Search arXiv⌕ Search

arXiv subjects

Chen Gong

Publications and source records attributed to Chen Gong.

At least 19 recordsLinked to original sources
arXiv2026

PrivCode: When Code Generation Meets Differential Privacy

Large language models (LLMs) have presented outstanding performance in code generation and completion. However, fine-tuning these models on private datasets can raise privacy and proprietary concerns, such as the leakage of sensitive personal information. Differentially private (DP) code generation provides theoretical guarantees for protecting sensitive code by generating synthetic datasets that preserve statistical properties while reducing privacy leakage concerns. However, DP code generation faces significant challenges due to the strict syntactic dependencies and the privacy-utility trade-off. We propose PrivCode, the first DP synthesizer specifically designed for code datasets. It incorporates a two-stage framework to improve both privacy and utility. In the first stage, termed "privacy-sanitizing", PrivCode generates DP-compliant synthetic code by training models using DP-SGD while introducing syntactic information to preserve code structure. The second stage, termed "utility-boosting", fine-tunes a larger pre-trained LLM on the synthetic privacy-free code to mitigate the utility loss caused by DP, enhancing the utility of the generated code. Extensive experiments on four LLMs show that PrivCode generates higher-utility code across various testing tasks under four benchmarks. The experiments also confirm its ability to protect sensitive data under varying privacy budgets. We provide the replication package at the code repository: https://github.com/Liuzzyg/PrivCode.

cs.CR↗
arXiv2026

UniGFM-RAG: Unify Graph Foundation Models and Large Language Models for End-to-End Retrieval and Reasoning

Large Language Models (LLMs) have shown remarkable capabilities across tasks, but they still struggle with complex reasoning that requires integrating scattered knowledge. Graph Retrieval-Augmented Generation (GraphRAG) addresses this challenge by organizing fragment knowledge as graphs, while recent Graph Foundation Models (GFMs) further improve GraphRAG with graph-native retrieval and reasoning capabilities. However, existing GraphRAG methods usually adopt decoupled retrieve-then-generate pipelines, which create objective misalignment between retrieval and reasoning, and training GFM retrievers typically depends on expensive human annotations. To address these limitations, we propose \ourmethod, a unified end-to-end GraphRAG framework that tightly couples a GFM retriever with an LLM generator. Given a query, the GFM retriever first reasons over the graph to score and retrieve relevant document nodes. These node embeddings are then passed through a fully differentiable graph bridge, which projects them into soft-weighted graph prefixes according to retrieval scores and combines them with raw document texts to form graph-augmented prompts. In this way, structural information is injected directly into LLM generation, while the final answer loss can be back-propagated through the bridge to jointly optimize retrieval and reasoning. To stabilize training and reduce reliance on costly annotations, we further introduce a semantic-guided distillation module that provides auxiliary supervision using pseudo-labels derived from query-document semantic similarities. Extensive experiments on three multi-hop QA datasets and three out-of-domain GraphRAG benchmarks show that \ourmethod consistently outperforms strong baseline methods and generalizes well, which demonstrates the effectiveness of end-to-end GraphRAG with tightly integrated GFM retrieval and LLM reasoning.

cs.AI↗
arXiv2026

Demystifying Prompt Stealing Attacks against Synthetic Video

Creators publicly showcase generated content to advertise and sell their work, but these examples may also enable unauthorized reproduction and customization. We study whether an adversary can derive reusable inputs from a single public video for faithful reconstruction and controlled editing, without access to the original prompt or generator. We adapt captioning and prompt optimization approaches to video prompt stealing attacks (PSAs) and find that prompt recovery alone provides limited reconstruction fidelity. Even the original prompt produces different spatial and temporal realizations when the initial noise changes. Using an independent generator enlarges the reconstruction gap, as reusing the original prompt and seed can still produce substantially different videos. We propose V-PSA, which uses a recovered caption and DDIM inversion to construct a reference trajectory under an independent generator. Null-Text Inversion then optimizes per-step unconditional embeddings to reduce deviations from this trajectory during reconstruction. The resulting inputs also guide generation when the caption is modified, supporting edits that change selected content while preserving the target's layout and motion. We construct a benchmark containing 12,000 video-prompt pairs from three victim generators and four prompt sources. V-PSA achieves video-to-video similarity around 0.9 between target and reconstructed videos, substantially improving over prompt-only baselines with similarity score of 0.6, and supports multiple controlled variations that preserve the target's structure. Additional experiments show that augmenting prior image PSAs with inversion-derived inputs also improves reconstruction. Our results show that evaluations restricted to prompt recovery can underestimate an attacker's ability to reconstruct public generative outputs.

cs.CV↗
arXiv2026

Out of the box age estimation through facial imagery: A Comprehensive Benchmark of Vision-Language Models vs. out-of-the-box Traditional Architectures

Facial age estimation plays a critical role in content moderation, age verification, and deepfake detection. However, no prior benchmark has systematically compared modern vision-language models (VLMs) with specialized age estimation architectures. We present the first large-scale cross-paradigm benchmark, evaluating 34 models - 22 specialized architectures with publicly available pretrained weights and 12 general-purpose VLMs - across eight standard datasets (UTKFace, IMDB-WIKI, MORPH, AFAD, CACD, FG-NET, APPA-REAL, and AgeDB), totaling 1,100 test images per model. Our key finding is striking: zero-shot VLMs significantly outperform most specialized models, achieving an average mean absolute error (MAE) of 5.65 years compared to 9.88 years for non-LLM models. The best-performing VLM (Gemini 3 Flash Preview, MAE 4.32) surpasses the strongest non-LLM model (MiVOLO, MAE 5.10) by 15%. MiVOLO - unique in combining face and body features using Vision Transformers - is the only specialized model that remains competitive with VLMs. We further analyze age verification at the 18-year threshold and find that most non-LLM models exhibit false adult rates between 39% and 100% for minors, whereas VLMs reduce this to 16%-29%. Additionally, coarse age binning (8-9 classes) consistently increases MAE beyond 13 years. Stratified analysis across 14 age groups reveals that all models struggle most at extreme ages (under 5 and over 65). Overall, these findings challenge the assumption that task-specific architectures are necessary for high-performance age estimation and suggest that future work should focus on distilling VLM capabilities into efficient specialized models.

cs.CV↗
arXiv2026

Attention Transfer Is Not Universally Effective for Vision Transformers

A recent work shows that Attention Transfer, which transfers only the attention patterns from a pre-trained teacher Vision Transformer (ViT) to a randomly initialized standard student ViT, is sufficient to recover the full benefit of the teacher's pre-trained weights. We revisit this finding on a comprehensive benchmark of 20 teachers from 11 well-known ViT families and reveal that Attention Transfer is not universally effective. While 7 families transfer successfully, 4 consistently fail, falling up to 5.1\% below the from-scratch no-transfer baseline. Further results demonstrate that this failure is family-consistent across model sizes, and persists under extended training durations, different transfer datasets, and out-of-distribution evaluations. Controlled analyses then consistently localize the problem to the attention-routing channel, indicating that the key issue is not whether the student can match the teacher's attention patterns, but whether the matched patterns remain functional for the student. Crucially, we identify architectural mismatch between the pre-trained teacher and the standard student as the primary mechanism. By adding only the teacher's native architectural components to the student in a randomly initialized state, we completely reverse the failure for all 4 families. Notably, these components alone do not improve from-scratch training, confirming that they specifically unlock the usability of the teacher's attention. We further show that this failure is not explained by the inadequate choice of transfer loss, differences in pre-training recipes, or the intrinsic quality of attention maps. Our findings refine the prevailing understanding of attention in ViT representations: attention is sufficient \textit{only} when the student architecture matches the teacher.

cs.CV↗
arXiv2026

STLQ: Spatio-Topological Linkage Quantization for Approximate Nearest Neighbor Search

Vector quantization is widely used for large-scale data compression and approximate nearest neighbor search (ANNS). Multi-codebook quantization (MCQ) is a practical paradigm that represents database vectors with compact codes and supports efficient retrieval through asymmetric distance computation (ADC). However, existing shallow MCQs encode vectors independently within each inverted file (IVF) cluster, without exploiting local references. Therefore, we propose Spatio-Topological Linkage Quantization (STLQ), a topological shallow MCQ method that organizes each IVF cluster as a directed-forest with compact codes and scaling coefficients. Its basic encoding phase first quantizes vectors independently, while the linked encoding phase re-encodes relative residuals with respect to selected reconstructed parents, reusing local reconstruction information to obtain finer representations. The corresponding linked-ADC preserves linear scanning complexity. STLQ further compresses the scaling coefficients and directed-forest structure to control the additional topology overhead. Extensive experiments show that STLQ achieves higher recall than the evaluated shallow MCQs at comparable bit budgets, although at a higher offline construction cost. Meanwhile, it has a competitive recall-QPS tradeoff against deep MCQs. The code is available at https://github.com/ighxiy/STLQ.

cs.CV↗
arXiv2026

Harmony in Diversity: Multi-domain Contrastive Policy Optimization for Large Reasoning Models

Post-training via Reinforcement Learning (RL) has enabled Large Reasoning Models (LRMs) to achieve strong performance in individual domain. However, real-world applications increasingly require general-purpose reasoners rendering strong performance across diverse domains. Mixed-domain post-training aims to achieve this goal by jointly training with mixed domain data, but this often induces capability compromise and degradation among different domains. Existing methods attribute this performance degradation to harmful cross-domain interactions and propose various strategies to mitigate them, but these strategies may also impede beneficial knowledge sharing across domains and in turn fail to match or surpass single-domain performance. To address this problem, we propose \textbf{M}ulti-domain \textbf{C}ontrastive \textbf{P}olicy \textbf{O}ptimization (MCPO), which uses contrastive learning to utilize both positive and negative cross-domain interactions for knowledge sharing and competition. Specifically, we partition each rollout generated by LRMs according to its underlying reasoning structures and use reasoning segments to capture these structures. We thus formulate positive and negative pairs of reasoning segments as mutually augmented examples, which provide supportive and competing signals for knowledge sharing. Subsequently, we design complementary contrastive objectives for cross-domain knowledge sharing and intra-domain knowledge consolidation, targeting compatibility across domains and discriminability within each domain to form a harmonious reasoning space. Experimental results across a broad range of domains show that MCPO alleviates performance degradation caused by mixed-domain training and outperforms single-domain training in most cases.

cs.CL↗
arXiv2026

Copy the Same, Distill the Difference: Initializing Linear Vision Transformers

Linear Vision Transformers (ViTs) are designed to replace the attention in Softmax ViTs with the linear-complexity attention operator for more efficient token routing, but they require from-scratch pre-training and typically underperform the original Softmax version. How to initialize linear ViTs both efficiently and effectively still remains unclear. In this work, we explicitly ask: given that most foundation ViTs are built on the mainstream Softmax attention, can linear ViTs benefit from their pre-trained weights? Recent works on Attention Transfer show that attention is the effective transferable component between Softmax ViTs, suggesting attention alone suffices for such reuse. However, we find the opposite for Softmax-to-linear transfer. The attention weights are operator-specific: copying them barely helps, and is sometimes even worse than random initialization. Instead, the attention's token routing behavior can be recovered through distillation with a proper loss design, letting linear ViTs reduce the gap and even match Softmax ones. In contrast, the MLP weights, which carry the learned representation, are operator-agnostic: they can be transferred by simple direct copying, which already carries most of the benefit of the pre-trained weights. Thus, copying MLPs can serve as an effective foundation for Softmax-to-linear transfer: paired with the distilled attention, linear ViTs eventually close the remaining gap and even surpass Softmax ones. These findings hold consistently across various linear ViT variants, different model sizes, and diverse datasets. We hope this study deepens the understanding of reusing pre-trained weights across attention operators: copy what stays the same and distill what differs, to recover the benefit across the Softmax-to-linear boundary.

cs.CV↗
arXiv2026

HyperReCo: Retrieving and Connecting Evidence with Hypergraph Neural Networks for LLM Multi-hop Reasoning

Large language models (LLMs) have shown strong capabilities, with retrieval-augmented generation (RAG) supporting complex multi-hop reasoning by retrieving evidence distributed across documents. Graph-based approaches exploit connections among evidence, and hypergraph-based retrieval further preserves higher-order entity associations within documents and connects documents through shared entities. However, existing hypergraph retrievers often rely on predefined structural expansion or diffusion, which may miss query-dependent interactions needed to identify relevant evidence. They also leave connections among retrieved evidence implicit, requiring LLMs to reconstruct these connections before reasoning. Therefore, we propose HyperReCo, a framework for retrieving and connecting evidence with a hypergraph neural network (HyperGNN). We represent each document as a hyperedge over its extracted entities, with shared entities connecting the hyperedges. Through hypergraph message passing with joint supervision over documents and entities, the HyperGNN learns query-dependent interactions to retrieve complementary evidence. We further introduce Gradient-Guided Hyper-Path Decoding (GGHD), which uses gradient attribution to interpret the learned interactions and translate them into explicit hyper-paths that help LLMs combine complementary facts for multi-hop reasoning. Experiments on six benchmarks show that HyperReCo achieves the best retrieval performance among the compared methods on all three multi-hop QA datasets, together with strong downstream QA performance. Case studies and further analyses demonstrate the utility of decoded hyper-paths for connecting retrieved evidence.

cs.AI↗
arXiv2026

ForgetMimic: Motion Unlearning for Reinforcement Learning Humanoid Control

Humanoid control, leveraging human demonstrations, has achieved diverse, agile, and natural locomotion behaviors through reinforcement learning (RL). While this paradigm has yielded remarkable performance in physical humanoid control, how to eliminate specific motions from learned policies remains insufficiently explored. Addressing this issue is motivated by pressing safety and privacy concerns: the removal of malicious, poisoned, or suboptimal motions, as well as copyright-protected motions subject to the right to be forgotten under regulations such as the GDPR, is of critical importance. To this end, we propose {ForgetMimic}, the first motion-level unlearning method designed specifically for physical-world humanoid control. The core idea of ForgetMimic is as follows: given a policy $π_θ$ trained on $N$ motions, our method degrades performance on a target subset of $K$ motions while preserving the effectiveness of the remaining $N-K$ motions. Furthermore, we identify and resolve two key training mechanisms in robot control that lead to unlearning failure. We conduct extensive experiments on the Unitree G1 and H2 humanoid robots across 12 motions, including Dance, Fight, Flip, and others. Experimental results demonstrate that ForgetMimic effectively eliminates memory of designated motions while maintaining the normal operation of all other motions.

cs.RO↗
arXiv2026

Beyond One-Size-Fits-All: Neural Networks for Differentially Private Tabular Data Synthesis

In differentially private (DP) tabular data synthesis, the consensus is that statistical models are better than neural network (NN)-based methods. However, we argue that this conclusion is incomplete and overlooks the challenge of densely correlated datasets, where intricate dependencies can overwhelm statistical models. In such complex scenarios, neural networks are more suitable due to their capacity to fit complex distributions by learning directly from samples. Despite this potential, existing NN-based algorithms still suffer from significant limitations. We therefore propose MargNet, incorporating successful algorithmic designs of statistical models into neural networks. MargNet applies an adaptive marginal selection strategy and trains the neural networks to generate data that conforms to the selected marginals. On sparsely correlated datasets, our approach achieves utility close to the best statistical method while offering an average 7$\times$ speedup over it. More importantly, on densely correlated datasets, MargNet establishes a new state-of-the-art, reducing fidelity error by up to 26\% compared to the previous best. We release our code on GitHub.\footnote{https://github.com/KaiChen9909/margnet}

cs.LG↗
arXiv2026

VI-Bench: Benchmarking Prompt Inversion from AIGC Videos

Recent advances in video generation have made prompt-based control increasingly central to AIGC video generation. Prompts specify what a video should depict and how it should be represented, controlling factors such as visual style or camera behavior. Understanding this recoverability is important both for creative reuse and editing, and for assessing prompt leakage risks. However, existing video understanding benchmarks do not measure this capability: a caption may describe what is visible, but a replayable prompt must recover the generation-relevant controls needed to reproduce the video. To address this gap, we introduce VI-Bench, a benchmark built from 16.1 million real-user prompts and 900 human-verified AIGC videos. VI-Bench spans three progressively harder settings, namely single-shot semantic grounding, control over style and camera behavior, and multi-shot compositional inversion, and evaluates five generation-critical dimensions: subject, action, scene, style, and camera. We evaluate 18 representative VLMs, including 2 proprietary and 16 open-source models on VI-Bench, using an Inversion Score that measures prompt-level alignment with the original prompt and video-level fidelity of the regenerated video. The results reveal substantial limitations: even the strongest model achieves only 0.632 on Inversion Score, performance degrades sharply as samples require richer control and multi-shot reasoning, and models often produce plausible prompts whose regenerated videos deviate from the reference. These findings show that video prompt inversion is a distinct and under-evaluated capability requiring models to transform visual understanding into replay-stable generative control.

cs.CV↗
arXiv2026

Amortized Neural SVD for XL-MIMO: Structure-Guided Factor Prediction for Beamforming and Multi-Stream Utility

Singular value decomposition (SVD) is a core operation in multiple-input multiple-output (MIMO) beamforming, but the cubic complexity of standard SVD routines can lead to a major latency bottleneck as array dimensions scale to extremely large sizes. This paper presents a fully learned neural operator that avoids explicit SVD computation by directly mapping channel matrices to truncated low-rank factors for precoder and combiner design. In contrast to iterative numerical solvers and algorithm-unrolled networks, the proposed structure-aware model, termed SVDNet, produces these factors in a single forward pass at inference, shifting the per-instance decomposition cost to offline training. The model also includes lightweight constraints to enforce basic algebraic properties required by beamforming, such as semi-unitarity of the singular vectors and nonnegative singular values, without invoking matrix factorization kernels. Experiments on extremely large-scale MIMO channels with matrix dimensions up to 512*512 show that the proposed approach achieves spectral efficiency close to exact SVD-based beamforming in single-stream transmission and consistently improves multi-stream sum-rate over representative learned baselines, indicating good scalability for low-latency wireless processing.

eess.SP↗
arXiv2026

Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference

As large language models (LLMs) grow in scale and are predominantly served from remote platforms, verifying faithful inference execution becomes critical (i.e., ensuring that a provider actually executes the advertised model and computational workload rather than a tampered or downsized variant). Zero-knowledge (ZK) LLM inference offers an appealing approach. It promises public verifiability and delivers per-instance guarantees of equational correctness by proving that an output is consistent with executing a public architecture under committed, private weights. Though, we show that it does not bind the effort expended to produce the output. In this paper, we formalize this overlooked effort gap and introduce the Hollow-LLM Attack, in which a dishonest provider retains the declared architecture and parameter count but embeds ghost weights whose algebraic structure collapses effective computation. These witnesses satisfy the verification circuit and yield valid proofs, even though the dishonest model owner, who serves as the prover, performs computation commensurate with a much smaller model than the declared public architecture. This creates a profitable equilibrium in which providers deliver provably correct outputs at small-model cost while overclaiming model size. Accordingly, we characterize concrete families of ghost weights that compose with standard transformer blocks and show that such hollow deployments substantially reduce serving cost with zero quality loss under the same verification circuit. These findings underscore that proof of correct inference is not proof of large-model execution and necessitate additional protections to bind correctness to verifiable computational work.

cs.CR↗
arXiv2026

Non-Archimedean Rigidity and Uniformity for Common Preperiodic Points

Let $k$ be an algebraically closed, complete non-Archimedean field of residue characteristic $0$. Let $f$ be a polynomial of degree at least $2$ over $k$ which does not have potential good reduction. We prove that if $g$ is any other polynomial with the same Julia set, then $f$ and $g$ must be dynamically related. As a consequence, we show that for any two complex polynomials $f,g$ of degree at least $2$, either their sets of preperiodic points coincide, or the number of their common preperiodic points is uniformly bounded above by a constant depending only on the degrees, thereby answering a conjecture of DeMarco--Krieger--Ye for polynomials. We also establish relative results, allowing us to prove special cases of the DeMarco--Mavraki conjecture.

math.DS↗
arXiv2026

Wave2Body: Rethinking mmWave Human Pose Estimation as Radar-to-Body Token Translation

Millimeter-wave (mmWave) radar enables privacy-friendly human sensing, but its sparse point clouds are physical measurements of view-dependent electromagnetic reflections and only indirectly characterize body articulation. Recovering a complete 3D pose from such partial, geometry-dependent observations is therefore under-constrained. Existing methods directly regress joint coordinates from paired radar-pose data, relying on the same limited paired supervision to learn radar perception, human-body structure, and their alignment. This coupling can encourage dataset-specific shortcuts under ambiguous radar observations. We propose Wave2Body, a radar-to-body token translation framework that decouples these learning targets using a self-supervised mmWave tokenizer, a pretrained compositional body tokenizer that defines the output space, and a lightweight translator between them. Experiments on M4Human and mmBody show that Wave2Body achieves stronger cross-domain generalization than previous methods while incurring much lower computational costs for training and inference. All the code and experiment results are publicly available at https://github.com/Galaxywalk/Wave2Body.

cs.CV↗
arXiv2026

Spectral Attack on Continuous-Variable Quantum Key Distribution Systems

Continuous-variable quantum key distribution (CVQKD) has attracted extensive attention due to its compatibility and low costs. However, bandwidth mismatch exists to varying degrees between the transmitter and receiver. This may prevent frequency components carrying modulation information from being fully perceived by the legitimate party. In this paper, we identify a practical security loophole caused by bandwidth mismatch and propose a corresponding spectral attack scheme. Different from previous approaches that exploit security loopholes to conceal the excess noise introduced by intercept-resend attacks, this scheme can directly obtain raw-key information without introducing additional disturbances. A proof-of-principle attack on a CVQKD system with filtering operation is constructed to verify the feasibility. Experimental results indicate that Eve can obtain enough information to render the system insecure if this practical security loophole is ignored. Based on the identified security loophole, corresponding defense strategies are proposed. This work helps bridge the gap between theoretical models and practical implementations, providing a reference for defense design in practical quantum communication systems.

quant-ph↗
arXiv2026

Non-Archimedean techniques and dynamical degenerations

We develop non-Archimedean techniques to analyze the degeneration of a sequence of rational maps of the complex projective line. We provide an alternative to Luo's method which was based on ultra-limits of the hyperbolic 3-space. We build hybrid spaces using Berkovich theory which enable us to prove the convergence of equilibrium measures, and to determine the asymptotics of Lyapunov exponents.

math.DS↗