arXiv subjects

Yifan Jia

Publications and source records attributed to Yifan Jia.

At least 19 recordsLinked to original sources

arXiv2026

Bridging Source Code and Bytecode for Smart Contract Vulnerability Detection via Dual-Perspective Cross-Modal Distillation

Smart contract vulnerabilities have caused substantial financial losses, yet most deployed contracts are closed-source, forcing detection to operate on bytecode --- which lacks the high-level semantic information available in source code. To compensate, recent cross-modal methods distill knowledge from source-code models into bytecode detectors by aligning the two modalities. However, these methods align source and bytecode only at the graph level through global embedding matching, whereas vulnerability is a property of specific nodes and control-flow regions --- so the student preserves global structure but loses the fine-grained node-level correspondences that decide a contract's safety. We propose ExDoS, a dual-focus cross-modal distillation framework that addresses this limitation from three angles. First, We introduce aligned vulnerability patterns that mark corresponding nodes in both source and bytecode graphs, establishing the missing node-level supervision and cross-modal correspondence; with these correspondences in place, we then propose a dual-attention graph network that applies relation-aware attention and adaptive node weighting so that vulnerable nodes survive into the graph-level embedding rather than being diluted by uniform aggregation; given preserved node-level signals and established correspondences, we also propose a dual-focus distillation objective whose global loss retains whole-graph alignment while its local loss matches expert-paired nodes to enforce region-level consistency. On real-world contracts, ExDoS reaches F1 of 90.86%, 90.23%, and 83.94% for reentrancy, timestamp dependency, and infinite loop --- improving by 2.7--5.1 points over the strongest per-type baseline. Our ablations confirm that the pattern annotations, local alignment, and attentive encoding each address a distinct limitation of prior distillation-based approaches.

cs.CR↗

arXiv2026

Lossless Tensor Compression as Program Synthesis

Model checkpoints are growing in both number and size, which makes archival, transfer, and deployment increasingly costly. General-purpose compressors can reduce storage requirements but ignore tensor structure, whereas existing tensor-specific compressors rely on fixed and format-specific pipelines. We present Brevis, which formulates lossless tensor compression as program synthesis. We design a typed domain-specific language (DSL) that captures recurring tensor structures, such as repeated regions and floating-point fields, through a set of reversible operators. Given a tensor, Brevis synthesizes a self-contained DSL program that reconstructs it bit-exactly. A checkpoint-specific production prior, learned from a small representative sample of tensors, guides a bounded A* search to synthesize compact programs, which can later be executed directly for bit-exact decompression. On 10 public checkpoints spanning language, audio, and image generation models, Brevis reduces 2.13 TB of checkpoint data to 1.41 TB, a 33.93% storage reduction. It produces archives up to 30.87% smaller than those of four general-purpose compressors, including zstd and gzip, and smaller archives than the tensor-specific compressors ZipNN and DFloat11. Under a practical concurrency configuration, Brevis achieves 3.60 GB/s compression and 6.61 GB/s decompression while preserving every source byte.

cs.SE↗

arXiv2026

Almost free algebras: from the word problem to elimination of quantifiers

Term algebras are important objects in computer science and are correspondingly well-studied. A natural generalization is to quotient these algebras by finitely many ground term equations, obtaining what we call almost free algebras. One of the earliest results on almost free algebras is that their word problem is polynomial time decidable. In this paper, we show that other natural problems: finding canonical representatives; computing the cardinality of a congruence class; checking if all congruence classes are infinite; checking if the algebra is finite; checking if two algebras are isomorphic, are all polynomial time decidable. Another famous result regarding term algebras is that they admit quantifier elimination in a suitably expanded language. Following this pattern, we also show that almost free algebras admit quantifier elimination by expanding the language with the standard tester predicates. While this is implied by existing results, we view our main contribution here as providing a different approach, which we posit can be easily extended to a larger class that is not covered by existing works. Finally, we provide an application to the quantifier elimination procedure, constructing examples of non-initial algebras over arbitrary signatures with a polynomial time word problem.

math.LO↗

arXiv2026

Model theory of term algebras revisited

Building on work of Maltsev on locally free algebras in finite purely functional languages, we revisit the model theory of (absolutely free) term algebras and their completions. Maltsev's analysis yields a natural axiomatization together with quantifier elimination to positive Boolean combinations of special formulas, and shows that the complete extensions are parametrized exactly by the number $k\in\{0,1,\dots,ω\}$ of indecomposable elements; for $1\le k\leω$ the standard model is the free term algebra on $k$ generators. We give a new, quantifier-elimination--free proof of completeness using Ehrenfeucht--Fraïssé games, and we establish several further structural properties of the standard models and theories. In particular, for $1\le k\leω$ we prove first-order rigidity and atomicity of the standard model. For every $0\le k\leω$ we show that the corresponding theory does not have the finite cover property and weakly eliminates imaginaries. We also provide new proofs of stability-theoretic features previously obtained by Belegradek: the theories are stable but not superstable, normal (hence $1$-based), and have trivial forking; consequently, no infinite group is interpretable in any model. Finally, we analyze model completeness and show that $T_0$ is the model companion of the theory of locally free algebras, while the theories with $k\ge 1$ are not model complete.

math.LO↗

arXiv2026

TeachPro: Multi-Label Qualitative Teaching Evaluation via Cross-View Graph Synergy and Semantic Anchored Evidence Encoding

Standardized Student Evaluation of Teaching often suffer from low reliability, restricted response options, and response distortion. Existing machine learning methods that mine open-ended comments usually reduce feedback to binary sentiment, which overlooks concrete concerns such as content clarity, feedback timeliness, and instructor demeanor, and provides limited guidance for instructional improvement.We propose TeachPro, a multi-label learning framework that systematically assesses five key teaching dimensions: professional expertise, instructional behavior, pedagogical efficacy, classroom experience, and other performance metrics. We first propose a Dimension-Anchored Evidence Encoder, which integrates three core components: (i) a pre-trained text encoder that transforms qualitative feedback annotations into contextualized embeddings; (ii) a prompt module that represents five teaching dimensions as learnable semantic anchors; and (iii) a cross-attention mechanism that aligns evidence with pedagogical dimensions within a structured semantic space. We then propose a Cross-View Graph Synergy Network to represent student comments. This network comprises two components: (i) a Syntactic Branch that extracts explicit grammatical dependencies from parse trees, and (ii) a Semantic Branch that models latent conceptual relations derived from BERT-based similarity graphs. BiAffine fusion module aligns syntactic and semantic units, while a differential regularizer disentangles embeddings to encourage complementary representations. Finally, a cross-attention mechanism bridges the dimension-anchored evidence with the multi-view comment representations. We also contribute a novel benchmark dataset featuring expert qualitative annotations and multi-label scores. Extensive experiments demonstrate that TeachPro offers superior diagnostic granularity and robustness across diverse evaluation settings.

cs.CL↗

arXiv2025

URL2Graph++: Unified Semantic-Structural-Character Learning for Malicious URL Detection

Malicious URL detection remains a major challenge in cybersecurity, primarily due to two factors: (1) the exponential growth of the Internet has led to an immense diversity of URLs, making generalized detection increasingly difficult; and (2) attackers are increasingly employing sophisticated obfuscation techniques to evade detection. We advocate that addressing these challenges fundamentally requires: (1) obtaining semantic understanding to improve generalization across vast and diverse URL sets, and (2) accurately modeling contextual relationships within the structural composition of URLs. In this paper, we propose a novel malicious URL detection method combining multi-granularity graph learning with semantic embedding to jointly capture semantic, character-level, and structural features for robust URL analysis. To model internal dependencies within URLs, we first construct dual-granularity URL graphs at both subword and character levels, where nodes represent URL tokens/characters and edges encode co-occurrence relationships. To obtain fine-grained embeddings, we initialize node representations using a character-level convolutional network. The two graphs are then processed through jointly trained Graph Convolutional Networks to learn consistent graph-level representations, enabling the model to capture complementary structural features that reflect co-occurrence patterns and character-level dependencies. Furthermore, we employ BERT to derive semantic representations of URLs for semantically aware understanding. Finally, we introduce a gated dynamic fusion network to combine the semantically enriched BERT representations with the jointly optimized graph vectors, further enhancing detection performance. We extensively evaluate our method across multiple challenging dimensions. Results show our method exceeds SOTA performance, including against large language models.

cs.CR↗

arXiv2025

KGBERT4Eth: A Feature-Complete Transformer Powered by Knowledge Graph for Multi-Task Ethereum Fraud Detection

Ethereum's rapid ecosystem expansion and transaction anonymity have triggered a surge in malicious activity. Detection mechanisms currently bifurcate into three technical strands: expert-defined features, graph embeddings, and sequential transaction patterns, collectively spanning the complete feature sets of Ethereum's native data layer. Yet the absence of cross-paradigm integration mechanisms forces practitioners to choose between sacrificing sequential context awareness, structured fund-flow patterns, or human-curated feature insights in their solutions. To bridge this gap, we propose KGBERT4Eth, a feature-complete pre-training encoder that synergistically combines two key components: (1) a Transaction Semantic Extractor, where we train an enhanced Transaction Language Model (TLM) to learn contextual semantic representations from conceptualized transaction records, and (2) a Transaction Knowledge Graph (TKG) that incorporates expert-curated domain knowledge into graph node embeddings to capture fund flow patterns and human-curated feature insights. We jointly optimize pre-training objectives for both components to fuse these complementary features, generating feature-complete embeddings. To emphasize rare anomalous transactions, we design a biased masking prediction task for TLM to focus on statistical outliers, while the Transaction TKG employs link prediction to learn latent transaction relationships and aggregate knowledge. Furthermore, we propose a mask-invariant attention coordination module to ensure stable dynamic information exchange between TLM and TKG during pre-training. KGBERT4Eth significantly outperforms state-of-the-art baselines in both phishing account detection and de-anonymization tasks, achieving absolute F1-score improvements of 8-16% on three phishing detection benchmarks and 6-26% on four de-anonymization datasets.

cs.CR↗

arXiv2025

LMAE4Eth: Generalizable and Robust Ethereum Fraud Detection by Exploring Transaction Semantics and Masked Graph Embedding

Current Ethereum fraud detection methods rely on context-independent, numerical transaction sequences, failing to capture semantic of account transactions. Furthermore, the pervasive homogeneity in Ethereum transaction records renders it challenging to learn discriminative account embeddings. Moreover, current self-supervised graph learning methods primarily learn node representations through graph reconstruction, resulting in suboptimal performance for node-level tasks like fraud account detection, while these methods also encounter scalability challenges. To tackle these challenges, we propose LMAE4Eth, a multi-view learning framework that fuses transaction semantics, masked graph embedding, and expert knowledge. We first propose a transaction-token contrastive language model (TxCLM) that transforms context-independent numerical transaction records into logically cohesive linguistic representations. To clearly characterize the semantic differences between accounts, we also use a token-aware contrastive learning pre-training objective together with the masked transaction model pre-training objective, learns high-expressive account representations. We then propose a masked account graph autoencoder (MAGAE) using generative self-supervised learning, which achieves superior node-level account detection by focusing on reconstructing account node features. To enable MAGAE to scale for large-scale training, we propose to integrate layer-neighbor sampling into the graph, which reduces the number of sampled vertices by several times without compromising training quality. Finally, using a cross-attention fusion network, we unify the embeddings of TxCLM and MAGAE to leverage the benefits of both. We evaluate our method against 21 baseline approaches on three datasets. Experimental results show that our method outperforms the best baseline by over 10% in F1-score on two of the datasets.

cs.CR↗

arXiv2025

Attributing Data for Sharpness-Aware Minimization

Sharpness-aware Minimization (SAM) improves generalization in large-scale model training by linking loss landscape geometry to generalization. However, challenges such as mislabeled noisy data and privacy concerns have emerged as significant issues. Data attribution, which identifies the contributions of specific training samples, offers a promising solution. However, directly rendering existing data influence evaluation tools such as influence functions (IF) to SAM will be inapplicable or inaccurate as SAM utilizes an inner loop to find model perturbations that maximize loss, which the outer loop then minimizes, resulting in a doubled computational structure. Additionally, this bilevel structure complicates the modeling of data influence on the parameters. In this paper, based on the IF, we develop two innovative data valuation methods for SAM, each offering unique benefits in different scenarios: the Hessian-based IF and the Gradient Trajectory-based IF. The first one provides a comprehensive estimation of data influence using a closed-form measure that relies only on the trained model weights. In contrast, the other IF for SAM utilizes gradient trajectory information during training for more accurate and efficient data assessment. Extensive experiments demonstrate their effectiveness in data evaluation and parameter tuning, with applications in identifying mislabeled data, model editing, and enhancing interpretability.

cs.LG↗

arXiv2025

DiT-SGCR: Directed Temporal Structural Representation with Global-Cluster Awareness for Ethereum Malicious Account Detection

The detection of malicious accounts on Ethereum - the preeminent DeFi platform - is critical for protecting digital assets and maintaining trust in decentralized finance. Recent advances highlight that temporal transaction evolution reveals more attack signatures than static graphs. However, current methods either fail to model continuous transaction dynamics or incur high computational costs that limit scalability to large-scale transaction networks. Furthermore, current methods fail to consider two higher-order behavioral fingerprints: (1) direction in temporal transaction flows, which encodes money movement trajectories, and (2) account clustering, which reveals coordinated behavior of organized malicious collectives. To address these challenges, we propose DiT-SGCR, an unsupervised graph encoder for malicious account detection. Specifically, DiT-SGCR employs directional temporal aggregation to capture dynamic account interactions, then coupled with differentiable clustering and graph Laplacian regularization to generate high-quality, low-dimensional embeddings. Our approach simultaneously encodes directional temporal dynamics, global topology, and cluster-specific behavioral patterns, thereby enhancing the discriminability and robustness of account representations. Furthermore, DiT-SGCR bypasses conventional graph propagation mechanisms, yielding significant scalability advantages. Extensive experiments on three datasets demonstrate that DiT-SGCR consistently outperforms state-of-the-art methods across all benchmarks, achieving F1-score improvements ranging from 3.62% to 10.83%.

cs.CE↗

arXiv2025

WebGuard++:Interpretable Malicious URL Detection via Bidirectional Fusion of HTML Subgraphs and Multi-Scale Convolutional BERT

URL+HTML feature fusion shows promise for robust malicious URL detection, since attacker artifacts persist in DOM structures. However, prior work suffers from four critical shortcomings: (1) incomplete URL modeling, failing to jointly capture lexical patterns and semantic context; (2) HTML graph sparsity, where threat-indicative nodes (e.g., obfuscated scripts) are isolated amid benign content, causing signal dilution during graph aggregation; (3) unidirectional analysis, ignoring URL-HTML feature bidirectional interaction; and (4) opaque decisions, lacking attribution to malicious DOM components. To address these challenges, we present WebGuard++, a detection framework with 4 novel components: 1) Cross-scale URL Encoder: Hierarchically learns local-to-global and coarse to fine URL features based on Transformer network with dynamic convolution. 2) Subgraph-aware HTML Encoder: Decomposes DOM graphs into interpretable substructures, amplifying sparse threat signals via Hierarchical feature fusion. 3) Bidirectional Coupling Module: Aligns URL and HTML embeddings through cross-modal contrastive learning, optimizing inter-modal consistency and intra-modal specificity. 4) Voting Module: Localizes malicious regions through consensus voting on malicious subgraph predictions. Experiments show WebGuard++ achieves significant improvements over state-of-the-art baselines, achieving 1.1x-7.9x higher TPR at fixed FPR of 0.001 and 0.0001 across both datasets.

cs.CR↗

arXiv2025

Interpreting Social Bias in LVLMs via Information Flow Analysis and Multi-Round Dialogue Evaluation

Large Vision Language Models (LVLMs) have achieved remarkable progress in multimodal tasks, yet they also exhibit notable social biases. These biases often manifest as unintended associations between neutral concepts and sensitive human attributes, leading to disparate model behaviors across demographic groups. While existing studies primarily focus on detecting and quantifying such biases, they offer limited insight into the underlying mechanisms within the models. To address this gap, we propose an explanatory framework that combines information flow analysis with multi-round dialogue evaluation, aiming to understand the origin of social bias from the perspective of imbalanced internal information utilization. Specifically, we first identify high-contribution image tokens involved in the model's reasoning process for neutral questions via information flow analysis. Then, we design a multi-turn dialogue mechanism to evaluate the extent to which these key tokens encode sensitive information. Extensive experiments reveal that LVLMs exhibit systematic disparities in information usage when processing images of different demographic groups, suggesting that social bias is deeply rooted in the model's internal reasoning dynamics. Furthermore, we complement our findings from a textual modality perspective, showing that the model's semantic representations already display biased proximity patterns, thereby offering a cross-modal explanation of bias formation.

cs.AI↗

arXiv2025

Benchmarking Multimodal Knowledge Conflict for Large Multimodal Models

Large Multimodal Models(LMMs) face notable challenges when encountering multimodal knowledge conflicts, particularly under retrieval-augmented generation(RAG) frameworks where the contextual information from external sources may contradict the model's internal parametric knowledge, leading to unreliable outputs. However, existing benchmarks fail to reflect such realistic conflict scenarios. Most focus solely on intra-memory conflicts, while context-memory and inter-context conflicts remain largely investigated. Furthermore, commonly used factual knowledge-based evaluations are often overlooked, and existing datasets lack a thorough investigation into conflict detection capabilities. To bridge this gap, we propose MMKC-Bench, a benchmark designed to evaluate factual knowledge conflicts in both context-memory and inter-context scenarios. MMKC-Bench encompasses three types of multimodal knowledge conflicts and includes 1,573 knowledge instances and 3,381 images across 23 broad types, collected through automated pipelines with human verification. We evaluate three representative series of LMMs on both model behavior analysis and conflict detection tasks. Our findings show that while current LMMs are capable of recognizing knowledge conflicts, they tend to favor internal parametric knowledge over external evidence. We hope MMKC-Bench will foster further research in multimodal knowledge conflict and enhance the development of multimodal RAG systems. The source code is available at https://github.com/MLLMKCBENCH/MLLMKC.

cs.LG↗

arXiv2025

Continuous Multi-Task Pre-training for Malicious URL Detection and Webpage Classification

Malicious URL detection and webpage classification are critical tasks in cybersecurity and information management. In recent years, extensive research has explored using BERT or similar language models to replace traditional machine learning methods for detecting malicious URLs and classifying webpages. While previous studies show promising results, they often apply existing language models to these tasks without accounting for the inherent differences in domain data (e.g., URLs being loosely structured and semantically sparse compared to text), leaving room for performance improvement. Furthermore, current approaches focus on single tasks and have not been tested in multi-task scenarios. To address these challenges, we propose urlBERT, a pre-trained URL encoder leveraging Transformer to encode foundational knowledge from billions of unlabeled URLs. To achieve it, we propose to use 5 unsupervised pretraining tasks to capture multi-level information of URL lexical, syntax, and semantics, and generate contrastive and adversarial representations. Furthermore, to avoid inter-pre-training competition and interference, we proposed a grouped sequential learning method to ensure effective training across multi-tasks. Finally, we leverage a two-stage fine-tuning approach to improve the training stability and efficiency of the task model. To assess the multitasking potential of urlBERT, we fine-tune the task model in both single-task and multi-task modes. The former creates a classification model for a single task, while the latter builds a classification model capable of handling multiple tasks. We evaluate urlBERT on three downstream tasks: phishing URL detection, advertising URL detection, and webpage classification. The results demonstrate that urlBERT outperforms standard pre-trained models, and its multi-task mode is capable of addressing the real-world demands of multitasking.

cs.CR↗

arXiv2025

Accelerating Diffusion Transformer via Increment-Calibrated Caching with Channel-Aware Singular Value Decomposition

Diffusion transformer (DiT) models have achieved remarkable success in image generation, thanks for their exceptional generative capabilities and scalability. Nonetheless, the iterative nature of diffusion models (DMs) results in high computation complexity, posing challenges for deployment. Although existing cache-based acceleration methods try to utilize the inherent temporal similarity to skip redundant computations of DiT, the lack of correction may induce potential quality degradation. In this paper, we propose increment-calibrated caching, a training-free method for DiT acceleration, where the calibration parameters are generated from the pre-trained model itself with low-rank approximation. To deal with the possible correction failure arising from outlier activations, we introduce channel-aware Singular Value Decomposition (SVD), which further strengthens the calibration effect. Experimental results show that our method always achieve better performance than existing naive caching methods with a similar computation resource budget. When compared with 35-step DDIM, our method eliminates more than 45% computation and improves IS by 12 at the cost of less than 0.06 FID increase. Code is available at https://github.com/ccccczzy/icc.

cs.CV↗

arXiv2025

BioD2C: A Dual-level Semantic Consistency Constraint Framework for Biomedical VQA

Biomedical visual question answering (VQA) has been widely studied and has demonstrated significant application value and potential in fields such as assistive medical diagnosis. Despite their success, current biomedical VQA models perform multimodal information interaction only at the model level within large language models (LLMs), leading to suboptimal multimodal semantic alignment when dealing with complex tasks. To address this issue, we propose BioD2C: a novel Dual-level Semantic Consistency Constraint Framework for Biomedical VQA, which achieves dual-level semantic interaction alignment at both the model and feature levels, enabling the model to adaptively learn visual features based on the question. Specifically, we firstly integrate textual features into visual features via an image-text fusion mechanism as feature-level semantic interaction, obtaining visual features conditioned on the given text; and then introduce a text-queue-based cross-modal soft semantic loss function to further align the image semantics with the question semantics. Specifically, in this work, we establish a new dataset, BioVGQ, to address inherent biases in prior datasets by filtering manually-altered images and aligning question-answer pairs with multimodal context, and train our model on this dataset. Extensive experimental results demonstrate that BioD2C achieves state-of-the-art (SOTA) performance across multiple downstream datasets, showcasing its robustness, generalizability, and potential to advance biomedical VQA research.

cs.CV↗

arXiv2025

Adaptive H&E-IHC information fusion staining framework based on feature extra

Immunohistochemistry (IHC) staining plays a significant role in the evaluation of diseases such as breast cancer. The H&E-to-IHC transformation based on generative models provides a simple and cost-effective method for obtaining IHC images. Although previous models can perform digital coloring well, they still suffer from (i) coloring only through the pixel features that are not prominent in HE, which is easy to cause information loss in the coloring process; (ii) The lack of pixel-perfect H&E-IHC groundtruth pairs poses a challenge to the classical L1 loss.To address the above challenges, we propose an adaptive information enhanced coloring framework based on feature extractors. We first propose the VMFE module to effectively extract the color information features using multi-scale feature extraction and wavelet transform convolution, while combining the shared decoder for feature fusion. The high-performance dual feature extractor of H&E-IHC is trained by contrastive learning, which can effectively perform feature alignment of HE-IHC in high latitude space. At the same time, the trained feature encoder is used to enhance the features and adaptively adjust the loss in the HE section staining process to solve the problems related to unclear and asymmetric information. We have tested on different datasets and achieved excellent performance.Our code is available at https://github.com/babyinsunshine/CEFF

cs.CV↗

arXiv2025

Ethereum Fraud Detection via Joint Transaction Language Model and Graph Representation Learning

Ethereum faces growing fraud threats. Current fraud detection methods, whether employing graph neural networks or sequence models, fail to consider the semantic information and similarity patterns within transactions. Moreover, these approaches do not leverage the potential synergistic benefits of combining both types of models. To address these challenges, we propose TLMG4Eth that combines a transaction language model with graph-based methods to capture semantic, similarity, and structural features of transaction data in Ethereum. We first propose a transaction language model that converts numerical transaction data into meaningful transaction sentences, enabling the model to learn explicit transaction semantics. Then, we propose a transaction attribute similarity graph to learn transaction similarity information, enabling us to capture intuitive insights into transaction anomalies. Additionally, we construct an account interaction graph to capture the structural information of the account transaction network. We employ a deep multi-head attention network to fuse transaction semantic and similarity embeddings, and ultimately propose a joint training approach for the multi-head attention network and the account interaction graph to obtain the synergistic benefits of both.

cs.CR↗