arXiv subjects

Li Zhang

Publications and source records attributed to Li Zhang.

At least 289 records · Page 16Linked to original sources

arXiv2025

Unifying Inference-Time Planning Language Generation

A line of work in planning uses LLM not to generate a plan, but to generate a formal representation in some planning language, which can be input into a symbolic solver to deterministically find a plan. While showing improved trust and promising performance, dozens of recent publications have proposed scattered methods on a variety of benchmarks under different experimental settings. We attempt to unify the inference-time LLM-as-formalizer methodology for classical planning by proposing a unifying framework based on intermediate representations. We thus systematically evaluate more than a dozen pipelines that subsume most existing work, while proposing novel ones that involve syntactically similar but high resource intermediate languages (such as a Python wrapper of PDDL). We provide recipes for planning language generation pipelines, draw a series of conclusions showing the efficacy of their various components, and evidence their robustness against problem complexity.

cs.CL↗

arXiv2025

Prototype-Based Dynamic Steering for Large Language Models

Despite impressive breadth, LLMs still rely on explicit reasoning instructions or static, one-fits-all steering methods, leaving a gap for adaptive, instruction-free reasoning amplification. We present Prototype-Based Dynamic Steering (PDS), a test-time method that amplifies large language model (LLM) reasoning without adding or altering instructions. We introduce "reasoning prototypes" by clustering activation differences between Chain-of-Thought (CoT) and neutral prompts. At inference, an input's hidden state is projected onto these prototypes to form an instance-specific steering vector. Evaluated on GSM8K, AQuA-RAT, and BIG-Bench tasks, PDS consistently improves accuracy without fine-tuning or prompt engineering. Notably, the gains persist even when CoT is explicitly suppressed to improve cost-efficiency, indicating that the intervention strengthens latent reasoning processes rather than inducing a superficial behavioral shift. These results position dynamic, prototype-guided steering as a lightweight alternative to training-time approaches for enhancing LLM reasoning.

cs.CL↗

arXiv2025

Break the ID-Language Barrier: An Adaption Framework for LLM-based Sequential Recommendation

The recent breakthrough of large language models (LLMs) in natural language processing has sparked exploration in recommendation systems, however, their limited domain-specific knowledge remains a critical bottleneck. Specifically, LLMs lack key pieces of information crucial for sequential recommendations, such as user behavior patterns. To address this critical gap, we propose IDLE-Adapter, a novel framework that integrates pre-trained ID embeddings, rich in domain-specific knowledge, into LLMs to improve recommendation accuracy. IDLE-Adapter acts as a bridge, transforming sparse user-item interaction data into dense, LLM-compatible representations through a Pre-trained ID Sequential Model, Dimensionality Alignment, Layer-wise Embedding Refinement, and Layer-wise Distribution Alignment. Furthermore, IDLE-Adapter demonstrates remarkable flexibility by seamlessly integrating ID embeddings from diverse ID-based sequential models and LLM architectures. Extensive experiments across various datasets demonstrate the superiority of IDLE-Adapter, achieving over 10\% and 20\% improvements in HitRate@5 and NDCG@5 metrics, respectively, compared to state-of-the-art methods.

cs.IR↗

arXiv2025

FastCoder: Accelerating Repository-level Code Generation via Efficient Retrieval and Verification

Code generation is a latency-sensitive task that demands high timeliness. However, with the growing interest and inherent difficulty in repository-level code generation, most existing code generation studies focus on improving the correctness of generated code while overlooking the inference efficiency, which is substantially affected by the overhead during LLM generation. Although there has been work on accelerating LLM inference, these approaches are not tailored to the specific characteristics of code generation; instead, they treat code the same as natural language sequences and ignore its unique syntax and semantic characteristics, which are also crucial for improving efficiency. Consequently, these approaches exhibit limited effectiveness in code generation tasks, particularly for repository-level scenarios with considerable complexity and difficulty. To alleviate this issue, following draft-verification paradigm, we propose FastCoder, a simple yet highly efficient inference acceleration approach specifically designed for code generation, without compromising the quality of the output. FastCoder constructs a multi-source datastore, providing access to both general and project-specific knowledge, facilitating the retrieval of high-quality draft sequences. Moreover, FastCoder reduces the retrieval cost by controlling retrieval timing, and enhances efficiency through parallel retrieval and a context- and LLM preference-aware cache. Experimental results show that FastCoder can reach up to 2.53x and 2.54x speedup compared to autoregressive decoding in repository-level and standalone code generation tasks, respectively, outperforming state-of-the-art inference acceleration approaches by up to 88%. FastCoder can also be integrated with existing correctness-focused code generation approaches to accelerate the LLM generation process, and reach a speedup exceeding 2.6x.

cs.AI↗

arXiv2025

EfficientEdit: Accelerating Code Editing via Edit-Oriented Speculative Decoding

Large Language Models (LLMs) have demonstrated remarkable capabilities in code editing, substantially enhancing software development productivity. However, the inherent complexity of code editing tasks forces existing approaches to rely on LLMs' autoregressive end-to-end generation, where decoding speed plays a critical role in efficiency. While inference acceleration techniques like speculative decoding are applied to improve the decoding efficiency, these methods fail to account for the unique characteristics of code editing tasks where changes are typically localized and existing code segments are reused. To address this limitation, we propose EfficientEdit, a novel method that improves LLM-based code editing efficiency through two key mechanisms based on speculative decoding: (1) effective reuse of original code segments while identifying potential edit locations, and (2) efficient generate edit content via high-quality drafts from edit-oriented draft models and a dynamic verification mechanism that balances quality and acceleration. Experimental results show that EfficientEdit can achieve up to 10.38$\times$ and 13.09$\times$ speedup compared to standard autoregressive decoding in CanItEdit and CodeIF-Bench, respectively, outperforming state-of-the-art inference acceleration approaches by up to 90.6%. The code and data are available at https://github.com/zhu-zhu-ding/EfficientEdit.

cs.SE↗

arXiv2025

Explainable Fault Localization for Programming Assignments via LLM-Guided Annotation

Providing timely and personalized guidance for students' programming assignments, offers significant practical value for helping students complete assignments and enhance their learning. In recent years, various automated Fault Localization (FL) techniques have demonstrated promising results in identifying errors in programs. However, existing FL techniques face challenges when applied to educational contexts. Most approaches operate at the method level without explanatory feedback, resulting in granularity too coarse for students who need actionable insights to identify and fix their errors. While some approaches attempt line-level fault localization, they often depend on predicting line numbers directly in numerical form, which is ill-suited to LLMs. To address these challenges, we propose FLAME, a fine-grained, explainable Fault Localization method tailored for programming assignments via LLM-guided Annotation and Model Ensemble. FLAME leverages rich contextual information specific to programming assignments to guide LLMs in identifying faulty code lines. Instead of directly predicting line numbers, we prompt the LLM to annotate faulty code lines with detailed explanations, enhancing both localization accuracy and educational value. To further improve reliability, we introduce a weighted multi-model voting strategy that aggregates results from multiple LLMs to determine the suspiciousness of each code line. Extensive experimental results demonstrate that FLAME outperforms state-of-the-art fault localization baselines on programming assignments, successfully localizing 207 more faults at top-1 over the best-performing baseline. Beyond educational contexts, FLAME also generalizes effectively to general-purpose software codebases, outperforming all baselines on the Defects4J benchmark.

cs.SE↗

arXiv2025

Documentation Retrieval Improves Planning Language Generation

Certain strong LLMs have shown promise for zero-shot formal planning by generating planning languages like PDDL. Yet, the performance of most open-source models under 50B parameters has been reported to be close to zero due to the low-resource nature of these languages. We significantly improve their performance via a series of lightweight pipelines that integrates documentation retrieval with modular code generation and error refinement. With models like Llama-4-Maverick, our best pipeline improves plan correctness from 0% to over 80% on the common BlocksWorld domain. However, while syntactic errors are substantially reduced, semantic errors persist in more challenging domains, revealing fundamental limitations in current models' reasoning capabilities.

cs.IR↗

arXiv2025

SVAC: Scaling Is All You Need For Referring Video Object Segmentation

Referring Video Object Segmentation (RVOS) aims to segment target objects in video sequences based on natural language descriptions. While recent advances in Multi-modal Large Language Models (MLLMs) have improved RVOS performance through enhanced text-video understanding, several challenges remain, including insufficient exploitation of MLLMs' prior knowledge, prohibitive computational and memory costs for long-duration videos, and inadequate handling of complex temporal dynamics. In this work, we propose SVAC, a unified model that improves RVOS by scaling up input frames and segmentation tokens to enhance video-language interaction and segmentation precision. To address the resulting computational challenges, SVAC incorporates the Anchor-Based Spatio-Temporal Compression (ASTC) module to compress visual tokens while preserving essential spatio-temporal structure. Moreover, the Clip-Specific Allocation (CSA) strategy is introduced to better handle dynamic object behaviors across video clips. Experimental results demonstrate that SVAC achieves state-of-the-art performance on multiple RVOS benchmarks with competitive efficiency. Our code is available at https://github.com/lizhang1998/SVAC.

cs.CV↗

arXiv2025

TurnaboutLLM: A Deductive Reasoning Benchmark from Detective Games

This paper introduces TurnaboutLLM, a novel framework and dataset for evaluating the deductive reasoning abilities of Large Language Models (LLMs) by leveraging the interactive gameplay of detective games Ace Attorney and Danganronpa. The framework tasks LLMs with identifying contradictions between testimonies and evidences within long narrative contexts, a challenging task due to the large answer space and diverse reasoning types presented by its questions. We evaluate twelve state-of-the-art LLMs on the dataset, hinting at limitations of popular strategies for enhancing deductive reasoning such as extensive thinking and Chain-of-Thought prompting. The results also suggest varying effects of context size, the number of reasoning step and answer space size on model performance. Overall, TurnaboutLLM presents a substantial challenge for LLMs' deductive reasoning abilities in complex, narrative-rich environments.

cs.CL↗

arXiv2025

An Empirical Study on Failures in Automated Issue Solving

Automated issue solving seeks to autonomously identify and repair defective code snippets across an entire codebase. SWE-Bench has emerged as the most widely adopted benchmark for evaluating progress in this area. While LLM-based agentic tools show great promise, they still fail on a substantial portion of tasks. Moreover, current evaluations primarily report aggregate issue-solving rates, which obscure the underlying causes of success and failure, making it challenging to diagnose model weaknesses or guide targeted improvements. To bridge this gap, we first analyze the performance and efficiency of three SOTA tools, spanning both pipeline-based and agentic architectures, in automated issue solving tasks of SWE-Bench-Verified under varying task characteristics. Furthermore, to move from high-level performance metrics to underlying cause analysis, we conducted a systematic manual analysis of 150 failed instances. From this analysis, we developed a comprehensive taxonomy of failure modes comprising 3 primary phases, 9 main categories, and 25 fine-grained subcategories. Then we systematically analyze the distribution of the identified failure modes, the results reveal distinct failure fingerprints between the two architectural paradigms, with the majority of agentic failures stemming from flawed reasoning and cognitive deadlocks. Motivated by these insights, we propose a collaborative Expert-Executor framework. It introduces a supervisory Expert agent tasked with providing strategic oversight and course-correction for a primary Executor agent. This architecture is designed to correct flawed reasoning and break the cognitive deadlocks that frequently lead to failure. Experiments show that our framework solves 22.2% of previously intractable issues for a leading single agent. These findings pave the way for building more robust agents through diagnostic evaluation and collaborative design.

cs.SE↗

arXiv2025

UAR-NVC: A Unified AutoRegressive Framework for Memory-Efficient Neural Video Compression

Implicit Neural Representations (INRs) have demonstrated significant potential in video compression by representing videos as neural networks. However, as the number of frames increases, the memory consumption for training and inference increases substantially, posing challenges in resource-constrained scenarios. Inspired by the success of traditional video compression frameworks, which process video frame by frame and can efficiently compress long videos, we adopt this modeling strategy for INRs to decrease memory consumption, while aiming to unify the frameworks from the perspective of timeline-based autoregressive modeling. In this work, we present a novel understanding of INR models from an autoregressive (AR) perspective and introduce a Unified AutoRegressive Framework for memory-efficient Neural Video Compression (UAR-NVC). UAR-NVC integrates timeline-based and INR-based neural video compression under a unified autoregressive paradigm. It partitions videos into several clips and processes each clip using a different INR model instance, leveraging the advantages of both compression frameworks while allowing seamless adaptation to either in form. To further reduce temporal redundancy between clips, we design two modules to optimize the initialization, training, and compression of these model parameters. UAR-NVC supports adjustable latencies by varying the clip length. Extensive experimental results demonstrate that UAR-NVC, with its flexible video clip setting, can adapt to resource-constrained environments and significantly improve performance compared to different baseline models. The project page: "https://wj-inf.github.io/UAR-NVC-page/".

cs.CV↗

arXiv2025

Balanced Multimodal Learning: An Unidirectional Dynamic Interaction Perspective

Multimodal learning typically utilizes multimodal joint loss to integrate different modalities and enhance model performance. However, this joint learning strategy can induce modality imbalance, where strong modalities overwhelm weaker ones and limit exploitation of individual information from each modality and the inter-modality interaction information. Existing strategies such as dynamic loss weighting, auxiliary objectives and gradient modulation mitigate modality imbalance based on joint loss. These methods remain fundamentally reactive, detecting and correcting imbalance after it arises, while leaving the competitive nature of the joint loss untouched. This limitation drives us to explore a new strategy for multimodal imbalance learning that does not rely on the joint loss, enabling more effective interactions between modalities and better utilization of information from individual modalities and their interactions. In this paper, we introduce Unidirectional Dynamic Interaction (UDI), a novel strategy that abandons the conventional joint loss in favor of a proactive, sequential training scheme. UDI first trains the anchor modality to convergence, then uses its learned representations to guide the other modality via unsupervised loss. Furthermore, the dynamic adjustment of modality interactions allows the model to adapt to the task at hand, ensuring that each modality contributes optimally. By decoupling modality optimization and enabling directed information flow, UDI prevents domination by any single modality and fosters effective cross-modal feature learning. Our experimental results demonstrate that UDI outperforms existing methods in handling modality imbalance, leading to performance improvement in multimodal learning tasks.

cs.LG↗

arXiv2025

Multimodal Foundation Model-Driven User Interest Modeling and Behavior Analysis on Short Video Platforms

With the rapid expansion of user bases on short video platforms, personalized recommendation systems are playing an increasingly critical role in enhancing user experience and optimizing content distribution. Traditional interest modeling methods often rely on unimodal data, such as click logs or text labels, which limits their ability to fully capture user preferences in a complex multimodal content environment. To address this challenge, this paper proposes a multimodal foundation model-based framework for user interest modeling and behavior analysis. By integrating video frames, textual descriptions, and background music into a unified semantic space using cross-modal alignment strategies, the framework constructs fine-grained user interest vectors. Additionally, we introduce a behavior-driven feature embedding mechanism that incorporates viewing, liking, and commenting sequences to model dynamic interest evolution, thereby improving both the timeliness and accuracy of recommendations. In the experimental phase, we conduct extensive evaluations using both public and proprietary short video datasets, comparing our approach against multiple mainstream recommendation algorithms and modeling techniques. Results demonstrate significant improvements in behavior prediction accuracy, interest modeling for cold-start users, and recommendation click-through rates. Moreover, we incorporate interpretability mechanisms using attention weights and feature visualization to reveal the model's decision basis under multimodal inputs and trace interest shifts, thereby enhancing the transparency and controllability of the recommendation system.

cs.IR↗

arXiv2025

Neural Video Compression with In-Loop Contextual Filtering and Out-of-Loop Reconstruction Enhancement

This paper explores the application of enhancement filtering techniques in neural video compression. Specifically, we categorize these techniques into in-loop contextual filtering and out-of-loop reconstruction enhancement based on whether the enhanced representation affects the subsequent coding loop. In-loop contextual filtering refines the temporal context by mitigating error propagation during frame-by-frame encoding. However, its influence on both the current and subsequent frames poses challenges in adaptively applying filtering throughout the sequence. To address this, we introduce an adaptive coding decision strategy that dynamically determines filtering application during encoding. Additionally, out-of-loop reconstruction enhancement is employed to refine the quality of reconstructed frames, providing a simple yet effective improvement in coding efficiency. To the best of our knowledge, this work presents the first systematic study of enhancement filtering in the context of conditional-based neural video compression. Extensive experiments demonstrate a 7.71% reduction in bit rate compared to state-of-the-art neural video codecs, validating the effectiveness of the proposed approach.

eess.IV↗

arXiv2025

TexVerse: A Universe of 3D Objects with High-Resolution Textures

We introduce TexVerse, a large-scale 3D dataset featuring high-resolution textures. While recent advances in large-scale 3D datasets have enhanced high-resolution geometry generation, creating high-resolution textures end-to-end remains underexplored due to the lack of suitable datasets. TexVerse fills this gap with a curated collection of over 858K unique high-resolution 3D models sourced from Sketchfab, including more than 158K models with physically based rendering (PBR) materials. Each model encompasses all of its high-resolution variants, bringing the total to 1.6M 3D instances. TexVerse also includes specialized subsets: TexVerse-Skeleton, with 69K rigged models, and TexVerse-Animation, with 54K animated models, both preserving original skeleton and animation data uploaded by the user. We also provide detailed model annotations describing overall characteristics, structural components, and intricate features. TexVerse offers a high-quality data resource with wide-ranging potential applications in texture synthesis, PBR material development, animation, and various 3D vision and graphics tasks.

cs.CV↗

arXiv2025

GOSU: Retrieval-Augmented Generation with Global-Level Optimized Semantic Unit-Centric Framework

Building upon the standard graph-based Retrieval-Augmented Generation (RAG), the introduction of heterogeneous graphs and hypergraphs aims to enrich retrieval and generation by leveraging the relationships between multiple entities through the concept of semantic units (SUs). But this also raises a key issue: The extraction of high-level SUs limited to local text chunks is prone to ambiguity, complex coupling, and increased retrieval overhead due to the lack of global knowledge or the neglect of fine-grained relationships. To address these issues, we propose GOSU, a semantic unit-centric RAG framework that efficiently performs global disambiguation and utilizes SUs to capture interconnections between different nodes across the global context. In the graph construction phase, GOSU performs global merging on the pre-extracted SUs from local text chunks and guides entity and relationship extraction, reducing the difficulty of coreference resolution while uncovering global semantic objects across text chunks. In the retrieval and generation phase, we introduce hierarchical keyword extraction and semantic unit completion. The former uncovers the fine-grained binary relationships overlooked by the latter, while the latter compensates for the coarse-grained n-ary relationships missing from the former. Evaluation across multiple tasks demonstrates that GOSU outperforms the baseline RAG methods in terms of generation quality.

cs.CL↗

arXiv2025

Matrixed-Spectrum Decomposition Accelerated Linear Boltzmann Transport Equation Solver for Fast Scatter Correction in Multi-Spectral CT

X-ray scatter has been a serious concern in computed tomography (CT), leading to image artifacts and distortion of CT values. The linear Boltzmann transport equation (LBTE) is recognized as a fast and accurate approach for scatter estimation. However, for multi-spectral CT, it is cumbersome to compute multiple scattering components for different spectra separately when applying LBTE-based scatter correction. In this work, we propose a Matrixed-Spectrum Decomposition accelerated LBTE solver (MSD-LBTE) that can be used to compute X-ray scatter distributions from CT acquisitions at two or more different spectra simultaneously, in a unified framework with no sacrifice in accuracy and nearly no increase in computation in theory. First, a matrixed-spectrum solver of LBTE is obtained by introducing an additional label dimension to expand the phase space. Then, we propose a ``spectrum basis'' for LBTE and a principle of selection of basis using the QR decomposition, along with the above solver to construct the MSD-LBTE. Based on MSD-LBTE, a unified scatter correction method can be established for multi-spectral CT. We validate the effectiveness and accuracy of our method by comparing it with the Monte Carlo method, including the computational time. We also evaluate the scatter correction performance using two different phantoms for fast-kV switching based dual-energy CT, and using an elliptical phantom in a numerical simulation for kV-modulation enabled CT scans, validating that our proposed method can significantly reduce the computational cost at multiple spectra and effectively reduce scatter artifact in reconstructed CT images.

physics.med-ph↗

arXiv2025

Periodic Vibration Gaussian: Dynamic Urban Scene Reconstruction and Real-time Rendering

Modeling dynamic, large-scale urban scenes is challenging due to their highly intricate geometric structures and unconstrained dynamics in both space and time. Prior methods often employ high-level architectural priors, separating static and dynamic elements, resulting in suboptimal capture of their synergistic interactions. To address this challenge, we present a unified representation model, called Periodic Vibration Gaussian (PVG). PVG builds upon the efficient 3D Gaussian splatting technique, originally designed for static scene representation, by introducing periodic vibration-based temporal dynamics. This innovation enables PVG to elegantly and uniformly represent the characteristics of various objects and elements in dynamic urban scenes. To enhance temporally coherent and large scene representation learning with sparse training data, we introduce a novel temporal smoothing mechanism and a position-aware adaptive control strategy respectively. Extensive experiments on Waymo Open Dataset and KITTI benchmarks demonstrate that PVG surpasses state-of-the-art alternatives in both reconstruction and novel view synthesis for both dynamic and static scenes. Notably, PVG achieves this without relying on manually labeled object bounding boxes or expensive optical flow estimation. Moreover, PVG exhibits 900-fold acceleration in rendering over the best alternative.

cs.CV↗