resolved sensor. In particular for non-line-of-sight (NLOS) scenarios, transients of hidden objects are measured beyond the sensor's direct line of sight. Using NLOS transients, the majority of previous works optimize volume density or surfaces to reconstruct the hidden objects and do not transfer priors learned from datasets. In this work, we present a masked autoencoder for modeling transient imaging, or MARMOT, to facilitate NLOS applications. Our MARMOT is a self-supervised model pretrianed on massive and diverse NLOS transient datasets. Using a Transformer-based encoder-decoder, MARMOT learns features from partially masked transients via a scanning pattern mask (SPM), where the unmasked subset is functionally equivalent to arbitrary sampling, and predicts full measurements. Pretrained on TransVerse-a synthesized transient dataset of 500K 3D models-MARMOT adapts to downstream imaging tasks using direct feature transfer or decoder finetuning. Comprehensive experiments are carried out in comparisons with state-of-the-art methods. Quantitative and qualitative results demonstrate the efficiency of our MARMOT.
We present the B-spline Encoded Action Sequence Tokenizer (BEAST), a novel action tokenizer that encodes action sequences into compact discrete or continuous tokens using B-splines. In contrast to existing action tokenizers based on vector quantization or byte pair encoding, BEAST requires no separate tokenizer training and consistently produces tokens of uniform length, enabling fast action sequence generation via parallel decoding. Leveraging our B-spline formulation, BEAST inherently ensures generating smooth trajectories without discontinuities between adjacent segments. We extensively evaluate BEAST by integrating it with three distinct model architectures: a Variational Autoencoder (VAE) with continuous tokens, a decoder-only Transformer with discrete tokens, and Florence-2, a pretrained Vision-Language Model with an encoder-decoder architecture, demonstrating BEAST's compatibility and scalability with large pretrained models. We evaluate BEAST across three established benchmarks consisting of 166 simulated tasks and on three distinct robot settings with a total of 8 real-world tasks. Experimental results demonstrate that BEAST (i) significantly reduces both training and inference computational costs, and (ii) consistently generates smooth, high-frequency control signals suitable for continuous control tasks while (iii) reliably achieves competitive task success rates compared to state-of-the-art methods.
Real-time reconstruction of dynamic 3D scenes from uncalibrated video streams is crucial for numerous real-world applications. However, existing methods struggle to jointly address three key challenges: 1) processing uncalibrated inputs in real time, 2) accurately modeling dynamic scene evolution, and 3) maintaining long-term stability and computational efficiency. To this end, we introduce StreamSplat, the first fully feed-forward framework that transforms uncalibrated video streams of arbitrary length into dynamic 3D Gaussian Splatting (3DGS) representations in an online manner, capable of recovering scene dynamics from temporally local observations. We propose two key technical innovations: a probabilistic sampling mechanism in the static encoder for 3DGS position prediction, and a bidirectional deformation field in the dynamic decoder that enables robust and efficient dynamic modeling. Extensive experiments on static and dynamic benchmarks demonstrate that StreamSplat consistently outperforms prior works in both reconstruction quality and dynamic scene modeling, while uniquely supporting online reconstruction of arbitrarily long video streams. Code and models are available at https://github.com/nickwzk/StreamSplat.
A significant portion of the energy consumed by Large Language Models (LLMs) arises from their inference processes; hence developing energy-efficient methods for inference is crucial. While several techniques exist for inference optimization, output compression remains relatively unexplored, with only a few preliminary efforts addressing this aspect. In this work, we first benchmark 12 decoder-only LLMs across 5 datasets, revealing that these models often produce responses that are substantially longer than necessary. We then conduct a comprehensive quality assessment of LLM responses, formally defining six information categories present in LLM responses. We show that LLMs often tend to include redundant or additional information besides the minimal answer. To address this issue of long responses by LLMs, we explore several simple and intuitive prompt-engineering strategies. Empirical evaluation shows that appropriate prompts targeting length reduction and controlling information content can achieve significant energy optimization between 25-60\% by reducing the response length while preserving the quality of LLM responses.