Yichen Lu

Xuankai Chang, Brian Yan, Kwanghee Choi et al. · cmu, meta-ai

Speech signals, typically sampled at rates in the tens of thousands per second, contain redundancies, evoking inefficiencies in sequence modeling. High-dimensional speech features such as spectrograms are often used as the input for the subsequent model. However, they can still be redundant. Recent investigations proposed the use of discrete speech units derived from self-supervised learning representations, which significantly compresses the size of speech data. Applying various methods, such as de-duplication and subword modeling, can further compress the speech sequence length. Hence, training time is significantly reduced while retaining notable performance. In this study, we undertake a comprehensive and systematic exploration into the application of discrete units within end-to-end speech processing models. Experiments on 12 automatic speech recognition, 3 speech translation, and 1 spoken language understanding corpora demonstrate that discrete units achieve reasonably good results in almost all the settings. We intend to release our configurations and trained models to foster future research efforts.

Yihan Wu, Yifan Peng, Yichen Lu et al. · nvidia

Visual signals can enhance audiovisual speech recognition accuracy by providing additional contextual information. Given the complexity of visual signals, an audiovisual speech recognition model requires robust generalization capabilities across diverse video scenarios, presenting a significant challenge. In this paper, we introduce EVA, leveraging the mixture-of-Experts for audioVisual ASR to perform robust speech recognition for ``in-the-wild'' videos. Specifically, we first encode visual information into visual tokens sequence and map them into speech space by a lightweight projection. Then, we build EVA upon a robust pretrained speech recognition model, ensuring its generalization ability. Moreover, to incorporate visual information effectively, we inject visual information into the ASR model through a mixture-of-experts module. Experiments show our model achieves state-of-the-art results on three benchmarks, which demonstrates the generalization ability of EVA across diverse video domains.

Ruijie Wang, Baoyu Li, Yichen Lu et al.

This paper studies speculative reasoning task on real-world knowledge graphs (KG) that contain both \textit{false negative issue} (i.e., potential true facts being excluded) and \textit{false positive issue} (i.e., unreliable or outdated facts being included). State-of-the-art methods fall short in the speculative reasoning ability, as they assume the correctness of a fact is solely determined by its presence in KG, making them vulnerable to false negative/positive issues. The new reasoning task is formulated as a noisy Positive-Unlabeled learning problem. We propose a variational framework, namely nPUGraph, that jointly estimates the correctness of both collected and uncollected facts (which we call \textit{label posterior}) and updates model parameters during training. The label posterior estimation facilitates speculative reasoning from two perspectives. First, it improves the robustness of a label posterior-aware graph encoder against false positive links. Second, it identifies missing facts to provide high-quality grounds of reasoning. They are unified in a simple yet effective self-training procedure. Empirically, extensive experiments on three benchmark KG and one Twitter dataset with various degrees of false negative/positive cases demonstrate the effectiveness of nPUGraph.

Xu Ling, Yichen Lu, Wenqi Xu et al.

Although deep learning has significantly improved Face Recognition (FR), dramatic performance deterioration may occur when processing Low Resolution (LR) faces. To alleviate this, approaches based on unified feature space are proposed with the sacrifice under High Resolution (HR) circumstances. To deal with the huge domain gap between HR and LR domains and achieve the best on both domains, we first took a closer look at the impacts of several resolution augmentations and then analyzed the difficulty of LR samples from the perspective of the model gradient produced by different resolution samples. Besides, we also find that the introduction of some resolutions could help the learning of lower resolutions. Based on these, we divide the LR samples into three difficulties according to the resolution and propose a more effective Multi-Resolution Augmentation. Then, due to the rapidly increasing domain gap as the resolution decreases, we carefully design a novel and effective metric loss based on a LogExp distance function that provides decent gradients to prevent oscillation near the convergence point or tolerance to small distance errors; it could also dynamically adjust the penalty for errors in different dimensions, allowing for more optimization of dimensions with large errors. Combining these two insights, our model could learn more general knowledge in a wide resolution range of images and balanced results can be achieved by our extremely simple framework. Moreover, the augmentations and metrics are the cornerstones of LRFR, so our method could be considered a new baseline for the LRFR task. Experiments on the LRFR datasets: SCface, XQLFW, and large-scale LRFR dataset: TinyFace demonstrate the effectiveness of our methods, while the degradation on HRFR datasets is significantly reduced.

Yichen Lu, Jiaqi Song, Xuankai Chang et al.

In this work, we present SynesLM, an unified model which can perform three multimodal language understanding tasks: audio-visual automatic speech recognition(AV-ASR) and visual-aided speech/machine translation(VST/VMT). Unlike previous research that focused on lip motion as visual cues for speech signals, our work explores more general visual information within entire frames, such as objects and actions. Additionally, we use synthetic image data to enhance the correlation between image and speech data. We benchmark SynesLM against the How2 dataset, demonstrating performance on par with state-of-the-art (SOTA) models dedicated to AV-ASR while maintaining our multitasking framework. Remarkably, for zero-shot AV-ASR, SynesLM achieved SOTA performance by lowering the Word Error Rate (WER) from 43.4% to 39.4% on the VisSpeech Dataset. Furthermore, our results in VST and VMT outperform the previous results, improving the BLEU score to 43.5 from 37.2 for VST, and to 54.8 from 54.4 for VMT.

Jiatong Shi, Jionghao Han, Yichen Lu et al.

Role-play has become a key testbed for generative models, expanding from text-only dialogue to multimodal interaction. Extending role-play to speech captures prosody, emotion, and delivery, but also poses new evaluation challenges. Current pipelines often use audio large language models (ALLMs) as zero-shot judges, which miss paralinguistic cues, collapse multiple aspects into coarse scores, and rely on synthetic speech references that fail to reflect real-world roles. We present Speech-DRAME, a unified framework that contributes at three levels: (i) Speech-DRAME-EvalBench, an evaluation benchmark with bilingual human-annotated data and protocols for training and testing speech evaluation models (SEMs), (ii) DRAME-Eval, a fine-tuned evaluation model, which substantially outperforms zero-shot and few-shot ALLMs, and (iii) Speech-DRAME-RoleBench, a speech role-play benchmark that leverages DRAME-Eval as an automatic judge to compare speech foundation models (SFMs). Speech-DRAME distinguishes between two complementary evaluation strategies: Archetype Evaluation, a top-down approach measuring adherence to broad role archetypes, and Realism Evaluation, a bottom-up approach grounded in real human speech that emphasizes nuanced role quality. Compared to zero-shot ALLM judges, DRAME-Eval achieves stronger agreement with human ratings (Pearson correlation from 0.480 to 0.629 in archetypes, and 0.390 to 0.625 in realism). By integrating transparent benchmark resources, modeling approaches, and system-level evaluation, Speech-DRAME provides the first comprehensive, reproducible foundation for assessing spoken role-play.

Yichen Lu, Wei Dai, Jiaen Liu et al.

LLM-based translation agents have achieved highly human-like translation results and are capable of handling longer and more complex contexts with greater efficiency. However, they are typically limited to text-only inputs. In this paper, we introduce ViDove, a translation agent system designed for multimodal input. Inspired by the workflow of human translators, ViDove leverages visual and contextual background information to enhance the translation process. Additionally, we integrate a multimodal memory system and long-short term memory modules enriched with domain-specific knowledge, enabling the agent to perform more accurately and adaptively in real-world scenarios. As a result, ViDove achieves significantly higher translation quality in both subtitle generation and general translation tasks, with a 28% improvement in BLEU scores and a 15% improvement in SubER compared to previous state-of-the-art baselines. Moreover, we introduce DoveBench, a new benchmark for long-form automatic video subtitling and translation, featuring 17 hours of high-quality, human-annotated data. Our code is available here: https://github.com/pigeonai-org/ViDove

Yuan Zhang, Weihua Chen, Yichen Lu et al.

Knowledge distillation is an effective paradigm for boosting the performance of pocket-size model, especially when multiple teacher models are available, the student would break the upper limit again. However, it is not economical to train diverse teacher models for the disposable distillation. In this paper, we introduce a new concept dubbed Avatars for distillation, which are the inference ensemble models derived from the teacher. Concretely, (1) For each iteration of distillation training, various Avatars are generated by a perturbation transformation. We validate that Avatars own higher upper limit of working capacity and teaching ability, aiding the student model in learning diverse and receptive knowledge perspectives from the teacher model. (2) During the distillation, we propose an uncertainty-aware factor from the variance of statistical differences between the vanilla teacher and Avatars, to adjust Avatars' contribution on knowledge transfer adaptively. Avatar Knowledge Distillation AKD is fundamentally different from existing methods and refines with the innovative view of unequal training. Comprehensive experiments demonstrate the effectiveness of our Avatars mechanism, which polishes up the state-of-the-art distillation methods for dense prediction without more extra computational cost. The AKD brings at most 0.7 AP gains on COCO 2017 for Object Detection and 1.83 mIoU gains on Cityscapes for Semantic Segmentation, respectively. Code is available at https://github.com/Gumpest/AvatarKD.

Siyi Qian, Xiongfei Bai, Bingtao Fu et al.

In this paper, we propose Agentar-Fin-OCR, a document parsing system tailored to financial-domain documents, transforming ultra-long financial PDFs into semantically consistent, highly accurate, structured outputs with auditing-grade provenance. To address finance-specific challenges such as complex layouts, cross-page structural discontinuities, and cell-level referencing capability, Agentar-Fin-OCR combines (1) a Cross-page Contents Consolidation algorithm to restore continuity across pages and a Document-level Heading Hierarchy Reconstruction (DHR) module to build a globally consistent Table of Contents (TOC) tree for structure-aware retrieval, and (2) a difficulty-adaptive curriculum learning training strategy for table parsing, together with a CellBBoxRegressor module that uses structural anchor tokens to localize table cells from decoder hidden states without external detectors. Experiments demonstrate that our model shows high performance on the table parsing metrics of OmniDocBench. To enable realistic evaluation in the financial vertical, we further introduce FinDocBench, a benchmark that includes six financial document categories with expert-verified annotations and evaluation metrics including Table of Contents edit-distance-based similarity (TocEDS), cross-page concatenated TEDS, and Table Cell Intersection over Union (C-IoU). We evaluate a wide range of state-of-the-art models on FinDocBench to assess their capabilities and remaining limitations on financial documents. Overall, Agentar-Fin-OCR and FinDocBench provide a practical foundation for reliable downstream financial document applications.

Yichen Lu, Siwei Nie, Minlong Lu et al.

Image Copy Detection (ICD) aims to identify manipulated content between image pairs through robust feature representation learning. While self-supervised learning (SSL) has advanced ICD systems, existing view-level contrastive methods struggle with sophisticated edits due to insufficient fine-grained correspondence learning. We address this limitation by exploiting the inherent geometric traceability in edited content through two key innovations. First, we propose PixTrace - a pixel coordinate tracking module that maintains explicit spatial mappings across editing transformations. Second, we introduce CopyNCE, a geometrically-guided contrastive loss that regularizes patch affinity using overlap ratios derived from PixTrace's verified mappings. Our method bridges pixel-level traceability with patch-level similarity learning, suppressing supervision noise in SSL training. Extensive experiments demonstrate not only state-of-the-art performance (88.7% uAP / 83.9% RP90 for matcher, 72.6% uAP / 68.4% RP90 for descriptor on DISC21 dataset) but also better interpretability over existing methods.

Yihan Wu, Yichen Lu, Yifan Peng et al. · nvidia

Audiovisual Automatic Speech Recognition (AV-ASR) aims to improve speech recognition accuracy by leveraging visual signals. It is particularly challenging in unconstrained real-world scenarios across various domains due to noisy acoustic environments, spontaneous speech, and the uncertain use of visual information. Most previous works fine-tune audio-only ASR models on audiovisual datasets, optimizing them for conventional ASR objectives. However, they often neglect visual features and common errors in unconstrained video scenarios. In this paper, we propose using a preference optimization strategy to improve speech recognition accuracy for real-world videos. First, we create preference data via simulating common errors that occurred in AV-ASR from two focals: manipulating the audio or vision input and rewriting the output transcript. Second, we propose BPO-AVASR, a Bifocal Preference Optimization method to improve AV-ASR models by leveraging both input-side and output-side preference. Extensive experiments demonstrate that our approach significantly improves speech recognition accuracy across various domains, outperforming previous state-of-the-art models on real-world video speech recognition.

Yichen Lu, Kai Cheng, Yue Zhang et al.

The freeway on-ramp merging section is often identified as a crash-prone spot due to the high frequency of traffic conflicts. Very few traffic conflict analysis studies comprehensively consider different vehicle types at freeway merging section. Thus, the main objective of this study is to analyse conflicts between different vehicle types at freeway merging section. Field data are collected by Unmanned Aerial Vehicle (UAV) at merging areas in Shanghai, China. Vehicle extraction method is utilized to obtain vehicle trajectories. Time-to-collision (TTC) is utilized as the surrogate safety measure. TTC of car-car conflicts are the smallest while TTC of truck-truck conflicts are the largest. Traffic conflicts frequently occur at on-ramp and acceleration lane. Results show the spatial distribution of lane-change conflicts is significantly different between different vehicle types, suggesting that vehicle drivers should maintain safe distance especially car drivers. Besides, in order to decrease lane-change conflict at merging area, traffic management agencies are suggested to change dotted lie to solid lane at the beginning of acceleration lane.

Yichen Lu, Jane Fridlyand, Tiffany Tang et al.

Finding translational biomarkers stands center stage of the future of personalized medicine in healthcare. We observed notable challenges in identifying robust biomarkers as some with great performance in one scenario often fail to perform well in new trials (e.g. different population, indications). With rapid development in the clinical trial world (e.g. assay, disease definition), new trials very likely differ from legacy ones in many perspectives and in development of biomarkers this heterogeneity should be considered. In response, we recommend considering building in the heterogeneity when evaluating biomarkers. In this paper, we present one evaluation strategy by using leave-one-study-out (LOSO) in place of conventional cross-validation (cv) methods to account for the potential heterogeneity across trials used for building and testing the biomarkers. To demonstrate the performance of K-fold vs LOSO cv in estimating the effect size of biomarkers, we leveraged data from clinical trials and simulation studies. In our assessment, LOSO cv provided a more objective estimate of the future performance. This conclusion remained true across different evaluation metrics and different statistical methods.