arXiv subjects

Shikun Zhang

Publications and source records attributed to Shikun Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2025

TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them

The adoption of Large Language Models (LLMs) as automated evaluators (LLM-as-a-judge) has revealed critical inconsistencies in current evaluation frameworks. We identify two fundamental types of inconsistencies: (1) Score-Comparison Inconsistency, where lower-rated responses outperform higher-scored ones in pairwise comparisons, and (2) Pairwise Transitivity Inconsistency, manifested through circular preference chains (A>B>C>A) and equivalence contradictions (A=B=C\neq A). We argue that these issues come from information loss in discrete rating systems and ambiguous tie judgments during pairwise evaluation. We propose TrustJudge, a probabilistic framework that addresses these limitations through two key innovations: 1) distribution-sensitive scoring that computes continuous expectations from discrete rating probabilities, preserving information entropy for more precise scoring, and 2) likelihood-aware aggregation that resolves transitivity violations using bidirectional preference probabilities or perplexity. We also formalize the theoretical limitations of current LLM-as-a-judge frameworks and demonstrate how TrustJudge's components overcome them. When evaluated with Llama-3.1-70B-Instruct as judge using our dataset, TrustJudge reduces Score-Comparison inconsistency by 8.43% (from 23.32% to 14.89%) and Pairwise Transitivity inconsistency by 10.82% (from 15.22% to 4.40%), while maintaining higher evaluation accuracy. Our work provides the first systematic analysis of evaluation framework inconsistencies in LLM-as-a-judge paradigms, offering both theoretical insights and practical solutions for reliable automated assessment. The framework demonstrates consistent improvements across various model architectures and scales, enabling more trustworthy LLM evaluation without requiring additional training or human annotations. The codes can be found at https://github.com/TrustJudge/TrustJudge.

cs.AI↗

arXiv2025

Click Without Compromise: Online Advertising Measurement via Per User Differential Privacy

Online advertising is a cornerstone of the Internet ecosystem, with advertising measurement playing a crucial role in optimizing efficiency. Ad measurement entails attributing desired behaviors, such as purchases, to ad exposures across various platforms, necessitating the collection of user activities across these platforms. As this practice faces increasing restrictions due to rising privacy concerns, safeguarding user privacy in this context is imperative. Our work is the first to formulate the real-world challenge of advertising measurement systems with real-time reporting of streaming data in advertising campaigns. We introduce AdsBPC, a novel user-level differential privacy protection scheme for online advertising measurement results. This approach optimizes global noise power and results in a non-identically distributed noise distribution that preserves differential privacy while enhancing measurement accuracy. Through experiments on both real-world advertising campaigns and synthetic datasets, AdsBPC achieves a 33% to 95% increase in accuracy over existing streaming DP mechanisms applied to advertising measurement. This highlights our method's effectiveness in achieving superior accuracy alongside a formal privacy guarantee, thereby advancing the state-of-the-art in privacy-preserving advertising measurement.

cs.CR↗

arXiv2025

EAvatar: Expression-Aware Head Avatar Reconstruction with Generative Geometry Priors

High-fidelity head avatar reconstruction plays a crucial role in AR/VR, gaming, and multimedia content creation. Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated effectiveness in modeling complex geometry with real-time rendering capability and are now widely used in high-fidelity head avatar reconstruction tasks. However, existing 3DGS-based methods still face significant challenges in capturing fine-grained facial expressions and preserving local texture continuity, especially in highly deformable regions. To mitigate these limitations, we propose a novel 3DGS-based framework termed EAvatar for head reconstruction that is both expression-aware and deformation-aware. Our method introduces a sparse expression control mechanism, where a small number of key Gaussians are used to influence the deformation of their neighboring Gaussians, enabling accurate modeling of local deformations and fine-scale texture transitions. Furthermore, we leverage high-quality 3D priors from pretrained generative models to provide a more reliable facial geometry, offering structural guidance that improves convergence stability and shape accuracy during training. Experimental results demonstrate that our method produces more accurate and visually coherent head reconstructions with improved expression controllability and detail fidelity.

cs.CV↗

arXiv2025

Temporal Self-Rewarding Language Models: Decoupling Chosen-Rejected via Past-Future

Self-Rewarding Language Models propose an architecture in which the Large Language Models(LLMs) both generates responses and evaluates its own outputs via LLM-as-a-Judge prompting, dynamically improving its generative capabilities through iterative Direct Preference Optimization (DPO). However, our analysis reveals a critical limitation in existing Self-Rewarding paradigms: the synchronized improvement of chosen and rejected responses progressively narrows the representational difference between contrasting samples, undermining effective preference learning. We propose \textbf{Temporal Self-Rewarding Language Models} that strategically coordinate past, present, and future model generations to sustain learning signals. Our dual-phase framework introduces: (1) \textit{Anchored Rejection} - fixing rejected responses using the past initial model's outputs and (2) \textit{Future-Guided Chosen} - dynamically curating chosen samples using next-generation model predictions. Extensive experiments across three model families (Llama, Qwen, Mistral) and different model sizes (Llama3B/8B/70B) demonstrate significant improvements when trained with our method compared to Self-Rewarding using same computation resources. For example, Llama3.1-8B reaches a 29.44 win rate on AlpacaEval 2.0 with our method, outperforming the Self-Rewarding baseline (19.69) by 9.75. Notably, our method also demonstrates superior out-of-distribution generalization across mathematical reasoning (GSM8K), knowledge-based QA (ARC, TruthfulQA), and code generation (HumanEval) tasks, even though we do not specifically collect such training data.

cs.CL↗

arXiv2025

Seed&Steer: Guiding Large Language Models with Compilable Prefix and Branch Signals for Unit Test Generation

Unit tests play a vital role in the software development lifecycle. Recent advances in Large Language Model (LLM)-based approaches have significantly improved automated test generation, garnering attention from both academia and industry. We revisit LLM-based unit test generation from a novel perspective by decoupling prefix generation and assertion generation. To characterize their respective challenges, we define Initialization Complexity and adopt Cyclomatic Complexity to measure the difficulty of prefix and assertion generation, revealing that the former primarily affects compilation success, while the latter influences test coverage. To address these challenges, we propose Seed&Steer, a two-step approach that combines traditional unit testing techniques with the capabilities of large language models. Seed&Steer leverages conventional unit testing tools (e.g., EvoSuite) to generate method invocations with high compilation success rates, which serve as seeds to guide LLMs in constructing effective test contexts. It then introduces branching cues to help LLMs explore diverse execution paths (e.g., normal, boundary, and exception cases) and generate assertions with high coverage. We evaluate Seed&Steer on five real-world Java projects against state-of-the-art baselines. Results show that Seed&Steer improves the compilation pass rate by approximately 7%, successfully compiling 792 and 887 previously failing cases on two LLMs. It also achieves up to ~73% branch and line coverage across focal methods of varying complexity, with coverage improvements ranging from 1.09* to 1.26*. Our code, dataset, and experimental scripts will be publicly released to support future research and reproducibility.

cs.SE↗

arXiv2025

RewardAnything: Generalizable Principle-Following Reward Models

Reward Models, essential for guiding Large Language Model optimization, are typically trained on fixed preference datasets, resulting in rigid alignment to single, implicit preference distributions. This prevents adaptation to diverse real-world needs-from conciseness in one task to detailed explanations in another. The standard practice of collecting task-specific preference data and retraining reward models is resource-intensive, often producing biased rewards, and limits practical application. We introduce generalizable, principle-following reward models. We propose that RMs should understand and adhere to dynamically provided natural language specifications of reward principles, similar to instruction-following in LLMs. To measure this capability, we develop RABench, a comprehensive benchmark for RMs focusing on generalization across diverse principles. Evaluations on RABench reveal poor generalization of current RMs. As a solution, we present RewardAnything, a novel RM designed and trained to explicitly follow natural language principles. We achieve SotA performance with RewardAnything in traditional RM benchmark simply by specifying a well-defined principle, and results on RABench show we excel in adapting to novel principles without retraining. Furthermore, RewardAnything integrates seamlessly with existing RLHF methods and we show by a case study on how to automatically and efficiently align LLMs with only natural language principles.

cs.CL↗

arXiv2025

PrivacyGo: Privacy-Preserving Ad Measurement with Multidimensional Intersection

This paper tackles the challenging and practical problem of multi-identifier private user profile matching for privacy-preserving ad measurement, a cornerstone of modern advertising analytics. We introduce a comprehensive cryptographic framework leveraging reversed Oblivious Pseudorandom Functions (OPRF) and novel blind key rotation techniques to support secure matching across multiple identifiers. Our design prevents cross-identifier linkages and includes a differentially private mechanism to obfuscate intersection sizes, mitigating risks such as membership inference attacks. We present a concrete construction of our protocol that achieves both strong privacy guarantees and high efficiency. It scales to large datasets, offering a practical and scalable solution for privacy-centric applications like secure ad conversion tracking. By combining rigorous cryptographic principles with differential privacy, our work addresses a critical need in the advertising industry, setting a new standard for privacy-preserving ad measurement frameworks.

cs.CR↗

arXiv2025

Reasoning Through Execution: Unifying Process and Outcome Rewards for Code Generation

Large Language Models excel at code generation yet struggle with complex programming tasks that demand sophisticated reasoning. To bridge this gap, traditional process supervision relies on learned reward models requiring costly training data and suffering from reward misalignment, while outcome supervision fails for complex tasks needing coordinated intermediate steps. We introduce Outcome Refining Process Supervision, which unifies process and outcome supervision by leveraging executable verification: a tree-structured search framework generates strategic alternatives, profiles execution metrics, and scores candidates via self-critique mechanisms that integrate runtime feedback with reasoning. Experiments across 5 models and 3 benchmarks show consistent gains, with 26.9% higher correctness and 42.2% improved code efficiency. The results demonstrate that ORPS enables LLMs to overcome local optima in code generation, suggesting a promising direction for combining verifiable outcomes with structured reasoning to tackle complex challenges. We open-source at: https://github.com/zhuohaoyu/ORPS

cs.CL↗

arXiv2025

VLM-R$^3$: Region Recognition, Reasoning, and Refinement for Enhanced Multimodal Chain-of-Thought

Recently, reasoning-based MLLMs have achieved a degree of success in generating long-form textual reasoning chains. However, they still struggle with complex tasks that necessitate dynamic and iterative focusing on and revisiting of visual regions to achieve precise grounding of textual reasoning in visual evidence. We introduce \textbf{VLM-R$^3$} (\textbf{V}isual \textbf{L}anguage \textbf{M}odel with \textbf{R}egion \textbf{R}ecognition and \textbf{R}easoning), a framework that equips an MLLM with the ability to (i) decide \emph{when} additional visual evidence is needed, (ii) determine \emph{where} to ground within the image, and (iii) seamlessly weave the relevant sub-image content back into an interleaved chain-of-thought. The core of our method is \textbf{Region-Conditioned Reinforcement Policy Optimization (R-GRPO)}, a training paradigm that rewards the model for selecting informative regions, formulating appropriate transformations (e.g.\ crop, zoom), and integrating the resulting visual context into subsequent reasoning steps. To bootstrap this policy, we compile a modest but carefully curated Visuo-Lingual Interleaved Rationale (VLIR) corpus that provides step-level supervision on region selection and textual justification. Extensive experiments on MathVista, ScienceQA, and other benchmarks show that VLM-R$^3$ sets a new state of the art in zero-shot and few-shot settings, with the largest gains appearing on questions demanding subtle spatial reasoning or fine-grained visual cue extraction.

cs.CV↗

arXiv2025

MPL: Multiple Programming Languages with Large Language Models for Information Extraction

Recent research in information extraction (IE) focuses on utilizing code-style inputs to enhance structured output generation. The intuition behind this is that the programming languages (PLs) inherently exhibit greater structural organization than natural languages (NLs). This structural advantage makes PLs particularly suited for IE tasks. Nevertheless, existing research primarily focuses on Python for code-style simulation, overlooking the potential of other widely-used PLs (e.g., C++ and Java) during the supervised fine-tuning (SFT) phase. In this research, we propose \textbf{M}ultiple \textbf{P}rogramming \textbf{L}anguages with large language models for information extraction (abbreviated as \textbf{MPL}), a novel framework that explores the potential of incorporating different PLs in the SFT phase. Additionally, we introduce \texttt{function-prompt} with virtual running to simulate code-style inputs more effectively and efficiently. Experimental results on a wide range of datasets demonstrate the effectiveness of MPL. Furthermore, we conduct extensive experiments to provide a comprehensive analysis. We have released our code for future research.

cs.CL↗

arXiv2025

Mitigating Spurious Correlations with Causal Logit Perturbation

Deep learning has seen widespread success in various domains such as science, industry, and society. However, it is acknowledged that certain approaches suffer from non-robustness, relying on spurious correlations for predictions. Addressing these limitations is of paramount importance, necessitating the development of methods that can disentangle spurious correlations. {This study attempts to implement causal models via logit perturbations and introduces a novel Causal Logit Perturbation (CLP) framework to train classifiers with generated causal logit perturbations for individual samples, thereby mitigating the spurious associations between non-causal attributes (i.e., image backgrounds) and classes.} {Our framework employs a} perturbation network to generate sample-wise logit perturbations using a series of training characteristics of samples as inputs. The whole framework is optimized by an online meta-learning-based learning algorithm and leverages human causal knowledge by augmenting metadata in both counterfactual and factual manners. Empirical evaluations on four typical biased learning scenarios, including long-tail learning, noisy label learning, generalized long-tail learning, and subpopulation shift learning, demonstrate that CLP consistently achieves state-of-the-art performance. Moreover, visualization results support the effectiveness of the generated causal perturbations in redirecting model attention towards causal image attributes and dismantling spurious associations.

cs.LG↗

arXiv2025

Can You Really Trust Code Copilots? Evaluating Large Language Models from a Code Security Perspective

Code security and usability are both essential for various coding assistant applications driven by large language models (LLMs). Current code security benchmarks focus solely on single evaluation task and paradigm, such as code completion and generation, lacking comprehensive assessment across dimensions like secure code generation, vulnerability repair and discrimination. In this paper, we first propose CoV-Eval, a multi-task benchmark covering various tasks such as code completion, vulnerability repair, vulnerability detection and classification, for comprehensive evaluation of LLM code security. Besides, we developed VC-Judge, an improved judgment model that aligns closely with human experts and can review LLM-generated programs for vulnerabilities in a more efficient and reliable way. We conduct a comprehensive evaluation of 20 proprietary and open-source LLMs. Overall, while most LLMs identify vulnerable codes well, they still tend to generate insecure codes and struggle with recognizing specific vulnerability types and performing repairs. Extensive experiments and qualitative analyses reveal key challenges and optimization directions, offering insights for future research in LLM code security.

cs.CL↗

arXiv2025

3D Surface Reconstruction with Enhanced High-Frequency Details

Neural implicit 3D reconstruction can reproduce shapes without 3D supervision, and it learns the 3D scene through volume rendering methods and neural implicit representations. Current neural surface reconstruction methods tend to randomly sample the entire image, making it difficult to learn high-frequency details on the surface, and thus the reconstruction results tend to be too smooth. We designed a method (FreNeuS) based on high-frequency information to solve the problem of insufficient surface detail. Specifically, FreNeuS uses pixel gradient changes to easily acquire high-frequency regions in an image and uses the obtained high-frequency information to guide surface detail reconstruction. High-frequency information is first used to guide the dynamic sampling of rays, applying different sampling strategies according to variations in high-frequency regions. To further enhance the focus on surface details, we have designed a high-frequency weighting method that constrains the representation of high-frequency details during the reconstruction process. Qualitative and quantitative results show that our method can reconstruct fine surface details and obtain better surface reconstruction quality compared to existing methods. In addition, our method is more applicable and can be generalized to any NeuS-based work.

cs.CV↗

arXiv2025

Machine-Learning Insights into the Entanglement-trainability Correlation of Parameterized Quantum Circuits

Variational quantum algorithms (VQAs) have emerged as the leading strategy to obtain quantum advantage on the current noisy intermediate-scale devices. However, their entanglement-trainability correlation, as the major reason for the barren plateau (BP) phenomenon, poses a challenge to their applications. In this Letter, we suggest a gate-to-tensor (GTT) encoding method for parameterized quantum circuits (PQCs), with which two long short-term memory networks (L-G networks) are trained to predict both entanglement and trainability. The remarkable capabilities of the L-G networks afford a statistical way to delve into the entanglement-trainability correlation of PQCs within a dataset encompassing millions of instances. This machine-learning-driven method first confirms that the more entanglement, the more possible the BP problem. Then, we observe that there still exist PQCs with both high entanglement and high trainability. Furthermore, the trained L-G networks result in an impressive increase in time efficiency by about one million times when constructing a PQC with specific entanglement and trainability, demonstrating their practical applications in VQAs.

quant-ph↗

arXiv2025

SaRO: Enhancing LLM Safety through Reasoning-based Alignment

Current safety alignment techniques for large language models (LLMs) face two key challenges: (1) under-generalization, which leaves models vulnerable to novel jailbreak attacks, and (2) over-alignment, which leads to the excessive refusal of benign instructions. Our preliminary investigation reveals semantic overlap between jailbreak/harmful queries and normal prompts in embedding space, suggesting that more effective safety alignment requires a deeper semantic understanding. This motivates us to incorporate safety-policy-driven reasoning into the alignment process. To this end, we propose the Safety-oriented Reasoning Optimization Framework (SaRO), which consists of two stages: (1) Reasoning-style Warmup (RW) that enables LLMs to internalize long-chain reasoning through supervised fine-tuning, and (2) Safety-oriented Reasoning Process Optimization (SRPO) that promotes safety reflection via direct preference optimization (DPO). Extensive experiments demonstrate the superiority of SaRO over traditional alignment methods.

cs.CL↗

arXiv2025

A Survey on Evaluating Large Language Models in Code Generation Tasks

This paper provides a comprehensive review of the current methods and metrics used to evaluate the performance of Large Language Models (LLMs) in code generation tasks. With the rapid growth in demand for automated software development, LLMs have demonstrated significant potential in the field of code generation. The paper begins by reviewing the historical development of LLMs and their applications in code generation. Next, it details various methods and metrics for assessing the code generation capabilities of LLMs, including code correctness, efficiency, readability, and evaluation methods based on expert review and user experience. The paper also evaluates the widely used benchmark datasets, identifying their limitations and proposing directions for future improvements. Specifically, the paper analyzes the performance of code generation models across different tasks by combining multiple evaluation metrics, such as code compilation/interpretation success rates, unit test pass rates, and performance and efficiency metrics, to comprehensively assess the practical application of LLMs in code generation. Finally, the paper discusses the challenges faced in evaluating LLMs in code generation, particularly how to ensure the comprehensiveness and accuracy of evaluation methods and how to adapt to the evolving practices of software development. These analyses and discussions provide valuable insights for further optimizing and improving the application of LLMs in code generation tasks.

cs.SE↗

arXiv2025

SampleMix: A Sample-wise Pre-training Data Mixing Strategey by Coordinating Data Quality and Diversity

Existing pretraining data mixing methods for large language models (LLMs) typically follow a domain-wise methodology, a top-down process that first determines domain weights and then performs uniform data sampling across each domain. However, these approaches neglect significant inter-domain overlaps and commonalities, failing to control the global diversity of the constructed training dataset. Further, uniform sampling within domains ignores fine-grained sample-specific features, potentially leading to suboptimal data distribution. To address these shortcomings, we propose a novel sample-wise data mixture approach based on a bottom-up paradigm. This method performs global cross-domain sampling by systematically evaluating the quality and diversity of each sample, thereby dynamically determining the optimal domain distribution. Comprehensive experiments across multiple downstream tasks and perplexity assessments demonstrate that SampleMix surpasses existing domain-based methods. Meanwhile, SampleMix requires 1.4x to 2.1x training steps to achieves the baselines' performance, highlighting the substantial potential of SampleMix to optimize pre-training data.

cs.CL↗

arXiv2025

Diffusion-Enhanced Optimization of Variational Quantum Eigensolver for General Hamiltonians

Variational quantum algorithms (VQAs) have emerged as a promising approach for achieving quantum advantage on current noisy intermediate-scale quantum devices. However, their large-scale applications are significantly hindered by optimization challenges, such as the barren plateau (BP) phenomenon, local minima, and numerous iteration demands. In this work, we leverage denoising diffusion models (DMs) to address these difficulties. The DM is trained on a few data points in the Heisenberg model parameter space and then can be guided to generate high-performance parameters for parameterized quantum circuits (PQCs) in variational quantum eigensolver (VQE) tasks for general Hamiltonians. Numerical experiments demonstrate that DM-parameterized VQE can explore the ground-state energies of Heisenberg models with parameters not included in the training dataset. Even when applied to previously unseen Hamiltonians, such as the Ising and Hubbard models, it can generate the appropriate initial state to achieve rapid convergence and mitigate the BP and local minima problems. These results highlight the effectiveness of our proposed method in improving optimization efficiency for general Hamiltonians.

quant-ph↗