arXiv subjects

Shengda Luo

Publications and source records attributed to Shengda Luo.

8 recordsLinked to original sources

arXiv2026

SPARD: Defending Harmful Fine-Tuning Attack via Safety Projection with Relevance-Diversity Data Selection

Fine-tuning large language models often undermines their safety alignment, a problem further amplified by harmful fine-tuning attacks in which adversarial data removes safeguards and induces unsafe behaviors. We propose SPARD, a defense framework that integrates Safety-Projected Alternating optimization with Relevance-Diversity aware data selection. SPARD employs SPAG, which optimizes alternatively between utility updates and explicit safety projections with a set of safe data to enforce safety constraints. To curate safe data, we introduce a Relevance-Diversity Determinantal Point Process to select compact safe data, balancing task relevance and safety coverage. Experiments on GSM8K and OpenBookQA under four harmful fine-tuning attacks demonstrate that SPARD consistently achieves the lowest average attack success rates, substantially outperforming state-of-the-art defense methods, while maintaining high task accuracy. Code is available at https://github.com/shuhao02/SPARD.

cs.LG↗

arXiv2026

Reinforcing 3D Understanding in Point-VLMs via Geometric Reward Credit Assignment

Point-Vision-Language Models promise to empower embodied agents with executable spatial reasoning, yet they frequently succumb to geometric hallucination where predicted 3D structures contradict the observed 2D reality. We identify a key cause of this failure not as a representation bottleneck but as a structural misalignment in reinforcement learning, where sparse geometric tokens are drowned out by noisy and broadcasted sequence-level rewards. To resolve this causal dilution, we propose Geometric Reward Credit Assignment, a framework that disentangles holistic supervision into field-specific signals and routes them exclusively to their responsible token spans. This mechanism transforms vague feedback into precise gradient updates and effectively turns generic policy optimization into targeted structural alignment. Furthermore, we internalize physical constraints via a Reprojection-Consistency term which serves as a cross-modal verifier to penalize physically impossible geometries. Validated on a calibrated benchmark derived from ShapeNetCore, our approach bridges the reliability gap by boosting 3D KPA from 0.64 to 0.93, increasing 3D bounding box intersection over union to 0.686, and raising reprojection consistency scores to 0.852. Crucially, these gains are achieved while maintaining robust 2D localization performance, marking a meaningful step from plausible textual outputs toward physically verifiable spatial predictions.

cs.CV↗

arXiv2025

WorldGenBench: A World-Knowledge-Integrated Benchmark for Reasoning-Driven Text-to-Image Generation

Recent advances in text-to-image (T2I) generation have achieved impressive results, yet existing models still struggle with prompts that require rich world knowledge and implicit reasoning: both of which are critical for producing semantically accurate, coherent, and contextually appropriate images in real-world scenarios. To address this gap, we introduce \textbf{WorldGenBench}, a benchmark designed to systematically evaluate T2I models' world knowledge grounding and implicit inferential capabilities, covering both the humanities and nature domains. We propose the \textbf{Knowledge Checklist Score}, a structured metric that measures how well generated images satisfy key semantic expectations. Experiments across 21 state-of-the-art models reveal that while diffusion models lead among open-source methods, proprietary auto-regressive models like GPT-4o exhibit significantly stronger reasoning and knowledge integration. Our findings highlight the need for deeper understanding and inference capabilities in next-generation T2I systems. Project Page: \href{https://dwanzhang-ai.github.io/WorldGenBench/}{https://dwanzhang-ai.github.io/WorldGenBench/}

cs.CV↗

arXiv2024

Gradient-Guided Modality Decoupling for Missing-Modality Robustness

Multimodal learning with incomplete input data (missing modality) is practical and challenging. In this work, we conduct an in-depth analysis of this challenge and find that modality dominance has a significant negative impact on the model training, greatly degrading the missing modality performance. Motivated by Grad-CAM, we introduce a novel indicator, gradients, to monitor and reduce modality dominance which widely exists in the missing-modality scenario. In aid of this indicator, we present a novel Gradient-guided Modality Decoupling (GMD) method to decouple the dependency on dominating modalities. Specifically, GMD removes the conflicted gradient components from different modalities to achieve this decoupling, significantly improving the performance. In addition, to flexibly handle modal-incomplete data, we design a parameter-efficient Dynamic Sharing (DS) framework which can adaptively switch on/off the network parameters based on whether one modality is available. We conduct extensive experiments on three popular multimodal benchmarks, including BraTS 2018 for medical segmentation, CMU-MOSI, and CMU-MOSEI for sentiment analysis. The results show that our method can significantly outperform the competitors, showing the effectiveness of the proposed solutions. Our code is released here: https://github.com/HaoWang420/Gradient-guided-Modality-Decoupling.

cs.CV↗

arXiv2023

Strip-MLP: Efficient Token Interaction for Vision MLP

Token interaction operation is one of the core modules in MLP-based models to exchange and aggregate information between different spatial locations. However, the power of token interaction on the spatial dimension is highly dependent on the spatial resolution of the feature maps, which limits the model's expressive ability, especially in deep layers where the feature are down-sampled to a small spatial size. To address this issue, we present a novel method called \textbf{Strip-MLP} to enrich the token interaction power in three ways. Firstly, we introduce a new MLP paradigm called Strip MLP layer that allows the token to interact with other tokens in a cross-strip manner, enabling the tokens in a row (or column) to contribute to the information aggregations in adjacent but different strips of rows (or columns). Secondly, a \textbf{C}ascade \textbf{G}roup \textbf{S}trip \textbf{M}ixing \textbf{M}odule (CGSMM) is proposed to overcome the performance degradation caused by small spatial feature size. The module allows tokens to interact more effectively in the manners of within-patch and cross-patch, which is independent to the feature spatial size. Finally, based on the Strip MLP layer, we propose a novel \textbf{L}ocal \textbf{S}trip \textbf{M}ixing \textbf{M}odule (LSMM) to boost the token interaction power in the local region. Extensive experiments demonstrate that Strip-MLP significantly improves the performance of MLP-based models on small datasets and obtains comparable or even better results on ImageNet. In particular, Strip-MLP models achieve higher average Top-1 accuracy than existing MLP-based models by +2.44\% on Caltech-101 and +2.16\% on CIFAR-100. The source codes will be available at~\href{https://github.com/Med-Process/Strip_MLP{https://github.com/Med-Process/Strip\_MLP}.

cs.CV↗

arXiv2021

Imbalance-Aware Self-Supervised Learning for 3D Radiomic Representations

Radiomic representations can quantify properties of regions of interest in medical image data. Classically, they account for pre-defined statistics of shape, texture, and other low-level image features. Alternatively, deep learning-based representations are derived from supervised learning but require expensive annotations from experts and often suffer from overfitting and data imbalance issues. In this work, we address the challenge of learning representations of 3D medical images for an effective quantification under data imbalance. We propose a \emph{self-supervised} representation learning framework to learn high-level features of 3D volumes as a complement to existing radiomics features. Specifically, we demonstrate how to learn image representations in a self-supervised fashion using a 3D Siamese network. More importantly, we deal with data imbalance by exploiting two unsupervised strategies: a) sample re-weighting, and b) balancing the composition of training batches. When combining our learned self-supervised feature with traditional radiomics, we show significant improvement in brain tumor classification and lung cancer staging tasks covering MRI and CT imaging modalities.

cs.CV↗

arXiv2020

Searches for Pulsar-like Candidates from Unidentified Objects in the Third Catalog of Hard Fermi-LAT (3FHL) sources with Machine Learning Techniques

We report the results of searching pulsar-like candidates from the unidentified objects in the $3^{\rm rd}$ Catalog of Hard Fermi-LAT sources (3FHL). Using a machine-learning based classification scheme with a nominal accuracy of $\sim98\%$, we have selected 27 pulsar-like objects from 200 unidentified 3FHL sources for an identification campaign. Using archival data, X-ray sources are found within the $\gamma-$ray error ellipses of 10 3FHL pulsar-like candidates. Within the error circles of the much better constrained X-ray positions, we have also searched for the optical/infrared counterparts and examined their spectral energy distributions. Among our short-listed candidates, the most secure identification is the association of 3FHL J1823.3-1339 and its X-ray counterpart with the globular cluster Mercer 5. The $\gamma-$rays from the source can be contributed by a population of millisecond pulsars residing in the cluster. This makes Mercer 5 as one of the slowly growing hard $\gamma-$ray population of globular clusters with emission $>10$ GeV. Very recently, another candidate picked by our classification scheme, 3FHL J1405.1-6118, has been identified as a new $\gamma-$ray binary with an orbital period of $13.7$ days. Our X-ray analysis with a short Chandra observation has found a possible periodic signal candidate of $\sim1.4$ hrs and a putative extended X-ray tail of $\sim20$ arcsec long. Spectral energy distribution of its optical/infrared counterpart conforms with a blackbody of $T_{\rm bb}\sim40000$ K and $R_{\rm bb}\sim12R_{\odot}$ at a distance of 7.7 kpc. This is consistent with its identification as an early O star as found by infrared spectroscopy.

astro-ph.HE↗

arXiv2020

An investigation on the factors affecting machine learning classifications in $\gamma$-ray astronomy

We have investigated a number of factors that can have significant impacts on the classification performance of $\gamma$-ray sources detected by Fermi Large Area Telescope (LAT) with machine learning techniques. We show that a framework of automatic feature selection can construct a simple model with a small set of features which yields better performance over previous results. Secondly, because of the small sample size of the training/test sets of certain classes in $\gamma$-ray, nested re-sampling and cross-validations are suggested for quantifying the statistical fluctuations of the quoted accuracy. We have also constructed a test set by cross-matching the identified active galactic nuclei (AGNs) and the pulsars (PSRs) in the Fermi LAT eight-year point source catalog (4FGL) with those unidentified sources in the previous 3$^{\rm rd}$ Fermi LAT Source Catalog (3FGL). Using this cross-matched set, we show that some features used for building classification model with the identified source can suffer from the problem of covariate shift, which can be a result of various observational effects. This can possibly hamper the actual performance when one applies such model in classifying unidentified sources. Using our framework, both AGN/PSR and young pulsar (YNG)/millisecond pulsar (MSP) classifiers are automatically updated with the new features and the enlarged training samples in 4FGL catalog incorporated. Using a two-layer model with these updated classifiers, we have selected 20 promising MSP candidates with confidence scores $>98\%$ from the unidentified sources in 4FGL catalog which can provide inputs for a multi-wavelength identification campaign.

astro-ph.IM↗