Jiajun Sun

Qing Chang, Junran Peng, Lingxie Xie et al.

The paradigm of training models on massive data without label through self-supervised learning (SSL) and finetuning on many downstream tasks has become a trend recently. However, due to the high training costs and the unconsciousness of downstream usages, most self-supervised learning methods lack the capability to correspond to the diversities of downstream scenarios, as there are various data domains, different vision tasks and latency constraints on models. Neural architecture search (NAS) is one universally acknowledged fashion to conquer the issues above, but applying NAS on SSL seems impossible as there is no label or metric provided for judging model selection. In this paper, we present DATA, a simple yet effective NAS approach specialized for SSL that provides Domain-Aware and Task-Aware pre-training. Specifically, we (i) train a supernet which could be deemed as a set of millions of networks covering a wide range of model scales without any label, (ii) propose a flexible searching mechanism compatible with SSL that enables finding networks of different computation costs, for various downstream vision tasks and data domains without explicit metric provided. Instantiated With MoCo v2, our method achieves promising results across a wide range of computation costs on downstream tasks, including image classification, object detection and semantic segmentation. DATA is orthogonal to most existing SSL methods and endows them the ability of customization on downstream needs. Extensive experiments on other SSL methods demonstrate the generalizability of the proposed method. Code is released at https://github.com/GAIA-vision/GAIA-ssl

Changhao Jiang, Jiahao Chen, Zhenghao Xiang et al.

Recent commercial systems such as Suno demonstrate strong capabilities in long-form song generation, while academic research remains largely non-reproducible due to the lack of publicly available training data, hindering fair comparison and progress. To this end, we release a fully open-source system for long-form song generation with fine-grained style conditioning, including a licensed synthetic dataset, training and evaluation pipelines, and Muse, an easy-to-deploy song generation model. The dataset consists of 116k fully licensed synthetic songs with automatically generated lyrics and style descriptions paired with audio synthesized by SunoV5. We train Muse via single-stage supervised finetuning of a Qwen-based language model extended with discrete audio tokens using MuCodec, without task-specific losses, auxiliary objectives, or additional architectural components. Our evaluations find that although Muse is trained with a modest data scale and model size, it achieves competitive performance on phoneme error rate, text--music style similarity, and audio aesthetic quality, while enabling controllable segment-level generation across different musical structures. All data, model weights, and training and evaluation pipelines will be publicly released, paving the way for continued progress in controllable long-form song generation research. The project repository is available at https://github.com/yuhui1038/Muse.

Shuo Li, Jiajun Sun, Zhihao Zhang et al.

Recent advances in fine-tuning multimodal large language models (MLLMs) using reinforcement learning have achieved remarkable progress, particularly with the introduction of various entropy control techniques. However, the role and characteristics of entropy in perception-oriented tasks like visual grounding, as well as effective strategies for controlling it, remain largely unexplored. To address this issue, we focus on the visual grounding task and analyze the role and characteristics of entropy in comparison to reasoning tasks. Building on these findings, we introduce ECVGPO (Entropy Control Visual Grounding Policy Optimization), an interpretable algorithm designed for effective entropy regulation. Through entropy control, the trade-off between exploration and exploitation is better balanced. Experiments show that ECVGPO achieves broad improvements across various benchmarks and models.

Shuo Li, Jiajun Sun, Zhekai Wang et al.

Charts are a fundamental visualization format for structured data analysis. Enabling end-to-end chart editing according to user intent is of great practical value, yet remains challenging due to the need for both fine-grained control and global structural consistency. Most existing approaches adopt pipeline-based designs, where natural language or code serves as an intermediate representation, limiting their ability to faithfully execute complex edits. We introduce ChartE$^{3}$, an End-to-End Chart Editing benchmark that directly evaluates models without relying on intermediate natural language programs or code-level supervision. ChartE$^{3}$ focuses on two complementary editing dimensions: local editing, which involves fine-grained appearance changes such as font or color adjustments, and global editing, which requires holistic, data-centric transformations including data filtering and trend line addition. ChartE$^{3}$ contains over 1,200 high-quality samples constructed via a well-designed data pipeline with human curation. Each sample is provided as a triplet of a chart image, its underlying code, and a multimodal editing instruction, enabling evaluation from both objective and subjective perspectives. Extensive benchmarking of state-of-the-art multimodal large language models reveals substantial performance gaps, particularly on global editing tasks, highlighting critical limitations in current end-to-end chart editing capabilities.

Jiajun Sun, Zhe Gao

This paper addresses the expression (EXPR) recognition challenge in the 10th Affective Behavior Analysis in-the-Wild (ABAW) workshop and competition, which requires frame-level classification of eight facial emotional expressions from unconstrained videos. This task is challenging due to inaccurate face localization, large pose and scale variations, motion blur, temporal instability, and other confounding factors across adjacent frames. We propose a two-stage dual-modal (audio-visual) model to address these difficulties. Stage I focuses on robust visual feature extraction with a pretrained DINOv2-based encoder. Specifically, DINOv2 ViT-L/14 is used as the backbone, a padding-aware augmentation (PadAug) strategy is employed for image padding and data preprocessing from raw videos, and a mixture-of-experts (MoE) training head is introduced to enhance classifier diversity. Stage II addresses modality fusion and temporal consistency. For the visual modality, faces are re-cropped from raw videos at multiple scales, and the extracted visual features are averaged to form a robust frame-level representation. Concurrently, frame-aligned Wav2Vec 2.0 audio features are derived from short audio windows to provide complementary acoustic cues. These dual-modal features are integrated via a lightweight gated fusion module, followed by inference-time temporal smoothing. Experiments on the ABAW dataset demonstrate the effectiveness of the proposed method. The two-stage model achieves a Macro-F1 score of 0.5368 on the official validation set and 0.5122 +/- 0.0277 under 5-fold cross-validation, outperforming the official baselines.

Hui Li, Changhao Jiang, Hongyu Wang et al.

The ability to reason from audio, including speech, paralinguistic cues, environmental sounds, and music, is essential for AI agents to interact effectively in real-world scenarios. Existing benchmarks mainly focus on static or single-scene settings and do not fully capture scenarios where multiple speakers, unfolding events, and heterogeneous audio sources interact. To address these challenges, we introduce MDAR, a benchmark for evaluating models on complex, multi-scene, and dynamically evolving audio reasoning tasks. MDAR comprises 3,000 carefully curated question-answer pairs linked to diverse audio clips, covering five categories of complex reasoning and spanning three question types. We benchmark 26 state-of-the-art audio language models on MDAR and observe that they exhibit limitations in complex reasoning tasks. On single-choice questions, Qwen2.5-Omni (open-source) achieves 76.67% accuracy, whereas GPT-4o Audio (closed-source) reaches 68.47%; however, GPT-4o Audio substantially outperforms Qwen2.5-Omni on the more challenging multiple-choice and open-ended tasks. Across all three question types, no model achieves 80% performance. These findings underscore the unique challenges posed by MDAR and its value as a benchmark for advancing audio reasoning research.Code and benchmark can be found at https://github.com/luckyerr/MDAR.

Jiajun Sun, Zhen Yu, Siyuan Yan et al.

Skin images from real-world clinical practice are often limited, resulting in a shortage of training data for deep-learning models. While many studies have explored skin image synthesis, existing methods often generate low-quality images and lack control over the lesion's location and type. To address these limitations, we present LF-VAR, a model leveraging quantified lesion measurement scores and lesion type labels to guide the clinically relevant and controllable synthesis of skin images. It enables controlled skin synthesis with specific lesion characteristics based on language prompts. We train a multiscale lesion-focused Vector Quantised Variational Auto-Encoder (VQVAE) to encode images into discrete latent representations for structured tokenization. Then, a Visual AutoRegressive (VAR) Transformer trained on tokenized representations facilitates image synthesis. Lesion measurement from the lesion region and types as conditional embeddings are integrated to enhance synthesis fidelity. Our method achieves the best overall FID score (average 0.74) among seven lesion types, improving upon the previous state-of-the-art (SOTA) by 6.3%. The study highlights our controllable skin synthesis model's effectiveness in generating high-fidelity, clinically relevant synthetic skin images. Our framework code is available at https://github.com/echosun1996/LF-VAR.

Haohua Que, Mingkai Liu, Jiayue Xie et al.

High-resolution sensors are critical for robust autonomous perception but impose a severe memory wall on battery-constrained electric vehicles. In these systems, data movement energy often outweighs computation. Traditional image compression is ill-suited as it is semantically blind and optimizes for storage rather than bus switching activity. We propose MotiMem, a hardware-software co-designed interface. Exploiting temporal coherence,MotiMem uses lightweight 2D Motion Propagation to dynamically identify Regions of Interest (RoI). Complementing this, a Hybrid Sparsity-Aware Coding scheme leverages adaptive inversion and truncation to induce bitlevel sparsity. Extensive experiments across nuScenes, Waymo, and KITTI with 16 detection models demonstrate that MotiMem reduces memory-interface dynamic energy by approximately 43 percent while retaining approximately 93 percent of the object detection accuracy, establishing a new Pareto frontier significantly superior to standard codecs like JPEG and WebP.

Jiajun Sun, Yangyi Ou, Haoyuan Zheng et al.

In complex environments, autonomous robot navigation and environmental perception pose higher requirements for SLAM technology. This paper presents a novel method for semantically enhancing 3D point cloud maps with thermal information. By first performing pixel-level fusion of visible and infrared images, the system projects real-time LiDAR point clouds onto this fused image stream. It then segments heat source features in the thermal channel to instantly identify high temperature targets and applies this temperature information as a semantic layer on the final 3D map. This approach generates maps that not only have accurate geometry but also possess a critical semantic understanding of the environment, making it highly valuable for specific applications like rapid disaster assessment and industrial preventive maintenance.

Shuo Li, Jiajun Sun, Guodong Zheng et al.

Recently, multimodal large language models (MLLMs) have demonstrated remarkable performance in visual-language tasks. However, the authenticity of the responses generated by MLLMs is often compromised by object hallucinations. We identify that a key cause of these hallucinations is the model's over-susceptibility to specific image frequency features in detecting objects. In this paper, we introduce Multi-Frequency Perturbations (MFP), a simple, cost-effective, and pluggable method that leverages both low-frequency and high-frequency features of images to perturb visual feature representations and explicitly suppress redundant frequency-domain features during inference, thereby mitigating hallucinations. Experimental results demonstrate that our method significantly mitigates object hallucinations across various model architectures. Furthermore, as a training-time method, MFP can be combined with inference-time methods to achieve state-of-the-art performance on the CHAIR benchmark.

Fengjie Li, Jiajun Jiang, Jiajun Sun et al.

LLM-based automated program repair methods have attracted significant attention for their state-of-the-art performance. However, they were primarily evaluated on a few well known datasets like Defects4J, raising questions about their effectiveness on new datasets. In this study, we evaluate 11 top-performing LLMs on DEFECTS4J-TRANS, a new dataset derived from transforming Defects4J while maintaining the original semantics. Results from experiments on both Defects4J and DEFECTS4J-TRANS show that all studied LLMs have limited generalizability in APR tasks, with the average number of correct and plausible patches decreasing by 49.48% and 42.90%, respectively, on DEFECTS4J-TRANS. Further investigation into incorporating additional repair-relevant information in repair prompts reveals that, although this information significantly enhances the LLMs' capabilities (increasing the number of correct and plausible patches by up to 136.67% and 121.82%, respectively), performance still falls short of their original results. This indicates that prompt engineering alone is insufficient to substantially enhance LLMs' repair capabilities. Based on our study, we also offer several recommendations for future research.

Yifei Cao, Changhao Jiang, Jiabao Zhuang et al.

Assessing the perceptual quality of synthetic speech is crucial for guiding the development and refinement of speech generation models. However, it has traditionally relied on human subjective ratings such as the Mean Opinion Score (MOS), which depend on manual annotations and often suffer from inconsistent rating standards and poor reproducibility. To address these limitations, we introduce MOS-RMBench, a unified benchmark that reformulates diverse MOS datasets into a preference-comparison setting, enabling rigorous evaluation across different datasets. Building on MOS-RMBench, we systematically construct and evaluate three paradigms for reward modeling: scalar reward models, semi-scalar reward models, and generative reward models (GRMs). Our experiments reveal three key findings: (1) scalar models achieve the strongest overall performance, consistently exceeding 74% accuracy; (2) most models perform considerably worse on synthetic speech than on human speech; and (3) all models struggle on pairs with very small MOS differences. To improve performance on these challenging pairs, we propose a MOS-aware GRM that incorporates an MOS-difference-based reward function, enabling the model to adaptively scale rewards according to the difficulty of each sample pair. Experimental results show that the MOS-aware GRM significantly improves fine-grained quality discrimination and narrows the gap with scalar models on the most challenging cases. We hope this work will establish both a benchmark and a methodological framework to foster more rigorous and scalable research in automatic speech quality assessment.

Changhao Jiang, Jiajun Sun, Yifei Cao et al.

Recently, role-playing agents have emerged as a promising paradigm for achieving personalized interaction and emotional resonance. Existing research primarily focuses on the textual modality, neglecting the critical dimension of speech in realistic interactive scenarios. In particular, there is a lack of systematic evaluation for Speech Role-Playing Agents (SRPAs). To address this gap, we construct SpeechRole-Data, a large-scale, high-quality dataset that comprises 98 diverse roles and 112k speech-based single-turn and multi-turn conversations. Each role demonstrates distinct vocal characteristics, including timbre and prosody, thereby enabling more sophisticated speech role-playing. Furthermore, we propose SpeechRole-Eval, a multidimensional evaluation benchmark that systematically assesses SRPAs performance in key aspects such as fundamental interaction ability, speech expressiveness, and role-playing fidelity. Experimental results reveal the advantages and challenges of both cascaded and end-to-end speech role-playing agents in maintaining vocal style consistency and role coherence. We release all data, code, and baseline models to provide a solid foundation for speech-driven multimodal role-playing research and to foster further developments in this field.

Xiaoran Fan, Zhichao Sun, Yangfan Gao et al.

Speech-language models (SLMs) offer a promising path toward unifying speech and text understanding and generation. However, challenges remain in achieving effective cross-modal alignment and high-quality speech generation. In this work, we systematically investigate the role of speech tokenizer designs in LLM-centric SLMs, augmented by speech heads and speaker modeling. We compare coupled, semi-decoupled, and fully decoupled speech tokenizers under a fair SLM framework and find that decoupled tokenization significantly improves alignment and synthesis quality. To address the information density mismatch between speech and text, we introduce multi-token prediction (MTP) into SLMs, enabling each hidden state to decode multiple speech tokens. This leads to up to 12$\times$ faster decoding and a substantial drop in word error rate (from 6.07 to 3.01). Furthermore, we propose a speaker-aware generation paradigm and introduce RoleTriviaQA, a large-scale role-playing knowledge QA benchmark with diverse speaker identities. Experiments demonstrate that our methods enhance both knowledge understanding and speaker consistency.

Changhao Jiang, Ming Zhang, Yifei Cao et al.

The GPT-4 technical report suggests that downstream performance can be predicted from pre-training signals, but offers little methodological detail on how to quantify this. This work address this gap by modeling knowledge retention, the capacity of a pre-trained language model to memorize factual information from its corpus, and introduce a principled method to estimate it prior to training. We propose Size-dependent Mutual Information (SMI), an information-theoretic predictor that integrates knowledge frequency, knowledge specificity, and model size to forecast closed-book question answering (QA) accuracy. SMI is validated through large-scale document retrieval over the disclosed pre-training corpora of 21 public and 3 custom models, combined with a robust multi-template QA evaluation. Experiments show that SMI significantly outperforms repetition-based baselines and achieves $R^2$ > 0.7 in predicting QA accuracy for models above 1B parameters, without additional training. The analysis further reveals diminishing returns from scaling data and model size and provides evidence for an intrinsic upper bound on knowledge retention achievable by pre-training alone, motivating retrieval and other augmentation strategies.

Jiajun Sun

In crowdsourcing markets, there are two different type jobs, i.e. homogeneous jobs and heterogeneous jobs, which need to be allocated to workers. Incentive mechanisms are essential to attract extensive user participating for achieving good service quality, especially under a given budget constraint condition. To this end, recently, Singer et al. propose a novel class of auction mechanisms for determining near-optimal prices of tasks for crowdsourcing markets constrained by the given budget. Their mechanisms are very useful to motivate extensive user to truthfully participate in crowdsourcing markets. Although they are so important, there still exist many security and privacy challenges in real-life environments. In this paper, we present a general privacy-preserving verifiable incentive mechanism for crowdsourcing markets with the budget constraint, not only to exploit how to protect the bids and assignments' privacy, and the chosen winners' privacy in crowdsourcing markets with homogeneous jobs and heterogeneous jobs and identity privacy from users, but also to make the verifiable payment between the platform and users for crowdsourcing applications. Results show that our general privacy-preserving verifiable incentive mechanisms achieve the same results as the generic one without privacy preservation.