Ranking and recommendation systems are the foundation for numerous online experiences, ranging from search results to personalized content delivery. These systems have evolved into complex, multilayered architectures that leverage vast datasets and often incorporate thousands of predictive models. The maintenance and enhancement of these models is a labor intensive process that requires extensive feature engineering. This approach not only exacerbates technical debt but also hampers innovation in extending these systems to emerging problem domains. In this report, we present our research to address these challenges by utilizing a large foundation model with a textual interface for ranking and recommendation tasks. We illustrate several key advantages of our approach: (1) a single model can manage multiple predictive tasks involved in ranking and recommendation, (2) decoder models with textual interface due to their comprehension of reasoning capabilities, can generalize to new recommendation surfaces and out-of-domain problems, and (3) by employing natural language interfaces for task definitions and verbalizing member behaviors and their social connections, we eliminate the need for feature engineering and the maintenance of complex directed acyclic graphs of model dependencies. We introduce our research pre-production model, 360Brew V1.0, a 150B parameter, decoder-only model that has been trained and fine-tuned on LinkedIn's data and tasks. This model is capable of solving over 30 predictive tasks across various segments of the LinkedIn platform, achieving performance levels comparable to or exceeding those of current production systems based on offline metrics, without task-specific fine-tuning. Notably, each of these tasks is conventionally addressed by dedicated models that have been developed and maintained over multiple years by teams of a similar or larger size than our own.
Quantum computing is entering a transformative phase with the emergence of logical quantum processors, which hold the potential to tackle complex problems beyond classical capabilities. While significant progress has been made, applying quantum algorithms to real-world problems remains challenging. Hybrid quantum-classical techniques have been explored to bridge this gap, but they often face limitations in expressiveness, trainability, or scalability. In this work, we introduce conditional Generative Quantum Eigensolver (conditional-GQE), a context-aware quantum circuit generator powered by an encoder-decoder Transformer. Focusing on combinatorial optimization, we train our generator for solving problems with up to 10 qubits, exhibiting nearly perfect performance on new problems. By leveraging the high expressiveness and flexibility of classical generative models, along with an efficient preference-based training scheme, conditional-GQE provides a generalizable and scalable framework for quantum circuit generation. Our approach advances hybrid quantum-classical computing and contributes to accelerate the transition toward fault-tolerant quantum computing.
Lightweight, controllable, and physically plausible human motion synthesis is crucial for animation, virtual reality, robotics, and human-computer interaction applications. Existing methods often compromise between computational efficiency, physical realism, or spatial controllability. We propose FlexMotion, a novel framework that leverages a computationally lightweight diffusion model operating in the latent space, eliminating the need for physics simulators and enabling fast and efficient training. FlexMotion employs a multimodal pre-trained Transformer encoder-decoder, integrating joint locations, contact forces, joint actuations and muscle activations to ensure the physical plausibility of the generated motions. FlexMotion also introduces a plug-and-play module, which adds spatial controllability over a range of motion parameters (e.g., joint locations, joint actuations, contact forces, and muscle activations). Our framework achieves realistic motion generation with improved efficiency and control, setting a new benchmark for human motion synthesis. We evaluate FlexMotion on extended datasets and demonstrate its superior performance in terms of realism, physical plausibility, and controllability.
The Segment Anything Model (SAM) has set a new standard in interactive image segmentation, offering robust performance across various tasks. However, its significant computational requirements limit its deployment in real-time or resource-constrained environments. To address these challenges, we propose a novel knowledge distillation approach, KD SAM, which incorporates both encoder and decoder optimization through a combination of Mean Squared Error (MSE) and Perceptual Loss. This dual-loss framework captures structural and semantic features, enabling the student model to maintain high segmentation accuracy while reducing computational complexity. Based on the model evaluation on datasets, including Kvasir-SEG, ISIC 2017, Fetal Head Ultrasound, and Breast Ultrasound, we demonstrate that KD SAM achieves comparable or superior performance to the baseline models, with significantly fewer parameters. KD SAM effectively balances segmentation accuracy and computational efficiency, making it well-suited for real-time medical image segmentation applications in resource-constrained environments.
This paper presents a novel streaming end-to-end target-speaker speech recognition that addresses two critical limitations in systems: the handling of noisy enrollment utterances and specific enrollment phrase requirements. This paper proposes a robust Target-Speaker Recurrent Neural Network Transducer (TS-RNNT) with dual attention mechanisms for contextual biasing and overlapping enrollment processing. The model incorporates a text decoder and attention mechanism specifically designed to extract relevant speaker characteristics from noisy, overlapping enrollment audio. Experimental results on a synthesized dataset demonstrate the model's resilience, maintaining a Word Error Rate (WER) of 16.44% even with overlapping enrollment at 5dB Signal-to-Interference Ratio (SIR), compared to conventional approaches that degrade to WERs above 75% under similar conditions. This significant performance improvement, coupled with the model's semi-text-dependent enrollment capabilities, represents a substantial advancement toward more practical and versatile voice-controlled devices.
Accurate and real-time position estimation of the robotic arm on the patient's side is crucial for the success of remote robotic surgery in Tactile Internet environments. This paper proposes a predictive approach using the computationally efficient Transformer-based Informer model for position estimation, combined with a Four-State Hidden Markov Model (4-State HMM) to simulate realistic packet loss scenarios. The method effectively addresses network-induced delays, jitter, and packet loss, ensuring reliable performance in remote robotic surgery. The study evaluates the Informer model on the JIGSAWS dataset, demonstrating its capability to handle sequential data challenges caused by network uncertainties. Key features, including ProbSparse attention and a generative-style decoder, enhance prediction accuracy, computational speed, and memory efficiency. Results indicate that the proposed method achieves over 90 percent accuracy across varying network conditions. Furthermore, the Informer framework outperforms traditional models such as TCN, RNN, and LSTM, highlighting its suitability for real-time remote surgery applications.
We present FireRedASR, a family of large-scale automatic speech recognition (ASR) models for Mandarin, designed to meet diverse requirements in superior performance and optimal efficiency across various applications. FireRedASR comprises two variants: FireRedASR-LLM: Designed to achieve state-of-the-art (SOTA) performance and to enable seamless end-to-end speech interaction. It adopts an Encoder-Adapter-LLM framework leveraging large language model (LLM) capabilities. On public Mandarin benchmarks, FireRedASR-LLM (8.3B parameters) achieves an average Character Error Rate (CER) of 3.05%, surpassing the latest SOTA of 3.33% with an 8.4% relative CER reduction (CERR). It demonstrates superior generalization capability over industrial-grade baselines, achieving 24%-40% CERR in multi-source Mandarin ASR scenarios such as video, live, and intelligent assistant. FireRedASR-AED: Designed to balance high performance and computational efficiency and to serve as an effective speech representation module in LLM-based speech models. It utilizes an Attention-based Encoder-Decoder (AED) architecture. On public Mandarin benchmarks, FireRedASR-AED (1.1B parameters) achieves an average CER of 3.18%, slightly worse than FireRedASR-LLM but still outperforming the latest SOTA model with over 12B parameters. It offers a more compact size, making it suitable for resource-constrained applications. Moreover, both models exhibit competitive results on Chinese dialects and English speech benchmarks and excel in singing lyrics recognition. To advance research in speech processing, we release our models and inference code at https://github.com/FireRedTeam/FireRedASR.
Ensembling neural machine translation (NMT) models to produce higher-quality translations than the $L$ individual models has been extensively studied. Recent methods typically employ a candidate selection block (CSB) and an encoder-decoder fusion block (FB), requiring inference across \textit{all} candidate models, leading to significant computational overhead, generally $\Omega(L)$. This paper introduces \textbf{SmartGen}, a reinforcement learning (RL)-based strategy that improves the CSB by selecting a small, fixed number of candidates and identifying optimal groups to pass to the fusion block for each input sentence. Furthermore, previously, the CSB and FB were trained independently, leading to suboptimal NMT performance. Our DQN-based \textbf{SmartGen} addresses this by using feedback from the FB block as a reward during training. We also resolve a key issue in earlier methods, where candidates were passed to the FB without modification, by introducing a Competitive Correction Block (CCB). Finally, we validate our approach with extensive experiments on English-Hindi translation tasks in both directions.
Federated Class-Incremental Learning (FCIL) refers to a scenario where a dynamically changing number of clients collaboratively learn an ever-increasing number of incoming tasks. FCIL is known to suffer from local forgetting due to class imbalance at each client and global forgetting due to class imbalance across clients. We develop a mathematical framework for FCIL that formulates local and global forgetting. Then, we propose an approach called Hybrid Rehearsal (HR), which utilizes latent exemplars and data-free techniques to address local and global forgetting, respectively. HR employs a customized autoencoder designed for both data classification and the generation of synthetic data. To determine the embeddings of new tasks for all clients in the latent space of the encoder, the server uses the Lennard-Jones Potential formulations. Meanwhile, at the clients, the decoder decodes the stored low-dimensional latent space exemplars back to the high-dimensional input space, used to address local forgetting. To overcome global forgetting, the decoder generates synthetic data. Furthermore, our mathematical framework proves that our proposed approach HR can, in principle, tackle the two local and global forgetting challenges. In practice, extensive experiments demonstrate that while preserving privacy, our proposed approach outperforms the state-of-the-art baselines on multiple FCIL benchmarks with low compute and memory footprints.
The reasoning abilities of Large Language Models (LLMs) can be improved by structurally denoising their weights, yet existing techniques primarily focus on denoising the feed-forward network (FFN) of the transformer block, and can not efficiently utilise the Multi-head Attention (MHA) block, which is the core of transformer architectures. To address this issue, we propose a novel intuitive framework that, at its very core, performs MHA compression through a multi-head tensorisation process and the Tucker decomposition. This enables both higher-dimensional structured denoising and compression of the MHA weights, by enforcing a shared higher-dimensional subspace across the weights of the multiple attention heads. We demonstrate that this approach consistently enhances the reasoning capabilities of LLMs across multiple benchmark datasets, and for both encoder-only and decoder-only architectures, while achieving compression rates of up to $\sim 250$ times in the MHA weights, all without requiring any additional data, training, or fine-tuning. Furthermore, we show that the proposed method can be seamlessly combined with existing FFN-only-based denoising techniques to achieve further improvements in LLM reasoning performance.
We present a deep learning model, dubbed Glissando-Net, to simultaneously estimate the pose and reconstruct the 3D shape of objects at the category level from a single RGB image. Previous works predominantly focused on either estimating poses(often at the instance level), or reconstructing shapes, but not both. Glissando-Net is composed of two auto-encoders that are jointly trained, one for RGB images and the other for point clouds. We embrace two key design choices in Glissando-Net to achieve a more accurate prediction of the 3D shape and pose of the object given a single RGB image as input. First, we augment the feature maps of the point cloud encoder and decoder with transformed feature maps from the image decoder, enabling effective 2D-3D interaction in both training and prediction. Second, we predict both the 3D shape and pose of the object in the decoder stage. This way, we better utilize the information in the 3D point clouds presented only in the training stage to train the network for more accurate prediction. We jointly train the two encoder-decoders for RGB and point cloud data to learn how to pass latent features to the point cloud decoder during inference. In testing, the encoder of the 3D point cloud is discarded. The design of Glissando-Net is inspired by codeSLAM. Unlike codeSLAM, which targets 3D reconstruction of scenes, we focus on pose estimation and shape reconstruction of objects, and directly predict the object pose and a pose invariant 3D reconstruction without the need of the code optimization step. Extensive experiments, involving both ablation studies and comparison with competing methods, demonstrate the efficacy of our proposed method, and compare favorably with the state-of-the-art.
The dominance of large decoder-only language models has overshadowed encoder-decoder architectures, despite their fundamental efficiency advantages in sequence processing. For small language models (SLMs) - those with 1 billion parameters or fewer - our systematic analysis across GPU, CPU, and NPU platforms reveals that encoder-decoder architectures achieve 47% lower first-token latency and 4.7x higher throughput compared to decoder-only models on edge devices. These gains may be attributed to encoder-decoder's one-time input processing and efficient separation of understanding and generation phases. We introduce a novel knowledge distillation framework that enables encoder-decoder models to leverage capabilities from large scalable decoder-only teachers while preserving their architectural advantages, achieving up to 6 average performance points improvement across diverse tasks, with significant gains in asymmetric sequence tasks where input and output distributions can benefit from different processing approaches. When combined with modern advances like Rotary Positional Embeddings (RoPE) and Vision encoders, our systematic investigation demonstrates that encoder-decoder architectures provide a more practical path toward deploying capable language models in resource-constrained environments. Our findings challenge the prevailing trend toward decoder-only scaling, showing that architectural choices become increasingly crucial as parameter budgets decrease, particularly for on-device and edge deployments where computational efficiency is paramount.
This work introduces a novel, fully differentiable linear-time complexity transformer decoder and a transformer decoder to correct 5G New Radio (NR) LDPC. We propose a scalable approach to decode linear block codes with $O(n)$ complexity rather than $O(n^2)$ for regular transformers. The architectures' performances are compared to Belief Propagation (BP), the production-level decoding algorithm used for 5G New Radio (NR) LDPC codes. We achieve bit error rate performance that matches a regular Transformer decoder and surpases one iteration BP, also achieving competitive time performance against BP, even for larger block codes. We utilize Sionna, Nvidia's 5G & 6G physical layer research software, for reproducible results.
Dense prediction tasks have enjoyed a growing complexity of encoder architectures, decoders, however, have remained largely the same. They rely on individual blocks decoding intermediate feature maps sequentially. We introduce banks, shared structures that are used by each decoding block to provide additional context in the decoding process. These structures, through applying them via resampling and feature fusion, improve performance on depth estimation for state-of-the-art transformer-based architectures on natural and synthetic images whilst training on large-scale datasets.
Single-view novel view synthesis (NVS) is a notorious problem due to its ill-posed nature, and often requires large, computationally expensive approaches to produce tangible results. In this paper, we propose CheapNVS: a fully end-to-end approach for narrow baseline single-view NVS based on a novel, efficient multiple encoder/decoder design trained in a multi-stage fashion. CheapNVS first approximates the laborious 3D image warping with lightweight learnable modules that are conditioned on the camera pose embeddings of the target view, and then performs inpainting on the occluded regions in parallel to achieve significant performance gains. Once trained on a subset of Open Images dataset, CheapNVS outperforms the state-of-the-art despite being 10 times faster and consuming 6% less memory. Furthermore, CheapNVS runs comfortably in real-time on mobile devices, reaching over 30 FPS on a Samsung Tab 9+.
The accurate detection and segmentation of pavement distresses, particularly tiny and small cracks, are critical for early intervention and preventive maintenance in transportation infrastructure. Traditional manual inspection methods are labor-intensive and inconsistent, while existing deep learning models struggle with fine-grained segmentation and computational efficiency. To address these challenges, this study proposes Context-CrackNet, a novel encoder-decoder architecture featuring the Region-Focused Enhancement Module (RFEM) and Context-Aware Global Module (CAGM). These innovations enhance the model's ability to capture fine-grained local details and global contextual dependencies, respectively. Context-CrackNet was rigorously evaluated on ten publicly available crack segmentation datasets, covering diverse pavement distress scenarios. The model consistently outperformed 9 state-of-the-art segmentation frameworks, achieving superior performance metrics such as mIoU and Dice score, while maintaining competitive inference efficiency. Ablation studies confirmed the complementary roles of RFEM and CAGM, with notable improvements in mIoU and Dice score when both modules were integrated. Additionally, the model's balance of precision and computational efficiency highlights its potential for real-time deployment in large-scale pavement monitoring systems.
Vision-Language Models (VLMs) have achieved notable success in multimodal tasks but face practical limitations due to the quadratic complexity of decoder attention mechanisms and autoregressive generation. Existing methods like FASTV and VTW have achieved notable results in reducing redundant visual tokens, but these approaches focus on pruning tokens in a single forward pass without systematically analyzing the redundancy of visual tokens throughout the entire generation process. In this paper, we introduce a dynamic pruning strategy tailored for VLMs, namedDynamic Rate (DyRate), which progressively adjusts the compression rate during generation. Our analysis of the distribution of attention reveals that the importance of visual tokens decreases throughout the generation process, inspiring us to adopt a more aggressive compression rate. By integrating a lightweight predictor based on attention distribution, our approach enables flexible adjustment of pruning rates based on the attention distribution. Our experimental results demonstrate that our method not only reduces computational demands but also maintains the quality of responses.
Referring video object segmentation (RVOS) aims to segment target objects throughout a video based on a text description. Despite notable progress in recent years, current RVOS models remain struggle to handle complicated object descriptions due to their limited video-language understanding. To address this limitation, we present \textbf{ReferDINO}, an end-to-end RVOS model that inherits strong vision-language understanding from the pretrained visual grounding foundation models, and is further endowed with effective temporal understanding and object segmentation capabilities. In ReferDINO, we contribute three technical innovations for effectively adapting the foundation models to RVOS: 1) an object-consistent temporal enhancer that capitalizes on the pretrained object-text representations to enhance temporal understanding and object consistency; 2) a grounding-guided deformable mask decoder that integrates text and grounding conditions to generate accurate object masks; 3) a confidence-aware query pruning strategy that significantly improves the object decoding efficiency without compromising performance. We conduct extensive experiments on five public RVOS benchmarks to demonstrate that our proposed ReferDINO outperforms state-of-the-art methods significantly. Project page: \url{https://isee-laboratory.github.io/ReferDINO}
Existing formal verification methods for image-based neural network controllers in autonomous vehicles often struggle with high-dimensional inputs, computational inefficiency, and a lack of explainability. These challenges make it difficult to ensure safety and reliability, as processing high-dimensional image data is computationally intensive and neural networks are typically treated as black boxes. To address these issues, we propose SEVIN (Scalable and Explainable Verification of Image-Based Neural Network Controllers), a framework that leverages a Variational Autoencoders (VAE) to encode high-dimensional images into a lower-dimensional, explainable latent space. By annotating latent variables with corresponding control actions, we generate convex polytopes that serve as structured input spaces for verification, significantly reducing computational complexity and enhancing scalability. Integrating the VAE's decoder with the neural network controller allows for formal and robustness verification using these explainable polytopes. Our approach also incorporates robustness verification under real-world perturbations by augmenting the dataset and retraining the VAE to capture environmental variations. Experimental results demonstrate that SEVIN achieves efficient and scalable verification while providing explainable insights into controller behavior, bridging the gap between formal verification techniques and practical applications in safety-critical systems.
Bearing data compression is vital to manage the large volumes of data generated during condition monitoring. In this paper, a novel asymmetrical autoencoder with a lifting wavelet transform (LWT) layer is developed to compress bearing sensor data. The encoder part of the network consists of a convolutional layer followed by a wavelet filterbank layer. Specifically, a dual-channel convolutional block with diverse convolutional kernel sizes and varying processing depths is integrated into the wavelet filterbank layer to enable comprehensive feature extraction from the wavelet domain. Additionally, the adaptive hard-thresholding nonlinearity is applied to remove redundant components while denoising the primary wavelet coefficients. On the decoder side, inverse LWT, along with multiple linear layers and activation functions, is employed to reconstruct the original signals. Furthermore, to enhance compression efficiency, a sparsity constraint is introduced during training to impose sparsity on the latent representations. The experimental results demonstrate that the proposed approach achieves superior data compression performance compared to state-of-the-art methods.