Peiwen Sun

Zhen Ye, Peiwen Sun, Jiahe Lei et al.

Recent advancements in audio generation have been significantly propelled by the capabilities of Large Language Models (LLMs). The existing research on audio LLM has primarily focused on enhancing the architecture and scale of audio language models, as well as leveraging larger datasets, and generally, acoustic codecs, such as EnCodec, are used for audio tokenization. However, these codecs were originally designed for audio compression, which may lead to suboptimal performance in the context of audio LLM. Our research aims to address the shortcomings of current audio LLM codecs, particularly their challenges in maintaining semantic integrity in generated audio. For instance, existing methods like VALL-E, which condition acoustic token generation on text transcriptions, often suffer from content inaccuracies and elevated word error rates (WER) due to semantic misinterpretations of acoustic tokens, resulting in word skipping and errors. To overcome these issues, we propose a straightforward yet effective approach called X-Codec. X-Codec incorporates semantic features from a pre-trained semantic encoder before the Residual Vector Quantization (RVQ) stage and introduces a semantic reconstruction loss after RVQ. By enhancing the semantic ability of the codec, X-Codec significantly reduces WER in speech synthesis tasks and extends these benefits to non-speech applications, including music and sound generation. Our experiments in text-to-speech, music continuation, and text-to-sound tasks demonstrate that integrating semantic information substantially improves the overall performance of language models in audio generation. Our code and demo are available (Demo: https://x-codec-audio.github.io Code: https://github.com/zhenye234/xcodec)

Yaoting Wang, Peiwen Sun, Yuanchao Li et al.

The Audio-Visual Segmentation (AVS) task aims to segment sounding objects in the visual space using audio cues. However, in this work, it is recognized that previous AVS methods show a heavy reliance on detrimental segmentation preferences related to audible objects, rather than precise audio guidance. We argue that the primary reason is that audio lacks robust semantics compared to vision, especially in multi-source sounding scenes, resulting in weak audio guidance over the visual space. Motivated by the the fact that text modality is well explored and contains rich abstract semantics, we propose leveraging text cues from the visual scene to enhance audio guidance with the semantics inherent in text. Our approach begins by obtaining scene descriptions through an off-the-shelf image captioner and prompting a frozen large language model to deduce potential sounding objects as text cues. Subsequently, we introduce a novel semantics-driven audio modeling module with a dynamic mask to integrate audio features with text cues, leading to representative sounding object features. These features not only encompass audio cues but also possess vivid semantics, providing clearer guidance in the visual space. Experimental results on AVS benchmarks validate that our method exhibits enhanced sensitivity to audio when aided by text cues, achieving highly competitive performance on all three subsets. Project page: \href{https://github.com/GeWu-Lab/Sounding-Object-Segmentation-Preference}{https://github.com/GeWu-Lab/Sounding-Object-Segmentation-Preference}

Peiwen Sun, Shanshan Zhang, Zishan Liu et al.

It has already been observed that audio-visual embedding is more robust than uni-modality embedding for person verification. Here, we proposed a novel audio-visual strategy that considers aggregators from a fusion perspective. First, we introduced weight-enhanced attentive statistics pooling for the first time in face verification. We find that a strong correlation exists between modalities during pooling, so joint attentive pooling is proposed which contains cycle consistency to learn the implicit inter-frame weight. Finally, each modality is fused with a gated attention mechanism to gain robust audio-visual embedding. All the proposed models are trained on the VoxCeleb2 dev dataset and the best system obtains 0.18%, 0.27%, and 0.49% EER on three official trial lists of VoxCeleb1 respectively, which is to our knowledge the best-published results for person verification.

Yaoting Wang, Peiwen Sun, Dongzhan Zhou et al.

Traditional reference segmentation tasks have predominantly focused on silent visual scenes, neglecting the integral role of multimodal perception and interaction in human experiences. In this work, we introduce a novel task called Reference Audio-Visual Segmentation (Ref-AVS), which seeks to segment objects within the visual domain based on expressions containing multimodal cues. Such expressions are articulated in natural language forms but are enriched with multimodal cues, including audio and visual descriptions. To facilitate this research, we construct the first Ref-AVS benchmark, which provides pixel-level annotations for objects described in corresponding multimodal-cue expressions. To tackle the Ref-AVS task, we propose a new method that adequately utilizes multimodal cues to offer precise segmentation guidance. Finally, we conduct quantitative and qualitative experiments on three test subsets to compare our approach with existing methods from related tasks. The results demonstrate the effectiveness of our method, highlighting its capability to precisely segment objects using multimodal-cue expressions. Dataset is available at \href{https://gewu-lab.github.io/Ref-AVS}{https://gewu-lab.github.io/Ref-AVS}.

Kaituo Feng, Manyuan Zhang, Hongyu Li et al.

Reinforcement learning (RL) has recently achieved remarkable success in eliciting visual reasoning within Multimodal Large Language Models (MLLMs). However, existing approaches typically train separate models for different tasks and treat image and video reasoning as disjoint domains. This results in limited scalability toward a multimodal reasoning generalist, which restricts practical versatility and hinders potential knowledge sharing across tasks and modalities. To this end, we propose OneThinker, an all-in-one reasoning model that unifies image and video understanding across diverse fundamental visual tasks, including question answering, captioning, spatial and temporal grounding, tracking, and segmentation. To achieve this, we construct the OneThinker-600k training corpus covering all these tasks and employ commercial models for CoT annotation, resulting in OneThinker-SFT-340k for SFT cold start. Furthermore, we propose EMA-GRPO to handle reward heterogeneity in multi-task RL by tracking task-wise moving averages of reward standard deviations for balanced optimization. Extensive experiments on diverse visual benchmarks show that OneThinker delivers strong performance on 31 benchmarks, across 10 fundamental visual understanding tasks. Moreover, it exhibits effective knowledge transfer between certain tasks and preliminary zero-shot generalization ability, marking a step toward a unified multimodal reasoning generalist. All code, model, and data are released.

Zhen Ye, Zeqian Ju, Haohe Liu et al.

Recent progress in large-scale zero-shot speech synthesis has been significantly advanced by language models and diffusion models. However, the generation process of both methods is slow and computationally intensive. Efficient speech synthesis using a lower computing budget to achieve quality on par with previous work remains a significant challenge. In this paper, we present FlashSpeech, a large-scale zero-shot speech synthesis system with approximately 5\% of the inference time compared with previous work. FlashSpeech is built on the latent consistency model and applies a novel adversarial consistency training approach that can train from scratch without the need for a pre-trained diffusion model as the teacher. Furthermore, a new prosody generator module enhances the diversity of prosody, making the rhythm of the speech sound more natural. The generation processes of FlashSpeech can be achieved efficiently with one or two sampling steps while maintaining high audio quality and high similarity to the audio prompt for zero-shot speech generation. Our experimental results demonstrate the superior performance of FlashSpeech. Notably, FlashSpeech can be about 20 times faster than other zero-shot speech synthesis systems while maintaining comparable performance in terms of voice quality and similarity. Furthermore, FlashSpeech demonstrates its versatility by efficiently performing tasks like voice conversion, speech editing, and diverse speech sampling. Audio samples can be found in https://flashspeech.github.io/.

Peiwen Sun, Sitong Cheng, Xiangtai Li et al.

Recently, diffusion models have achieved great success in mono-channel audio generation. However, when it comes to stereo audio generation, the soundscapes often have a complex scene of multiple objects and directions. Controlling stereo audio with spatial contexts remains challenging due to high data costs and unstable generative models. To the best of our knowledge, this work represents the first attempt to address these issues. We first construct a large-scale, simulation-based, and GPT-assisted dataset, BEWO-1M, with abundant soundscapes and descriptions even including moving and multiple sources. Beyond text modality, we have also acquired a set of images and rationally paired stereo audios through retrieval to advance multimodal generation. Existing audio generation models tend to generate rather random and indistinct spatial audio. To provide accurate guidance for Latent Diffusion Models, we introduce the SpatialSonic model utilizing spatial-aware encoders and azimuth state matrices to reveal reasonable spatial guidance. By leveraging spatial guidance, our model not only achieves the objective of generating immersive and controllable spatial audio from text but also extends to other modalities as the pioneer attempt. Finally, under fair settings, we conduct subjective and objective evaluations on simulated and real-world data to compare our approach with prevailing methods. The results demonstrate the effectiveness of our method, highlighting its capability to generate spatial audio that adheres to physical rules.

Peiwen Sun, Shiqiang Lang, Dongming Wu et al.

With the current surge in spatial reasoning explorations, researchers have made significant progress in understanding indoor scenes, but still struggle with diverse applications such as robotics and autonomous driving. This paper aims to advance all-scale spatial reasoning across diverse scenarios by tackling two key challenges: 1) the heavy reliance on indoor 3D scans and labor-intensive manual annotations for dataset curation; 2) the absence of effective all-scale scene modeling, which often leads to overfitting to individual scenes. In this paper, we introduce a holistic solution that integrates a structured spatial reasoning knowledge system, scale-aware modeling, and a progressive training paradigm, as the first attempt to broaden the all-scale spatial intelligence of MLLMs to the best of our knowledge. Using a task-specific, specialist-driven automated pipeline, we curate over 38K video scenes across 5 spatial scales to create SpaceVista-1M, a dataset comprising approximately 1M spatial QA pairs spanning 19 diverse task types. While specialist models can inject useful domain knowledge, they are not reliable for evaluation. We then build an all-scale benchmark with precise annotations by manually recording, retrieving, and assembling video-based data. However, naive training with SpaceVista-1M often yields suboptimal results due to the potential knowledge conflict. Accordingly, we introduce SpaceVista-7B, a spatial reasoning model that accepts dense inputs beyond semantics and uses scale as an anchor for scale-aware experts and progressive rewards. Finally, extensive evaluations across 5 benchmarks, including our SpaceVista-Bench, demonstrate competitive performance, showcasing strong generalization across all scales and scenarios. Our dataset, model, and benchmark will be released on https://peiwensun2000.github.io/mm2km .

Nishant Kumar, Ziyan Tao, Jaikirat Singh et al.

Image fusion typically employs non-invertible neural networks to merge multiple source images into a single fused image. However, for clinical experts, solely relying on fused images may be insufficient for making diagnostic decisions, as the fusion mechanism blends features from source images, thereby making it difficult to interpret the underlying tumor pathology. We introduce FusionINN, a novel decomposable image fusion framework, capable of efficiently generating fused images and also decomposing them back to the source images. FusionINN is designed to be bijective by including a latent image alongside the fused image, while ensuring minimal transfer of information from the source images to the latent representation. To the best of our knowledge, we are the first to investigate the decomposability of fused images, which is particularly crucial for life-sensitive applications such as medical image fusion compared to other tasks like multi-focus or multi-exposure image fusion. Our extensive experimentation validates FusionINN over existing discriminative and generative fusion methods, both subjectively and objectively. Moreover, compared to a recent denoising diffusion-based fusion model, our approach offers faster and qualitatively better fusion results.

Peiwen Sun, Yifan Zhang, Zishan Liu et al.

The vanilla fusion methods still dominate a large percentage of mainstream audio-visual tasks. However, the effectiveness of vanilla fusion from a theoretical perspective is still worth discussing. Thus, this paper reconsiders the signal fused in the multimodal case from a bionics perspective and proposes a simple, plug-and-play, attention module for vanilla fusion based on fundamental signal theory and uncertainty theory. In addition, previous work on multimodal dynamic gradient modulation still relies on decoupling the modalities. So, a decoupling-free gradient modulation scheme has been designed in conjunction with the aforementioned attention module, which has various advantages over the decoupled one. Experiment results show that just a few lines of code can achieve up to 2.0% performance improvements to several multimodal classification methods. Finally, quantitative evaluation of other fusion tasks reveals the potential for additional application scenarios.