arXiv subjects

Liang Zhu

Publications and source records attributed to Liang Zhu.

At least 19 recordsLinked to original sources

arXiv2026

DEFT: Distribution-guided Efficient Fine-Tuning for Human Alignment

Reinforcement Learning from Human Feedback (RLHF), using algorithms like Proximal Policy Optimization (PPO), aligns Large Language Models (LLMs) with human values but is costly and unstable. Alternatives have been proposed to replace PPO or integrate Supervised Fine-Tuning (SFT) and contrastive learning for direct fine-tuning and value alignment. However, these methods still require voluminous data to learn preferences and may weaken the generalization ability of LLMs. To further enhance alignment efficiency and performance while mitigating the loss of generalization ability, this paper introduces Distribution-guided Efficient Fine-Tuning (DEFT), an efficient alignment framework incorporating data filtering and distributional guidance by calculating the differential distribution reward based on the output distribution of language model and the discrepancy distribution of preference data. A small yet high-quality subset is filtered from the raw data using a differential distribution reward, which is then incorporated into existing alignment methods to guide the model's output distribution. Experimental results demonstrate that the methods enhanced by DEFT outperform the original methods in both alignment capability and generalization ability, with significantly reduced training time.

cs.CL↗

arXiv2026

PLOT: Enhancing Preference Learning via Optimal Transport

Preference learning in Large Language Models (LLMs) has advanced significantly, yet existing methods remain limited by modest performance gains, high computational costs, hyperparameter sensitivity, and insufficient modeling of global token-level relationships. We introduce PLOT, which enhances Preference Learning in fine-tuning-based alignment through a token-level loss derived from Optimal Transport. By formulating preference learning as an Optimal Transport Problem, PLOT aligns model outputs with human preferences while preserving the original distribution of LLMs, ensuring stability and robustness. Furthermore, PLOT leverages token embeddings to capture semantic relationships, enabling globally informed optimization. Experiments across two preference categories - Human Values and Logic & Problem Solving - spanning seven subpreferences demonstrate that PLOT consistently improves alignment performance while maintaining fluency and coherence. These results substantiate optimal transport as a principled methodology for preference learning, establishing a theoretically grounded framework that provides new insights for preference learning of LLMs.

cs.CL↗

arXiv2026

From Guessing to Placeholding: A Cost-Theoretic Framework for Uncertainty-Aware Code Completion

While Large Language Models (LLMs) have demonstrated exceptional proficiency in code completion, they typically adhere to a Hard Completion (HC) paradigm, compelling the generation of fully concrete code even amidst insufficient context. Our analysis of 3 million real-world interactions exposes the limitations of this strategy: 61% of the generated suggestions were either edited after acceptance or rejected despite exhibiting over 80% similarity to the user's subsequent code, suggesting that models frequently make erroneous predictions at specific token positions. Motivated by this observation, we propose Adaptive Placeholder Completion (APC), a collaborative framework that extends HC by strategically outputting explicit placeholders at high-entropy positions, allowing users to fill directly via IDE navigation. Theoretically, we formulate code completion as a cost-minimization problem under uncertainty. Premised on the observation that filling placeholders incurs lower cost than correcting errors, we prove the existence of a critical entropy threshold above which APC achieves strictly lower expected cost than HC. We instantiate this framework by constructing training data from filtered real-world edit logs and design a cost-based reward function for reinforcement learning. Extensive evaluations across 1.5B--14B parameter models demonstrate that APC reduces expected editing costs from 19% to 50% while preserving standard HC performance. Our work provides both a theoretical foundation and a practical training framework for uncertainty-aware code completion, demonstrating that adaptive abstention can be learned end-to-end without sacrificing conventional completion quality.

cs.CL↗

arXiv2026

OmniDP: Beyond-FOV Large-Workspace Humanoid Manipulation with Omnidirectional 3D Perception

The deployment of humanoid robots for dexterous manipulation in unstructured environments remains challenging due to perceptual limitations that constrain the effective workspace. In scenarios where physical constraints prevent the robot from repositioning itself, maintaining omnidirectional awareness becomes far more critical than color or semantic information.While recent advances in visuomotor policy learning have improved manipulation capabilities, conventional RGB-D solutions suffer from narrow fields of view (FOV) and self-occlusion, requiring frequent base movements that introduce motion uncertainty and safety risks. Existing approaches to expanding perception, including active vision systems and third-view cameras, introduce mechanical complexity, calibration dependencies, and latency that hinder reliable real-time performance. In this work, We propose OmniDP, an end-to-end LiDAR-driven 3D visuomotor policy that enables robust manipulation in large workspaces. Our method processes panoramic point clouds through a Time-Aware Attention Pooling mechanism, efficiently encoding sparse 3D data while capturing temporal dependencies. This 360{\deg} perception allows the robot to interact with objects across wide areas without frequent repositioning. To support policy learning, we develop a whole-body teleoperation system for efficient data collection on full-body coordination. Extensive experiments in simulation and real-world environments show that OmniDP achieves robust performance in large-workspace and cluttered scenarios, outperforming baselines that rely on egocentric depth cameras.

cs.RO↗

arXiv2025

Fast Revocable Attribute-Based Encryption with Data Integrity for Internet of Things

Efficient and secure revocable attribute-based encryption (RABE) is vital for ensuring flexible and fine-grained access control and data sharing in cloud storage and outsourced data environments within the Internet of Things (IoT). However, current RABE schemes often struggle to achieve an optimal balance between efficiency, security, dynamic scalability, and other important features, which hampers their practical application. To overcome these limitations, we propose a fast RABE scheme with data integrity for IoT that achieves adaptive security with multiple challenge ciphertexts. Our scheme supports the revocation of authorized users and transfers the computationally heavy revocation processes to the cloud, thereby easing the computational burden on IoT devices. Moreover, it consistently guarantees the integrity and correctness of data. We have demonstrated its adaptive security within the defined security model with multiple challenge ciphertexts and optimized its performance. Experimental results indicate that our scheme provides better performance than existing solutions. Under the same access policy, our scheme reduces computational consumption by 7 to 9 times compared to previous schemes.

cs.CR↗

arXiv2025

Act-Adaptive Margin: Dynamically Calibrating Reward Models for Subjective Ambiguity

Currently, most reinforcement learning tasks focus on domains like mathematics and programming, where verification is relatively straightforward. However, in subjective tasks such as role-playing, alignment techniques struggle to make progress, primarily because subjective reward modeling using the Bradley-Terry model faces significant challenges when dealing with ambiguous preferences. To improve reward modeling in subjective tasks, this paper proposes AAM (\textbf{\underline{A}}ct-\textbf{\underline{A}}daptive \textbf{\underline{M}}argin), which enhances reward modeling by dynamically calibrating preference margins using the model's internal parameter knowledge. We design two versions of AAM that efficiently generate contextually-appropriate preference gaps without additional human annotation. This approach fundamentally improves how reward models handle subjective rewards by better integrating generative understanding with preference scoring. To validate AAM's effectiveness in subjective reward modeling, we conduct evaluations on RewardBench, JudgeBench, and challenging role-playing tasks. Results show that AAM significantly improves subjective reward modeling performance, enhancing Bradley-Terry reward models by 2.95\% in general tasks and 4.85\% in subjective role-playing tasks. Furthermore, reward models trained with AAM can help downstream alignment tasks achieve better results. Our test results show that applying rewards generated by AAM-Augmented RM to preference learning techniques (e.g., GRPO) achieves state-of-the-art results on CharacterEval and Charm. Code and dataset are available at https://github.com/calubkk/AAM.

cs.CL↗

arXiv2024

UTSRMorph: A Unified Transformer and Superresolution Network for Unsupervised Medical Image Registration

Complicated image registration is a key issue in medical image analysis, and deep learning-based methods have achieved better results than traditional methods. The methods include ConvNet-based and Transformer-based methods. Although ConvNets can effectively utilize local information to reduce redundancy via small neighborhood convolution, the limited receptive field results in the inability to capture global dependencies. Transformers can establish long-distance dependencies via a self-attention mechanism; however, the intense calculation of the relationships among all tokens leads to high redundancy. We propose a novel unsupervised image registration method named the unified Transformer and superresolution (UTSRMorph) network, which can enhance feature representation learning in the encoder and generate detailed displacement fields in the decoder to overcome these problems. We first propose a fusion attention block to integrate the advantages of ConvNets and Transformers, which inserts a ConvNet-based channel attention module into a multihead self-attention module. The overlapping attention block, a novel cross-attention method, uses overlapping windows to obtain abundant correlations with match information of a pair of images. Then, the blocks are flexibly stacked into a new powerful encoder. The decoder generation process of a high-resolution deformation displacement field from low-resolution features is considered as a superresolution process. Specifically, the superresolution module was employed to replace interpolation upsampling, which can overcome feature degradation. UTSRMorph was compared to state-of-the-art registration methods in the 3D brain MR (OASIS, IXI) and MR-CT datasets. The qualitative and quantitative results indicate that UTSRMorph achieves relatively better performance. The code and datasets are publicly available at https://github.com/Runshi-Zhang/UTSRMorph.

cs.CV↗

arXiv2024

PersonaMath: Boosting Mathematical Reasoning via Persona-Driven Data Augmentation

While closed-source Large Language Models (LLMs) demonstrate strong mathematical problem-solving abilities, open-source models still face challenges with such tasks. To bridge this gap, we propose a data augmentation approach and introduce PersonaMathQA, a dataset derived from MATH and GSM8K, on which we train the PersonaMath models. Our approach consists of two stages: the first stage focuses on learning from Persona Diversification, and the second stage emphasizes learning from Reflection. In the first stage, we regenerate detailed chain-of-thought (CoT) solutions as instructions using a closed-source LLM and introduce a persona-driven data augmentation technique. This technique innovatively classifies personas based on occupations, significantly enhancing the dataset's diversity and quality. In the second stage, we incorporate reflection to fully leverage more challenging and valuable questions. Evaluation of our PersonaMath models on MATH and GSM8K reveals that the PersonaMath-7B model (based on Qwen2.5-7B) achieves an accuracy of 61.2% on MATH and 87.8% on GSM8K, surpassing all baseline methods and achieving state-of-the-art performance. Notably, our dataset contains only 128.9K data points-merely 32.6% of MetaMathQA and 49.5% of MathInstruct-yet our model outperforms these baselines, demonstrating the high quality and diversity of our dataset, which enables more efficient model training. We open-source the PersonaMathQA dataset, PersonaMath models, and our code for public usage.

cs.CL↗

arXiv2024

MADial-Bench: Towards Real-world Evaluation of Memory-Augmented Dialogue Generation

Long-term memory is important for chatbots and dialogue systems (DS) to create consistent and human-like conversations, evidenced by numerous developed memory-augmented DS (MADS). To evaluate the effectiveness of such MADS, existing commonly used evaluation metrics, like retrieval accuracy and perplexity (PPL), mainly focus on query-oriented factualness and language quality assessment. However, these metrics often lack practical value. Moreover, the evaluation dimensions are insufficient for human-like assessment in DS. Regarding memory-recalling paradigms, current evaluation schemes only consider passive memory retrieval while ignoring diverse memory recall with rich triggering factors, e.g., emotions and surroundings, which can be essential in emotional support scenarios. To bridge the gap, we construct a novel Memory-Augmented Dialogue Benchmark (MADail-Bench) covering various memory-recalling paradigms based on cognitive science and psychology theories. The benchmark assesses two tasks separately: memory retrieval and memory recognition with the incorporation of both passive and proactive memory recall data. We introduce new scoring criteria to the evaluation, including memory injection, emotion support (ES) proficiency, and intimacy, to comprehensively assess generated responses. Results from cutting-edge embedding models and large language models on this benchmark indicate the potential for further advancement. Extensive testing further reveals correlations between memory injection, ES proficiency, and intimacy.

cs.CL↗

arXiv2024

DeliLaw: A Chinese Legal Counselling System Based on a Large Language Model

Traditional legal retrieval systems designed to retrieve legal documents, statutes, precedents, and other legal information are unable to give satisfactory answers due to lack of semantic understanding of specific questions. Large Language Models (LLMs) have achieved excellent results in a variety of natural language processing tasks, which inspired us that we train a LLM in the legal domain to help legal retrieval. However, in the Chinese legal domain, due to the complexity of legal questions and the rigour of legal articles, there is no legal large model with satisfactory practical application yet. In this paper, we present DeliLaw, a Chinese legal counselling system based on a large language model. DeliLaw integrates a legal retrieval module and a case retrieval module to overcome the model hallucination. Users can consult professional legal questions, search for legal articles and relevant judgement cases, etc. on the DeliLaw system in a dialogue mode. In addition, DeliLaw supports the use of English for counseling. we provide the address of the system: https://data.delilegal.com/lawQuestion.

cs.CL↗

arXiv2024

CLHA: A Simple yet Effective Contrastive Learning Framework for Human Alignment

Reinforcement learning from human feedback (RLHF) is a crucial technique in aligning large language models (LLMs) with human preferences, ensuring these LLMs behave in beneficial and comprehensible ways to users. However, a longstanding challenge in human alignment techniques based on reinforcement learning lies in their inherent complexity and difficulty in training. To address this challenge, we present a simple yet effective Contrastive Learning Framework for Human Alignment (CLHA) to align LLMs with human preferences directly. CLHA employs a novel rescoring strategy to evaluate the noise within the data by considering its inherent quality and dynamically adjusting the training process. Simultaneously, CLHA utilizes pairwise contrastive loss and adaptive supervised fine-tuning loss to adaptively modify the likelihood of generating responses, ensuring enhanced alignment with human preferences. Using advanced methods, CLHA surpasses other algorithms, showcasing superior performance in terms of reward model scores, automatic evaluations, and human assessments on the widely used ``Helpful and Harmless'' dataset.

cs.AI↗

arXiv2023

What's Behind the Mask: Understanding Masked Graph Modeling for Graph Autoencoders

The last years have witnessed the emergence of a promising self-supervised learning strategy, referred to as masked autoencoding. However, there is a lack of theoretical understanding of how masking matters on graph autoencoders (GAEs). In this work, we present masked graph autoencoder (MaskGAE), a self-supervised learning framework for graph-structured data. Different from standard GAEs, MaskGAE adopts masked graph modeling (MGM) as a principled pretext task - masking a portion of edges and attempting to reconstruct the missing part with partially visible, unmasked graph structure. To understand whether MGM can help GAEs learn better representations, we provide both theoretical and empirical evidence to comprehensively justify the benefits of this pretext task. Theoretically, we establish close connections between GAEs and contrastive learning, showing that MGM significantly improves the self-supervised learning scheme of GAEs. Empirically, we conduct extensive experiments on a variety of graph benchmarks, demonstrating the superiority of MaskGAE over several state-of-the-arts on both link prediction and node classification tasks.

cs.LG↗

arXiv2023

Less Can Be More: Unsupervised Graph Pruning for Large-scale Dynamic Graphs

The prevalence of large-scale graphs poses great challenges in time and storage for training and deploying graph neural networks (GNNs). Several recent works have explored solutions for pruning the large original graph into a small and highly-informative one, such that training and inference on the pruned and large graphs have comparable performance. Although empirically effective, current researches focus on static or non-temporal graphs, which are not directly applicable to dynamic scenarios. In addition, they require labels as ground truth to learn the informative structure, limiting their applicability to new problem domains where labels are hard to obtain. To solve the dilemma, we propose and study the problem of unsupervised graph pruning on dynamic graphs. We approach the problem by our proposed STEP, a self-supervised temporal pruning framework that learns to remove potentially redundant edges from input dynamic graphs. From a technical and industrial viewpoint, our method overcomes the trade-offs between the performance and the time & memory overheads. Our results on three real-world datasets demonstrate the advantages on improving the efficacy, robustness, and efficiency of GNNs on dynamic node classification tasks. Most notably, STEP is able to prune more than 50% of edges on a million-scale industrial graph Alipay (7M nodes, 21M edges) while approximating up to 98% of the original performance. Code is available at https://github.com/EdisonLeeeee/STEP.

cs.LG↗

arXiv2022

Joint 3-D Positioning and Power Allocation for UAV Relay Aided by Geographic Information

In this paper, we study to employ geographic information to address the blockage problem of air-to-ground links between UAV and terrestrial nodes. In particular, a UAV relay is deployed to establish communication links from a ground base station to multiple ground users. To improve communication capacity, we first model the blockage effect caused by buildings according to the three-dimensional (3-D) geographic information. Then, an optimization problem is formulated to maximize the minimum capacity among users by jointly optimizing the 3-D position and power allocation of the UAV relay, under the constraints of link capacity, maximum transmit power, and blockage. To solve this complex non-convex problem, a two-loop optimization framework is developed based on Lagrangian relaxation. The outer-loop aims to obtain proper Lagrangian multipliers to ensure the solution of the Lagrangian problem converge to the tightest upper bound on the original problem. The inner-loop solves the Lagrangian problem by applying the block coordinate descent (BCD) and successive convex approximation (SCA) techniques, where UAV 3-D positioning and power allocation are alternately optimized in each iteration. Simulation results confirm that the proposed solution significantly outperforms two benchmark schemes and achieves a performance close to the upper bound on the UAV relay system.

eess.SP↗

arXiv2022

Dative epitaxy of commensurate monocrystalline covalent-van der Waals moiré supercrystal

Realizing van der Waals (vdW) epitaxy in the 80s represents a breakthrough that circumvents the stringent lattice matching and processing compatibility requirements in conventional covalent heteroepitaxy. However, due to the weak vdW interactions, there is little control over film qualities by the substrate. Typically, discrete domains with a spread of misorientation angles are formed, limiting the applicability of vdW epitaxy. Here we report the epitaxial growth of monocrystalline, covalent Cr5Te8 2D crystals on monolayer vdW WSe2 by chemical vapor deposition, driven by interfacial dative bond formation. The lattice of Cr5Te8, with a lateral dimension of a few ten microns, is fully commensurate with that of WSe2 via 3 x 3 (Cr5Te8)-7 x 7 (WSe2) supercell matching, forming a single crystalline moire superlattice. Our work has established a conceptually distinct paradigm of thin film epitaxy termed dative epitaxy, which takes full advantage of covalent epitaxy with chemical bonding for fixing the atomic registry and crystal orientation, while circumventing its stringent lattice matching and processing compatibility requirements; conversely, it ensures the full flexibility of vdW epitaxy, while avoiding its poor orientation control. Cr5Te8 2D crystals grown by dative epitaxy exhibit square magnetic hysteresis, suggesting minimized interfacial defects that can serve as pinning sites.

cond-mat.mtrl-sci↗

arXiv2020

Towards in-store multi-person tracking using head detection and track heatmaps

Computer vision algorithms are being implemented across a breadth of industries to enable technological innovations. In this paper, we study the problem of computer vision based customer tracking in retail industry. To this end, we introduce a dataset collected from a camera in an office environment where participants mimic various behaviors of customers in a supermarket. In addition, we describe an illustrative example of the use of this dataset for tracking participants based on a head tracking model in an effort to minimize errors due to occlusion. Furthermore, we propose a model for recognizing customers and staff based on their movement patterns. The model is evaluated using a real-world dataset collected in a supermarket over a 24-hour period that achieves 98% accuracy during training and 93% accuracy during evaluation.

cs.CV↗

arXiv2020

An improved sample size calculation method for score tests in generalized linear models

Self and Mauritsen (1988) developed a sample size determination procedure for score tests in generalized linear models under contiguous alternatives. Its performance may deteriorate when the effect size is large. We propose a modification of the Self-Mauritsen method by taking into account of the variance of the score statistic under both the null and alternative hypotheses, and extend the method to noninferiority trials. The modified approach is employed to calculate the sample size for the logistic regression and negative binomial regression in superiority and noninferiority trials. We further explain why the formulae recently derived by Zhu and Lakkis tend to underestimate the required sample size for the negative binomial regression. Numerical examples are used to demonstrate the accuracy of the proposed method.

stat.ME↗

arXiv2020

MOTS: Multiple Object Tracking for General Categories Based On Few-Shot Method

Most modern Multi-Object Tracking (MOT) systems typically apply REID-based paradigm to hold a balance between computational efficiency and performance. In the past few years, numerous attempts have been made to perfect the systems. Although they presented favorable performance, they were constrained to track specified category. Drawing on the ideas of few shot method, we pioneered a new multi-target tracking system, named MOTS, which is based on metrics but not limited to track specific category. It contains two stages in series: In the first stage, we design the self-Adaptive-matching module to perform simple targets matching, which can complete 88.76% assignments without sacrificing performance on MOT16 training set. In the second stage, a Fine-match Network was carefully designed for unmatched targets. With a newly built TRACK-REID data-set, the Fine-match Network can perform matching of 31 category targets, even generalizes to unseen categories.

cs.CV↗