arXiv subjects

Zheng Zhao

Publications and source records attributed to Zheng Zhao.

At least 37 records · Page 2Linked to original sources

arXiv2026

CodeCircuit: Toward Inferring LLM-Generated Code Correctness via Attribution Graphs

Current paradigms for code verification rely heavily on external mechanisms-such as execution-based unit tests or auxiliary LLM judges-which are often labor-intensive or limited by the judging model's own capabilities. This raises a fundamental, yet unexplored question: Can an LLM's functional correctness be assessed purely from its internal computational structure? Our primary objective is to investigate whether the model's neural dynamics encode internally decodable signals that are predictive of logical validity during code generation. Inspired by mechanistic interpretability, we propose to treat code verification as a mechanistic diagnostic task, mapping the model's explicit algorithmic trajectory into line-level attribution graphs. By decomposing complex residual flows, we aim to identify the structural signatures that distinguish sound reasoning from logical failure within the model's internal circuits. Analysis across Python, C++, and Java confirms that intrinsic correctness signals are robust across diverse syntaxes. Topological features from these internal graphs predict correctness more reliably than surface heuristics and enable targeted causal interventions to fix erroneous logic. These findings establish internal introspection as a decodable property for verifying generated code. Our code is at https:// github.com/bruno686/CodeCircuit.

cs.SE↗

arXiv2026

Utilising Gradient-Based Proposals Within Sequential Monte Carlo Samplers for Training of Partial Bayesian Neural Networks

Partial Bayesian neural networks (pBNNs) have been shown to perform competitively with fully Bayesian neural networks while only having a subset of the parameters be stochastic. Using sequential Monte Carlo (SMC) samplers as the inference method for pBNNs gives a non-parametric probabilistic estimation of the stochastic parameters, and has shown improved performance over parametric methods. In this paper we introduce a new SMC-based training method for pBNNs by utilising a guided proposal and incorporating gradient-based Markov kernels, which gives us better scalability on high dimensional problems. We show that our new method outperforms the state-of-the-art in terms of predictive performance and optimal loss. We also show that pBNNs scale well with larger batch sizes, resulting in significantly reduced training times and often better performance.

cs.LG↗

arXiv2025

Iterative Multilingual Spectral Attribute Erasure

Multilingual representations embed words with similar meanings to share a common semantic space across languages, creating opportunities to transfer debiasing effects between languages. However, existing methods for debiasing are unable to exploit this opportunity because they operate on individual languages. We present Iterative Multilingual Spectral Attribute Erasure (IMSAE), which identifies and mitigates joint bias subspaces across multiple languages through iterative SVD-based truncation. Evaluating IMSAE across eight languages and five demographic dimensions, we demonstrate its effectiveness in both standard and zero-shot settings, where target language data is unavailable, but linguistically similar languages can be used for debiasing. Our comprehensive experiments across diverse language models (BERT, LLaMA, Mistral) show that IMSAE outperforms traditional monolingual and cross-lingual approaches while maintaining model utility.

cs.CL↗

arXiv2025

Helicity amplitudes of $N(1520)$ and $N(1535)$ including pentaquark components

We investigate the helicity amplitudes of the $N(1520)$ and $N(1535)$ resonances in the electromagnetic transition process $γ^*p\to N^*$, where the $N(1520)$ and $N(1535)$ are assumed to include both the $L=1$ three-quark and $L=0$ spatial symmetric $q^4\bar q$ pentaquark components. The helicity transition amplitudes $A_{1/2}$, $A_{3/2}$ (for spin 3/2 states) and $S_{1/2}$ are computed within the constituent quark model. The inclusion of the $q^4 \bar q$ pentaquark components via the $γ^*\to q\bar q$ diagram significantly improves the theoretical description of the helicity amplitudes for both $N(1520)$ and $N(1535)$, yielding a closer agreement with experimental data compared to the pure three-quark picture. The work reveals that the $N(1520)$ and $N(1535)$ resonances may contain a considerable pentaquark components.

hep-ph↗

arXiv2025

Solving Linear-Gaussian Bayesian Inverse Problems with Decoupled Diffusion Sequential Monte Carlo

A recent line of research has exploited pre-trained generative diffusion models as priors for solving Bayesian inverse problems. We contribute to this research direction by designing a sequential Monte Carlo method for linear-Gaussian inverse problems which builds on "decoupled diffusion", where the generative process is designed such that larger updates to the sample are possible. The method is asymptotically exact and we demonstrate the effectiveness of our Decoupled Diffusion Sequential Monte Carlo (DDSMC) algorithm on both synthetic as well as protein and image data. Further, we demonstrate how the approach can be extended to discrete data.

cs.LG↗

arXiv2025

Taming Diffusion Models for Image Restoration: A Review

Diffusion models have achieved remarkable progress in generative modelling, particularly in enhancing image quality to conform to human preferences. Recently, these models have also been applied to low-level computer vision for photo-realistic image restoration (IR) in tasks such as image denoising, deblurring, dehazing, etc. In this review paper, we introduce key constructions in diffusion models and survey contemporary techniques that make use of diffusion models in solving general IR tasks. Furthermore, we point out the main challenges and limitations of existing diffusion-based IR frameworks and provide potential directions for future work.

cs.CV↗

arXiv2025

ESS-Flow: Training-free guidance of flow-based models as inference in source space

Guiding pretrained flow-based generative models for conditional generation or to produce samples with desired target properties enables solving diverse tasks without retraining on paired data. We present ESS-Flow, a gradient-free method that leverages the typically Gaussian prior of the source distribution in flow-based models to perform Bayesian inference directly in the source space using Elliptical Slice Sampling. ESS-Flow only requires forward passes through the generative model and observation process, no gradient or Jacobian computations, and is applicable even when gradients are unreliable or unavailable, such as with simulation-based observations or quantization in the generation or observation process. We demonstrate its effectiveness on designing materials with desired target properties and predicting protein structures from sparse inter-residue distance measurements.

cs.LG↗

arXiv2025

Interpolation-supplemented lattice Boltzmann simulation of thermal convection on non-uniform meshes

We present a systematic evaluation of an interpolation-supplemented lattice Boltzmann method (ISLBM) for simulating buoyancy-driven thermal convection on non-uniform meshes. The ISLBM extends the standard lattice Boltzmann framework by incorporating quadratic interpolation during the streaming step, enabling flexible mesh refinement near solid boundaries while maintaining algorithmic simplicity and parallel scalability. The method is implemented for a two-dimensional side-heated cavity at high Rayleigh numbers $10^6\leq Ra \leq 10^8$, and for a three-dimensional side-heated cavity at $10^5\leq Ra \leq 10^7$, with the Prandtl number fixed at $Pr=0.71$. Benchmark results show that the ISLBM accurately captures thermal and velocity boundary layers, yielding Nusselt and Reynolds numbers in close agreement with high-fidelity reference data. Grid-convergence studies demonstrate nearly third-order accuracy for global quantities and about second-order for local fields. We further assess the computational performance of the in-house LBM solver against two open-source solvers: Nek5000 based on the spectral element method, and OpenFOAM based on the finite volume method. Performance metrics, including million lattice updates per second (MLUPS) and wall-clock time per dimensionless time unit (WCTpDT), indicate that the ISLBM offers one to three orders of magnitude higher efficiency in large-scale simulations. On GPU architectures, the ISLBM retains high computational performance: throughput on non-uniform meshes reaches 60-70% of that on uniform meshes in terms of MLUPS, while the cost in WCTpDT is about three times higher. These results highlight the potential of interpolation-based LBM approaches for high-fidelity simulations of thermal convection on non-uniform meshes, providing a robust foundation for future extensions to turbulent flows.

physics.flu-dyn↗

arXiv2025

TASR: Timestep-Aware Diffusion Model for Image Super-Resolution

Diffusion models have recently achieved outstanding results in the field of image super-resolution. These methods typically inject low-resolution (LR) images via ControlNet.In this paper, we first explore the temporal dynamics of information infusion through ControlNet, revealing that the input from LR images predominantly influences the initial stages of the denoising process. Leveraging this insight, we introduce a novel timestep-aware diffusion model that adaptively integrates features from both ControlNet and the pre-trained Stable Diffusion (SD). Our method enhances the transmission of LR information in the early stages of diffusion to guarantee image fidelity and stimulates the generation ability of the SD model itself more in the later stages to enhance the detail of generated images. To train this method, we propose a timestep-aware training strategy that adopts distinct losses at varying timesteps and acts on disparate modules. Experiments on benchmark datasets demonstrate the effectiveness of our method. Code: https://github.com/SleepyLin/TASR

cs.CV↗

arXiv2025

Generative diffusion posterior sampling for informative likelihoods

Sequential Monte Carlo (SMC) methods have recently shown successful results for conditional sampling of generative diffusion models. In this paper we propose a new diffusion posterior SMC sampler achieving improved statistical efficiencies, particularly under outlier conditions or highly informative likelihoods. The key idea is to construct an observation path that correlates with the diffusion model and to design the sampler to leverage this correlation for more efficient sampling. Empirical results conclude the efficiency.

stat.ML↗

arXiv2025

CROP: Circuit Retrieval and Optimization with Parameter Guidance using LLMs

Modern very large-scale integration (VLSI) design requires the implementation of integrated circuits using electronic design automation (EDA) tools. Due to the complexity of EDA algorithms, the vast parameter space poses a huge challenge to chip design optimization, as the combination of even moderate numbers of parameters creates an enormous solution space to explore. Manual parameter selection remains industrial practice despite being excessively laborious and limited by expert experience. To address this issue, we present CROP, the first large language model (LLM)-powered automatic VLSI design flow tuning framework. Our approach includes: (1) a scalable methodology for transforming RTL source code into dense vector representations, (2) an embedding-based retrieval system for matching designs with semantically similar circuits, and (3) a retrieval-augmented generation (RAG)-enhanced LLM-guided parameter search system that constrains the search process with prior knowledge from similar designs. Experiment results demonstrate CROP's ability to achieve superior quality-of-results (QoR) with fewer iterations than existing approaches on industrial designs, including a 9.9% reduction in power consumption.

cs.LG↗

arXiv2025

Frequency-Domain Denoising-Based in Vivo Fluorescence Imaging

The second near-infrared window (NIR-II, 900-1,880 nm) has been pivotal in advancing in vivo fluorescence imaging due to its superior penetration depth and contrast. Yet, its clinical utility remains limited by insufficient imaging temporal-spatial resolution and the absence of U.S. Food and Drug Administration (FDA)-approved NIR-II contrast agents. This work presents a frequency-domain denoising (FDD)-based in vivo fluorescence imaging technique, which can improve signal-to-background ratio (SBR) and signal-to-noise ratio (SNR) by more than 2,500-fold and 300-fold, respectively. The great enhancement yields a doubled penetration depth and a 95% reduction in contrast agent dosage or excitation light intensity for mouse vascular imaging. Additionally, we achieved a SBR far exceeded the Rose criterion in the observation of tumor margins and vessels in mice using Indocyanine Green (ICG), demonstrating the feasibility of NIR-II surgical navigation with FDA-approved agents. Furthermore, a 600 Hz real-time video enables visualization of the entire contrast agent diffusion process within the mouse body and differentiation between arteries and veins. This innovative technique, characterized by exceptional sensitivity, efficiency, and robustness, presents a promising solution for clinical applications, particularly in NIR-II surgical navigation.

physics.bio-ph↗

arXiv2025

PersonaLens: A Benchmark for Personalization Evaluation in Conversational AI Assistants

Large language models (LLMs) have advanced conversational AI assistants. However, systematically evaluating how well these assistants apply personalization--adapting to individual user preferences while completing tasks--remains challenging. Existing personalization benchmarks focus on chit-chat, non-conversational tasks, or narrow domains, failing to capture the complexities of personalized task-oriented assistance. To address this, we introduce PersonaLens, a comprehensive benchmark for evaluating personalization in task-oriented AI assistants. Our benchmark features diverse user profiles equipped with rich preferences and interaction histories, along with two specialized LLM-based agents: a user agent that engages in realistic task-oriented dialogues with AI assistants, and a judge agent that employs the LLM-as-a-Judge paradigm to assess personalization, response quality, and task success. Through extensive experiments with current LLM assistants across diverse tasks, we reveal significant variability in their personalization capabilities, providing crucial insights for advancing conversational AI systems.

cs.CL↗

arXiv2025

What Is That Talk About? A Video-to-Text Summarization Dataset for Scientific Presentations

Transforming recorded videos into concise and accurate textual summaries is a growing challenge in multimodal learning. This paper introduces VISTA, a dataset specifically designed for video-to-text summarization in scientific domains. VISTA contains 18,599 recorded AI conference presentations paired with their corresponding paper abstracts. We benchmark the performance of state-of-the-art large models and apply a plan-based framework to better capture the structured nature of abstracts. Both human and automated evaluations confirm that explicit planning enhances summary quality and factual consistency. However, a considerable gap remains between models and human performance, highlighting the challenges of our dataset. This study aims to pave the way for future research on scientific video-to-text summarization.

cs.CL↗

arXiv2025

Humble your Overconfident Networks: Unlearning Overfitting via Sequential Monte Carlo Tempered Deep Ensembles

Sequential Monte Carlo (SMC) methods offer a principled approach to Bayesian uncertainty quantification but are traditionally limited by the need for full-batch gradient evaluations. We introduce a scalable variant by incorporating Stochastic Gradient Hamiltonian Monte Carlo (SGHMC) proposals into SMC, enabling efficient mini-batch based sampling. Our resulting SMCSGHMC algorithm outperforms standard stochastic gradient descent (SGD) and deep ensembles across image classification, out-of-distribution (OOD) detection, and transfer learning tasks. We further show that SMCSGHMC mitigates overfitting and improves calibration, providing a flexible, scalable pathway for converting pretrained neural networks into well-calibrated Bayesian models.

stat.ML↗

arXiv2025

Probabilistic Zeeman-Doppler imaging of stellar magnetic fields: I. Analysis of tau Scorpii in the weak-field limit

Zeeman-Doppler imaging (ZDI) is used to study the surface magnetic field topology of stars, based on high-resolution spectropolarimetric time series observations. Multiple ZDI inversions have been conducted for the early B-type star tau Sco, which has been found to exhibit a weak but complex non-dipolar surface magnetic field. The classical ZDI framework suffers from a significant limitation in that it provides little to no reliable uncertainty quantification for the reconstructed magnetic field maps, with essentially all published results being confined to point estimates. To fill this gap, we propose a Bayesian framework for probabilistic ZDI. Here, the proposed framework is demonstrated on tau Sco in the weak-field limit. We propose three distinct statistical models, and use archival ESPaDOnS high-resolution Stokes V observations to carry out the probabilistic magnetic inversion in closed form. The surface magnetic field is parameterised by a high-dimensional spherical-harmonic expansion. By comparing three different prior distributions over the latent variables in the spherical-harmonic decomposition, our results showcase the ZDI sensitivity to various hyperparameters. The mean magnetic field maps are qualitatively similar to previously published point estimates, but analysis of the magnetic energy distribution indicates high uncertainty and higher energy content at low angular degrees l. Our results effectively demonstrate that, for stars in the weak-field regime, reliable uncertainty quantification of recovered magnetic field maps can be obtained in closed form with natural assumptions on the statistical model. Future work will explore extending this framework beyond the weak-field approximation and incorporating prior uncertainty over multiple stellar parameters in more complex magnetic inversion problems.

astro-ph.SR↗

arXiv2025

Conditioning diffusion models by explicit forward-backward bridging

Given an unconditional diffusion model targeting a joint model $π(x, y)$, using it to perform conditional simulation $π(x \mid y)$ is still largely an open question and is typically achieved by learning conditional drifts to the denoising SDE after the fact. In this work, we express \emph{exact} conditional simulation within the \emph{approximate} diffusion model as an inference problem on an augmented space corresponding to a partial SDE bridge. This perspective allows us to implement efficient and principled particle Gibbs and pseudo-marginal samplers marginally targeting the conditional distribution $π(x \mid y)$. Contrary to existing methodology, our methods do not introduce any additional approximation to the unconditional diffusion model aside from the Monte Carlo error. We showcase the benefits and drawbacks of our approach on a series of synthetic and real data examples.

stat.ML↗

arXiv2025

Conditional sampling within generative diffusion models

Generative diffusions are a powerful class of Monte Carlo samplers that leverage bridging Markov processes to approximate complex, high-dimensional distributions, such as those found in image processing and language models. Despite their success in these domains, an important open challenge remains: extending these techniques to sample from conditional distributions, as required in, for example, Bayesian inverse problems. In this paper, we present a comprehensive review of existing computational approaches to conditional sampling within generative diffusion models. Specifically, we highlight key methodologies that either utilise the joint distribution, or rely on (pre-trained) marginal distributions with explicit likelihoods, to construct conditional generative samplers.

stat.ML↗