arXiv subjects

Jian Zhao

Publications and source records attributed to Jian Zhao.

At least 109 records · Page 6Linked to original sources

arXiv2025

Bias-Restrained Prefix Representation Finetuning for Mathematical Reasoning

Parameter-Efficient finetuning (PEFT) enhances model performance on downstream tasks by updating a minimal subset of parameters. Representation finetuning (ReFT) methods further improve efficiency by freezing model weights and optimizing internal representations with fewer parameters than PEFT, outperforming PEFT on several tasks. However, ReFT exhibits a significant performance decline on mathematical reasoning tasks. To address this problem, the paper demonstrates that ReFT's poor performance on mathematical tasks primarily stems from its struggle to generate effective reasoning prefixes during the early inference phase. Moreover, ReFT disturbs the numerical encoding and the error accumulats during the CoT stage. Based on these observations, this paper proposes Bias-REstrained Prefix Representation FineTuning (BREP ReFT), which enhances ReFT's mathematical reasoning capability by truncating training data to optimize the generation of initial reasoning prefixes, intervening on the early inference stage to prevent error accumulation, and constraining the intervention vectors' magnitude to avoid disturbing numerical encoding. Extensive experiments across diverse model architectures demonstrate BREP's superior effectiveness, efficiency, and robust generalization capability, outperforming both standard ReFT and weight-based PEFT methods on the task of mathematical reasoning. The source code is available at https://github.com/LiangThree/BREP.

cs.LG↗

arXiv2025

NoteEx: Interactive Visual Context Manipulation for LLM-Assisted Exploratory Data Analysis in Computational Notebooks

Computational notebooks have become popular for Exploratory Data Analysis (EDA), augmented by LLM-based code generation and result interpretation. Effective LLM assistance hinges on selecting informative context -- the minimal set of cells whose code, data, or outputs suffice to answer a prompt. As notebooks grow long and messy, users can lose track of the mental model of their analysis. They thus fail to curate appropriate contexts for LLM tasks, causing frustration and tedious prompt engineering. We conducted a formative study (n=6) that surfaced challenges in LLM context selection and mental model maintenance. Therefore, we introduce NoteEx, a JupyterLab extension that provides a semantic visualization of the EDA workflow, allowing analysts to externalize their mental model, specify analysis dependencies, and enable interactive selection of task-relevant contexts for LLMs. A user study (n=12) against a baseline shows that NoteEx improved mental model retention and context selection, leading to more accurate and relevant LLM responses.

cs.HC↗

arXiv2025

Spatial Insight: How Data-Driven Regions of Interest Selection Enhances Single-Trial P300 Classification in EEG-Based BCIs

EEG-based Brain-Computer Interfaces (BCIs) frequently face spatial specificity limitations in detecting single-trial P300 potentials, a neurophysiological hallmark leveraged for both BCI control and neurodegenerative disease diagnostics. We present a novel framework combining eLORETA source localization with cross-subject functional connectivity to identify stable regions of interest (ROIs) across sessions. Analyzing 62-channel EEG data from 31 subjects (63 sessions, 2,520 trials), we demonstrate that phase-lagged connectivity metrics can reliably isolate task-relevant hubs in deeper cortical-subcortical structures like the insula and parietal regions - critical for Alzheimer's disease biomarkers. By integrating spatially stable ROIs with dynamic temporal agreement, our hybrid classification systematically outperforms whole-brain approaches in different frequency bands (up to 5.4% depending on the connectivity method and the spectral range) while maintaining millisecond-level temporal precision. To the best of our knowledge, this is the first study to establish cross-subject ROI consensus through source-space connectivity, bypassing scalp EEG's depth constraints to probe Alzheimer's-relevant networks. The framework's robustness to noise and compatibility with portable systems offer significant potential for global deployment in early neurodegenerative disease detection. Future integration of individualized anatomical data or adaptive parameter optimization could refine this tool for clinical deployment, enhancing the current max accuracy of 81.57% in the 1-15 Hz range.

q-bio.QM↗

arXiv2025

Is Sora a World Simulator? A Comprehensive Survey on General World Models and Beyond

General world models represent a crucial pathway toward achieving Artificial General Intelligence (AGI), serving as the cornerstone for various applications ranging from virtual environments to decision-making systems. Recently, the emergence of the Sora model has attained significant attention due to its remarkable simulation capabilities, which exhibits an incipient comprehension of physical laws. In this survey, we embark on a comprehensive exploration of the latest advancements in world models. Our analysis navigates through the forefront of generative methodologies in video generation, where world models stand as pivotal constructs facilitating the synthesis of highly realistic visual content. Additionally, we scrutinize the burgeoning field of autonomous-driving world models, meticulously delineating their indispensable role in reshaping transportation and urban mobility. Furthermore, we delve into the intricacies inherent in world models deployed within autonomous agents, shedding light on their profound significance in enabling intelligent interactions within dynamic environmental contexts. At last, we examine challenges and limitations of world models, and discuss their potential future directions. We hope this survey can serve as a foundational reference for the research community and inspire continued innovation. This survey will be regularly updated at: https://github.com/GigaAI-research/General-World-Models-Survey.

cs.CV↗

arXiv2025

Evolving LLMs' Self-Refinement Capability via Synergistic Training-Inference Optimization

Self-Refinement refers to a model's ability to revise its own responses to produce improved outputs. This capability can also serve as a fundamental mechanism for Self-Improvement, for example, by reconstructing datasets with refined results to enhance intrinsic model performance. However, our comprehensive experiments reveal that large language models (LLMs) show no clear evidence of inherent Self-Refinement and may even experience response quality degradation after Self-Refinement. To address this issue, we propose EVOLVE, a simple and effective framework for eliciting and tracking the evolution of Self-Refinement through iterative training. We first explore optimization methods during training to activate the model's Self-Refinement capability. Then, at inference, we investigate various generation strategies to further enhance and utilize Self-Refinement while supplying the necessary data for training. Through synergistic optimization of training and inference stages, we continually evolve the model's Self-Refinement ability, enabling it to better refine its own responses. Moreover, we demonstrate the potential of leveraging Self-Refinement to achieve broader Self-Improvement of intrinsic model abilities. Experiments show that the evolved Self-Refinement ability enables the Llama-3.1-8B base model to surpass GPT-4o, achieving 62.3% length-controlled and 63.3% raw win rates on AlpacaEval 2, and 50.3% on Arena-Hard. It also generalizes effectively to out-of-domain reasoning tasks, improving performance on mathematical reasoning benchmarks such as GSM8K and MATH.

cs.CL↗

arXiv2025

RADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized Collaboration

Existing safety evaluation methods for large language models (LLMs) suffer from inherent limitations, including evaluator bias and detection failures arising from model homogeneity, which collectively undermine the robustness of risk evaluation processes. This paper seeks to re-examine the risk evaluation paradigm by introducing a theoretical framework that reconstructs the underlying risk concept space. Specifically, we decompose the latent risk concept space into three mutually exclusive subspaces: the explicit risk subspace (encompassing direct violations of safety guidelines), the implicit risk subspace (capturing potential malicious content that requires contextual reasoning for identification), and the non-risk subspace. Furthermore, we propose RADAR, a multi-agent collaborative evaluation framework that leverages multi-round debate mechanisms through four specialized complementary roles and employs dynamic update mechanisms to achieve self-evolution of risk concept distributions. This approach enables comprehensive coverage of both explicit and implicit risks while mitigating evaluator bias. To validate the effectiveness of our framework, we construct an evaluation dataset comprising 800 challenging cases. Extensive experiments on our challenging testset and public benchmarks demonstrate that RADAR significantly outperforms baseline evaluation methods across multiple dimensions, including accuracy, stability, and self-evaluation risk sensitivity. Notably, RADAR achieves a 28.87% improvement in risk identification accuracy compared to the strongest baseline evaluation method.

cs.AI↗

arXiv2025

A Parameter-Efficient Mixture-of-Experts Framework for Cross-Modal Geo-Localization

We present a winning solution to RoboSense 2025 Track 4: Cross-Modal Drone Navigation. The task retrieves the most relevant geo-referenced image from a large multi-platform corpus (satellite/drone/ground) given a natural-language query. Two obstacles are severe inter-platform heterogeneity and a domain gap between generic training descriptions and platform-specific test queries. We mitigate these with a domain-aligned preprocessing pipeline and a Mixture-of-Experts (MoE) framework: (i) platform-wise partitioning, satellite augmentation, and removal of orientation words; (ii) an LLM-based caption refinement pipeline to align textual semantics with the distinct visual characteristics of each platform. Using BGE-M3 (text) and EVA-CLIP (image), we train three platform experts using a progressive two-stage, hard-negative mining strategy to enhance discriminative power, and fuse their scores at inference. The system tops the official leaderboard, demonstrating robust cross-modal geo-localization under heterogeneous viewpoints.

cs.CV↗

arXiv2025

Planar or Spatial: Exploring Design Aspects and Challenges for Presentations in Virtual Reality with No-coding Interface

The proliferation of virtual reality (VR) has led to its increasing adoption as an immersive medium for delivering presentations, distinct from other VR experiences like games and 360-degree videos by sharing information in richly interactive environments. However, creating engaging VR presentations remains a challenging and time-consuming task for users, hindering the full realization of VR presentation's capabilities. This research aims to explore the potential of VR presentation, analyze users' opinions, and investigate these via providing a user-friendly no-coding authoring tool. Through an examination of popular presentation software and interviews with seven professionals, we identified five design aspects and four design challenges for VR presentations. Based on the findings, we developed VRStory, a prototype for presentation authoring without coding to explore the design aspects and strategies for addressing the challenges. VRStory offers a variety of predefined and customizable VR elements, as well as modules for layout design, navigation control, and asset generation. A user study was then conducted with 12 participants to investigate their opinions and authoring experience with VRStory. Our results demonstrated that, while acknowledging the advantages of immersive and spatial features in VR, users often have a consistent mental model for traditional 2D presentations and may still prefer planar and static formats in VR for better accessibility and efficient communication. We finally shared our learned design considerations for future development of VR presentation tools, emphasizing the importance of balancing of promoting immersive features and ensuring accessibility.

cs.HC↗

arXiv2025

AsynFusion: Towards Asynchronous Latent Consistency Models for Decoupled Whole-Body Audio-Driven Avatars

Whole-body audio-driven avatar pose and expression generation is a critical task for creating lifelike digital humans and enhancing the capabilities of interactive virtual agents, with wide-ranging applications in virtual reality, digital entertainment, and remote communication. Existing approaches often generate audio-driven facial expressions and gestures independently, which introduces a significant limitation: the lack of seamless coordination between facial and gestural elements, resulting in less natural and cohesive animations. To address this limitation, we propose AsynFusion, a novel framework that leverages diffusion transformers to achieve harmonious expression and gesture synthesis. The proposed method is built upon a dual-branch DiT architecture, which enables the parallel generation of facial expressions and gestures. Within the model, we introduce a Cooperative Synchronization Module to facilitate bidirectional feature interaction between the two modalities, and an Asynchronous LCM Sampling strategy to reduce computational overhead while maintaining high-quality outputs. Extensive experiments demonstrate that AsynFusion achieves state-of-the-art performance in generating real-time, synchronized whole-body animations, consistently outperforming existing methods in both quantitative and qualitative evaluations.

cs.SD↗

arXiv2025

Does Re-referencing Matter? Large Laplacian Filter Optimizes Single-Trial P300 BCI Performance

Electroencephalography (EEG) provides a non-invasive window into brain activity, enabling Brain-Computer Interfaces (BCIs) for communication and control. However, their performance is limited by signal fidelity issues, among which the choice of re-referencing strategy is a pervasive but often overlooked preprocessing bias. Addressing controversies about its necessity and optimal choice, we adopted a quantified approach to evaluate four strategies - no re-referencing, Common Average Reference (CAR), small Laplacian, and large Laplacian - using 62-channels EEG (31 subjects, 2,520 trials). To our knowledge, this is the first study systematically quantifying their impact on single-trial P300 classification accuracy. Our controlled pipeline isolated re-referencing effects for source-space reconstruction (eLORETA with Phase Lag Index) and anatomically constrained classification. The large Laplacian resolves distributed P3b networks while maintaining P3a specificity, achieving the best P300 peak classification accuracy (81.57% hybrid method; 75.97% majority regions of interest). Performance follows a consistent and statistically significant hierarchy: large Laplacian > CAR > no re-reference > small Laplacian, providing a foundation for unified methodological evaluation.

q-bio.NC↗

arXiv2025

The Cost of Simplicity: How Reducing EEG Electrodes Affects Source Localization and BCI Accuracy

Electrode density optimization in electroencephalography (EEG)-based Brain-Computer Interfaces (BCIs) requires balancing practical usability against signal fidelity, particularly for source localization. Reducing electrodes enhances portability but its effects on neural source reconstruction quality and source connectivity - treated as proxies to BCI performance - remain understudied. We address this gap through systematic evaluation of 62-, 32-, and 16-channel configurations using a fixed, fully automated processing pipeline applied to the well-characterized P300 potential. This approach's rationale is to minimize variability and bias inherent to EEG analysis by leveraging the P300's stimulus-locked reproducibility and pipeline standardization. Analyzing 63 sessions (31 subjects) from the Eye-BCI dataset with rigorous artifact correction and channel validation, we demonstrate: (1) Progressive degradation in source reconstruction quality with sparser configurations, including obscured deep neural generators and spatiotemporal distortions; (2) A novel sqrt(Re) scaling law linking electrode reduction ratio (Re) to localization accuracy - a previously unquantified relationship to the best of our knowledge; (3) While reduced configurations preserve basic P300 topography and may suffice for communicative BCIs, higher-density channels are essential for reliable deep source reconstruction. Overall, this study establishes a first step towards quantitative benchmarks for electrode selection, with critical implications for clinical BCIs requiring anatomical precision in applications like neurodegenerative disease monitoring, where compromised spatial resolution could mask pathological signatures. Most importantly, the sqrt(Re) scaling law may provide the first principled method to determine the minimal electrode density required based on acceptable error margins or expected effect sizes.

q-bio.NC↗

arXiv2025

Multiple Queries with Multiple Keys: A Precise Prompt Matching Paradigm for Prompt-based Continual Learning

Continual learning requires machine learning models to continuously acquire new knowledge in dynamic environments while avoiding the forgetting of previous knowledge. Prompt-based continual learning methods effectively address the issue of catastrophic forgetting through prompt expansion and selection. However, existing approaches often suffer from low accuracy in prompt selection, which can result in the model receiving biased knowledge and making biased predictions. To address this issue, we propose the Multiple Queries with Multiple Keys (MQMK) prompt matching paradigm for precise prompt selection. The goal of MQMK is to select the prompts whose training data distribution most closely matches that of the test sample. Specifically, Multiple Queries enable precise breadth search by introducing task-specific knowledge, while Multiple Keys perform deep search by representing the feature distribution of training samples at a fine-grained level. Each query is designed to perform local matching with a designated task to reduce interference across queries. Experiments show that MQMK enhances the prompt matching rate by over 30\% in challenging scenarios and achieves state-of-the-art performance on three widely adopted continual learning benchmarks. The code is available at https://github.com/DunweiTu/MQMK.

cs.CV↗

arXiv2025

OrthAlign: Orthogonal Subspace Decomposition for Non-Interfering Multi-Objective Alignment

Large language model (LLM) alignment faces a critical dilemma when addressing multiple human preferences: improvements in one dimension frequently come at the expense of others, creating unavoidable trade-offs between competing objectives like helpfulness and harmlessness. While prior work mainly focuses on constraint-based optimization algorithms and data selection strategies to mitigate conflicts, these approaches overlook the fundamental issue of resolving conflicts directly at the parameter level. In this paper, we present OrthAlign, an innovative approach that pioneers a new paradigm by leveraging orthogonal subspace decomposition to fundamentally resolve gradient-level conflicts in multi-objective preference alignment. OrthAlign strategically decomposes parameter update spaces into orthogonal subspaces, ensuring that optimization toward different preferences occurs in mathematically non-interfering directions. Building upon this, we provide theoretical guarantees demonstrating that when parameter increments satisfy both orthogonal subspace constraints and spectral norm bounds, the resulting updates exhibit linear Lipschitz growth rather than exponential instability, ensuring stable convergence across all preference dimensions. Extensive experiments show that: I. OrthAlign achieves maximum single-preference improvements ranging from 34.61% to 50.89% after multiple-objective alignment across helpful, harmless, and truthful dimensions. II. With an average overall reward improvement of 13.96%.

cs.LG↗

arXiv2025

Attention as a Compass: Efficient Exploration for Process-Supervised RL in Reasoning Models

Reinforcement Learning (RL) has shown remarkable success in enhancing the reasoning capabilities of Large Language Models (LLMs). Process-Supervised RL (PSRL) has emerged as a more effective paradigm compared to outcome-based RL. However, existing PSRL approaches suffer from limited exploration efficiency, both in terms of branching positions and sampling. In this paper, we introduce a novel PSRL framework (AttnRL), which enables efficient exploration for reasoning models. Motivated by preliminary observations that steps exhibiting high attention scores correlate with reasoning behaviors, we propose to branch from positions with high values. Furthermore, we develop an adaptive sampling strategy that accounts for problem difficulty and historical batch size, ensuring that the whole training batch maintains non-zero advantage values. To further improve sampling efficiency, we design a one-step off-policy training pipeline for PSRL. Extensive experiments on multiple challenging mathematical reasoning benchmarks demonstrate that our method consistently outperforms prior approaches in terms of performance and sampling and training efficiency.

cs.LG↗

arXiv2025

From Noise to Insight: Visualizing Neural Dynamics with Segmented SNR Topographies for Improved EEG-BCI Performance

Electroencephalography (EEG)-based wearable brain-computer interfaces (BCIs) face challenges due to low signal-to-noise ratio (SNR) and non-stationary neural activity. We introduce in this manuscript a mathematically rigorous framework that combines data-driven noise interval evaluation with advanced SNR visualization to address these limitations. Analysis of the publicly available Eye-BCI multimodal dataset demonstrates the method's ability to recover canonical P300 characteristics across frequency bands (delta: 0.5-4 Hz, theta: 4-7.5 Hz, broadband: 1-15 Hz), with precise spatiotemporal localization of both P3a (frontocentral) and P3b (parietal) subcomponents. To the best of our knowledge, this is the first study to systematically assess the impact of noise interval selection on EEG signal quality. Cross-session correlations for four different choices of noise intervals spanning from early to late pre-stimulus phases also indicate that alertness and task engagement states modulate noise interval sensitivity, suggesting broader applications for adaptive BCI systems. While validated in healthy participants, our results represent a first step towards providing clinicians with an interpretable tool for detecting neurophysiological abnormalities and provides quantifiable metrics for system optimization.

q-bio.NC↗

arXiv2025

MoA-Off: Adaptive Heterogeneous Modality-Aware Offloading with Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

Multimodal large language models (MLLMs) enable powerful cross-modal inference but impose significant computational and latency burdens, posing severe challenges for deployment in resource-constrained environments. In this paper, we propose MoA-Off, an adaptive heterogeneous modality-aware offloading framework with edge-cloud collaboration for efficient MLLM inference. MoA-Off introduces a lightweight heterogeneous modality-aware module that estimates the complexity of heterogeneous inputs through multi-dimensional feature analysis. Then, an adaptive edge-cloud collaborative offloading strategy is proposed that dynamically schedules workloads between edge and cloud based on modality-aware complexity scores and real-time system states. The experimental results demonstrate that MoA-Off can achieve over 30% reduction in latency and 30%-65% decrease in resource overhead while maintaining competitive accuracy compared to traditional approaches.

cs.DC↗

arXiv2025

IPF-RDA: An Information-Preserving Framework for Robust Data Augmentation

Data augmentation is widely utilized as an effective technique to enhance the generalization performance of deep models. However, data augmentation may inevitably introduce distribution shifts and noises, which significantly constrain the potential and deteriorate the performance of deep networks. To this end, we propose a novel information-preserving framework, namely IPF-RDA, to enhance the robustness of data augmentations in this paper. IPF-RDA combines the proposal of (i) a new class-discriminative information estimation algorithm that identifies the points most vulnerable to data augmentation operations and corresponding importance scores; And (ii) a new information-preserving scheme that preserves the critical information in the augmented samples and ensures the diversity of augmented data adaptively. We divide data augmentation methods into three categories according to the operation types and integrate these approaches into our framework accordingly. After being integrated into our framework, the robustness of data augmentation methods can be enhanced and their full potential can be unleashed. Extensive experiments demonstrate that although being simple, IPF-RDA consistently improves the performance of numerous commonly used state-of-the-art data augmentation methods with popular deep models on a variety of datasets, including CIFAR-10, CIFAR-100, Tiny-ImageNet, CUHK03, Market1501, Oxford Flower, and MNIST, where its performance and scalability are stressed. The implementation is available at https://github.com/Jackbrocp/IPF-RDA.

cs.CV↗

arXiv2025

When Safe Unimodal Inputs Collide: Optimizing Reasoning Chains for Cross-Modal Safety in Multimodal Large Language Models

Multimodal Large Language Models (MLLMs) are susceptible to the implicit reasoning risk, wherein innocuous unimodal inputs synergistically assemble into risky multimodal data that produce harmful outputs. We attribute this vulnerability to the difficulty of MLLMs maintaining safety alignment through long-chain reasoning. To address this issue, we introduce Safe-Semantics-but-Unsafe-Interpretation (SSUI), the first dataset featuring interpretable reasoning paths tailored for such a cross-modal challenge. A novel training framework, Safety-aware Reasoning Path Optimization (SRPO), is also designed based on the SSUI dataset to align the MLLM's internal reasoning process with human safety values. Experimental results show that our SRPO-trained models achieve state-of-the-art results on key safety benchmarks, including the proposed Reasoning Path Benchmark (RSBench), significantly outperforming both open-source and top-tier commercial MLLMs.

cs.AI↗