Searcharxiv⌕ Search

arXiv subjects

Shijie Zhang

Publications and source records attributed to Shijie Zhang.

At least 19 recordsLinked to original sources

Exposing and Mitigating Temporal Attack in Deepfake Video Detection

While spatiotemporal deepfake detectors achieve high AUC, our experiments reveal their susceptibility to evasion attacks. These models tend to overfit on fragile temporal spectrum cues, rather than learning robust semantic causality. To mitigate this vulnerability, we propose SpInShield, a temporal spectral-invariant defense framework explicitly designed to decouple semantic motion from manipulatable spectral artifacts. We propose a learnable spectral adversary that dynamically synthesizes severe spectral deformations, simulating extreme attack scenarios. By employing a shortcut suppression optimization strategy, SpInShield compels the encoder to extract reliable forensic cues while purging unstable spectral statistics from the latent space. Experiments show that SpInShield obtains competitive performance on widely used datasets and outperforms the strongest baseline by 21.30 percentage points in AUC under simulated amplitude spectral attacks.

cs.CV↗

ThinkWise: Learning When to Think in Large Multimodal Reasoning Models via Visual and Difficulty Perception

Large Multimodal Reasoning Models (LMRMs) extend large language models to heterogeneous modalities, demonstrating strong potential across a wide range of open-ended tasks. Reinforcement Learning with Verifiable Rewards (RLVR) has been widely adopted to improve reasoning performance in LMRMs; however, existing approaches overlook the critical role of visual perception uncertainty in guiding reasoning processes. We term this limitation Reasoning Allocation Misalignment, where the misalignment between visual perception uncertainty and reasoning depth leads to over-allocation of deep reasoning under uncertain visual grounding, resulting in unnecessarily long reasoning trajectories and degraded performance. To address this issue, we propose ThinkWise, a perception-aware RLVR framework that jointly models task difficulty and visual perception state to regulate reasoning depth. Specifically, we introduce an instance-adaptive visual perception modeling mechanism that enables context-aware calibration of perception uncertainty, leading to more accurate perception estimation. We further design a structured reward formulation that jointly considers perception, correctness, and task difficulty. In particular, an answer reward provides supervision for correctness, a perception- and difficulty-conditioned length reward adaptively regulates reasoning depth, and a perception reward encourages trajectories with stronger visual grounding. Extensive experiments on multiple multimodal reasoning benchmarks demonstrate that ThinkWise consistently improves the trade-off between accuracy and reasoning efficiency across both 3B and 7B models. These results highlight the importance of perception-aware reasoning and establish a principled framework for advancing LMRMs.

cs.CV↗

Inverse KLT kernel of AdS string integrals from the intersection matrix

We develop a topological and combinatorial interpretation of the Kawai--Lewellen--Tye (KLT) relations for $n$-point AdS string-amplitude building blocks, which are disk integrals with multivariable multiple polylogarithm insertions and sphere integrals with their single-valued counterparts. Building on the framework of twisted de Rham theory over a noncommutative ring, we reformulate the problem in terms of associahedral chambers of the moduli space loaded with solutions of the Knizhnik--Zamolodchikov equation. We construct regularized twisted cycles and express their intersection pairings through local monodromies and Drinfeld associators, which control transport between fusion-tree normalizations. Organizing the intersection contributions by common boundary faces yields the intersection matrix, which is identified with the inverse KLT kernel by the twisted period relation. The field-theory limit of the inverse KLT kernel is also given.

hep-th↗

On the (In)stability of Adaptive Fee Mechanisms in Multi-Chain Systems

EIP-1559 has become the dominant transaction fee mechanism in permissionless blockchains, regulating scarce block space through an adaptive base fee. The design balances a trade-off between responsiveness to demand fluctuations and stability of long-term resource usage. On a single chain, this trade-off is well understood and empirically stable. In multi-chain systems, however, cross-chain flows introduce delayed economic interactions across otherwise independent fee markets. These interactions alter the control dynamics and create stability risks that existing models do not capture. Deployed systems have already polarized. Polkadot enforces conservative update rates, effectively disabling fee adaptivity and reverting the market to inefficient first-price auctions. Cosmos adopts aggressive update rates without formal guarantees. To fill this gap, we propose a control-theoretic framework that formalizes multi-chain fee markets as a nonlinear dynamical system with delayed feedback. We identify delay-amplified over-correction as the structural root of instability. Experiments show that cross-chain coupling erodes stability relative to monolithic blockchains, reducing the safe parameter space for base fee update rates by up to 84.8%. Moreover, instability propagates across chains, contaminating the otherwise stable peer chains. At Ethereum's update rate, which leaves a monolithic chain stable, the multi-chain system exhibits extreme load oscillation between full and near-empty blocks, with over half of all blocks at full capacity. Mainstream Cosmos chains already exceed their safe bounds under their default configurations, and the resulting load oscillation raises their block failure rate by an order of magnitude. Guided by this mechanism, we design a lightweight protocol-level mitigation that removes the over-correction and stabilizes dynamics in multi-chain environments.

cs.CR↗

Nuclear mass table in deformed relativistic Hartree-Bogoliubov theory in continuum, III: nuclei with $8 \leq Z \leq 120$

The mass table in the deformed relativistic Hartree-Bogoliubov theory in continuum (DRHBc) with the PC-PK1 density functional has been established for nuclei with $8 \leq Z \leq 120$, extended from the previous works for even-even nuclei [Zhang et al. (DRHBc mass table collaboration), At. Data Nucl. Data Tables 144, 101488 (2022)] and for even-$Z$ nuclei [Guo et al. (DRHBc mass table collaboration), At. Data Nucl. Data Tables 158, 101661 (2024)]. The calculated binding energies, two- and one-nucleon separation energies, root-mean-square (rms) radii of neutron, proton, matter, and charge distributions, quadrupole deformations, neutron and proton Fermi surfaces, and the blocked neutron (proton) orbitals of odd-$N$ ($Z$) nuclei are tabulated and compared with the available experimental data. A total of 9495 nuclei are predicted to be bound, with an rms deviation of 1.444 MeV from the 2380 mass data. Good agreement with the available experimental pairing gaps, $α$ decay energies, and charge radii is also achieved. The accuracies of the calculated nuclear masses and nucleon separation energies as well as the prediction for drip lines are compared with those obtained by other relativistic and nonrelativistic density functional calculations. It turns out that the DRHBc theory with PC-PK1 provides one of the best microscopic descriptions for nuclear masses. The systematics of nucleon separation energies, pairing gaps, pairing energies, two-nucleon gaps, $α$ decay energies, rms radii, quadrupole deformations, potential energy curves, neutron density distributions, and neutron mean-field potentials are discussed.

nucl-th↗

Role-Guided MOE for Encoder-Level Pathology Representation Learning in WSI Classification

Whole slide image classification is a fundamental task in computational pathology, where patch representation quality directly affects downstream aggregation and slide-level discriminability. Pathology foundation models are widely adopted as frozen feature extractors for WSI classification; however, their fixed encoders may produce representations insufficiently adapted to target-specific tissue patterns and discriminative cues. Fine-tuning can improve target adaptation, but introduces a trade-off between pathology-specific representation capacity and adaptation efficiency, particularly in data-scarce settings. To address this, we propose a pathology role-guided mixture-of-experts feed-forward network (MoE-FFN) framework for efficient encoder-level representation learning. We design a two-stage training paradigm to establish and adapt pathology-aware expert specialization. In source-domain expert initialization, pathology-specific priors are distilled from a frozen Virchow2 teacher into a lightweight DINOv2-small student, while role prototypes serve as weak pathological anchors to encourage distinct expert functions. MoE-FFN blocks are introduced into selected high-level transformer layers to provide transformation diversity for heterogeneous pathological patterns. In target-domain adaptation, the initialized experts are refined through asymmetric prototype-guided optimization, enhancing task-relevant positive evidence and separating confusable hard negatives. The resulting encoder extracts offline patch representations that can be directly integrated with standard MIL aggregators. Experiments on the public BRACS dataset and a private PAROTID WSI dataset across five representative backbones demonstrate consistent improvements over the strongest baseline.

cs.CV↗

Robustness of Diffusion Models under Distribution Shift

Score-based diffusion models are increasingly considered in settings where the underlying data distribution may differ from the training distribution, yet existing theoretical guarantees largely focus on the no-shift setting. In this work, we study robust score estimation under Wasserstein perturbations of a reference distribution. For the Ornstein--Uhlenbeck diffusion, we show that robust estimation decomposes into two fundamental components: the statistical cost of learning the reference distribution and the intrinsic cost of distribution shift. The latter scales quadratically with the Wasserstein radius, and this dependence is minimax optimal. We construct an explicit finite-sample estimator achieving the resulting robust minimax rate without knowing the shift radius. When the reference distribution lies on an unknown low-dimensional subspace, the statistical term adapts to the intrinsic dimension while the shift cost remains unchanged. Finally, we show that the same decomposition governs positive-time reverse sampling and obtain matching minimax guarantees in KL divergence. Together, these results characterize how finite data, intrinsic dimension, and distribution shift affect the robustness of score-based diffusion models.

math.ST↗

Jailbreaking Text-to-Image Models Through Cracks: Navigating Heterogeneous Safety Filters via Multi-Agent Debate

Text-to-image (T2I) models remain vulnerable to jailbreak attacks that elicit Not-Safe-For-Work (NSFW) content, despite increasingly being guarded by heterogeneous, multi-layer safety stacks combining text filters, image classifiers, and cross-modal detectors. Existing jailbreak studies either optimize against individual filters or query the complete pipeline with aggregate feedback, making it difficult to identify the active constraint and adapt to conflicts across safety layers. In this paper, we introduce the Detection Surface, a unified geometric framework that characterizes the decision boundaries induced by heterogeneous T2I safety filters and their joint effect on the jailbreak search space. This formulation reveals that successful evasion is governed by a sparse and non-convex region shaped by cross-layer conflicts, where mutations that bypass one filter may increase exposure to another. Motivated by this analysis, we propose CRACK, a multi-agent debate framework for adaptive jailbreak search that decomposes jailbreak search into exploration, diagnosis, and arbitration. CRACK coordinates an Attack Agent, a Defense Agent, and a Judge Agent to iteratively generate prompt mutations, obtain layer-specific diagnostic feedback, and optimize mutation strategies through reward-guided refinement. Through repeated rounds of debate, CRACK adapts its search direction to the evolving cross-layer constraints while preserving the original harmful intent. Extensive experiments across multiple T2I models, datasets, and safety configurations show that CRACK achieves Attack Success Rates (ASR) of up to 99.63% under composite defenses, while requiring fewer queries than existing methods and maintaining semantic fidelity.

cs.AI↗

NFR-to-Code Traceability in a Blockchain-IoT System: An Empirical Study

Requirement-to-Code traceability has been widely studied, yet existing research and public benchmarks remain largely centered on functional requirements (FRs). In contrast, traceability for non-functional requirements (NFRs) remains more difficult and underexplored, which hinders the verification of critical quality concerns such as security and reliability.This paper studies NFR-to-Code traceability based on a real-world blockchain-IoT project. We design an annotation protocol for constructing trace links across heterogeneous artifacts and build a manually curated subset containing both FR and NFR links. Using this subset, we examine the characteristics of NFR traceability and further evaluate four representative retrieval baselines: TF-IDF, BM25, LSI, and WMD. The results show that FR-to-Code tracing is consistently easier than NFR-to-Code tracing, while security-related NFRs are the most difficult subset. They further indicate that the main challenge of NFR traceability lies not in requirement availability, but in implementation evidence that is distributed and not clearly localized in code.

cs.SE↗

ENVS: Environment-Native Verified Search for Long-Horizon GUI Agents

As multimodal agents move from interface understanding to real software control, successful trajectory discovery in live desktop environments becomes a key challenge. GUI tasks require long-horizon sequences of precise mouse and keyboard actions, while feedback is sparse, delayed, and costly to obtain through VM rollouts. We propose Environment-Native Verified Search (ENVS), a training-time search-and-filter pipeline that uses the environment to construct verified supervision before policy optimization: it branches over behaviorally distinct GUI actions in live OSWorld VMs, verifies successful leaves, and trains from globally balanced step-level supervision. To evaluate robustness under realistic desktop interruptions, we also introduce OSWorld-Noisy, a dynamic benchmark for recoverable desktop interruptions that preserves the original tasks while testing whether agents can refocus, dismiss, wait, or recover under live perturbations. On the 300-task OSWorld pool, ENVS reaches 30.3 pass@8 on original evaluations and 29.0 on OSWorld-Noisy, outperforming matched ARPO-style online RL while reducing compute from 184-192 to 138-153 GPU-hours; even with only 30% of its search data, ENVS reaches 27.0 pass@8, exceeding ARPO from the base model. Training from noisy environments also better preserves visual-reasoning abilities on auxiliary benchmarks, including OSWorld-G Refusal (16.7 vs. 1.9) and BLINK Functional Correspondence (26.2 vs. 23.1).

cs.AI↗

Randomized Midpoint Method for Log-Concave Sampling under Constraints

In this paper, we study the problem of sampling from log-concave distributions supported on convex and compact sets, with a particular focus on the randomized midpoint discretization of both overdamped and kinetic Langevin diffusions in constrained domains. We revisit the proximal framework for handling constraints through projection operators and develop a more general formulation that encompasses Euclidean, Bregman, and Gauge projections. The resulting smooth approximation allows a unified and tractable analysis of Langevin algorithms and their variants under constraints. Within this framework, we establish convergence guarantees in Wasserstein-$q$ $(q\geqslant 1)$ distances between the smooth surrogate and the target distribution. We further derive complementary lower bounds, showing that the results are near-optimal in order. Building upon this tight approximation analysis, we obtain new convergence guarantees for the randomized midpoint Langevin algorithms and refined bounds for both vanilla and kinetic Langevin Monte Carlo methods under constraints, thereby advancing the theoretical understanding of constrained diffusion-based sampling.

stat.ML↗

CLPO: Curriculum Learning meets Policy Optimization for LLM Reasoning

Online reinforcement learning with verifiable rewards (RLVR) has become an effective paradigm for improving the reasoning abilities of large language models, but most methods still optimize reasoning trajectories over the static problem set, wasting rollout budget on solved or overly difficult problems. We propose \textbf{CLPO (Curriculum Learning meets Policy Optimization)}, a self-evolving curriculum framework that uses on-policy rollout accuracy to identify solved, medium-difficulty, and hard problems, then restructures selected tasks according to the model's current capability. Hard problems are simplified to become learnable, while medium-difficulty problems are diversified to provide useful training variation. This allows the learning curriculum to co-evolve with the policy rather than remaining fixed as the model's capability boundary shifts. Rather than treating these rewrites as static data augmentation, CLPO optimizes restructuring trajectories with credit assigned by the downstream accuracy gain of the rewritten problem, requiring no additional human annotations beyond the original verifiable answers. Experiments across mathematical reasoning and out-of-domain general reasoning benchmarks show that CLPO substantially outperforms GRPO and DAPO on Qwen3-8B by 10.21 and 7.75 average points, respectively. Ablation studies on math and code domains further show that both the restructuring mode and the rewriting loss contribute to the final gains, demonstrating that CLPO provides a scalable and robust pathway for eliciting stronger reasoning capabilities through a self-evolving curriculum.

cs.AI↗

Gemini Embedding 2: A Native Multimodal Embedding Model from Gemini

We introduce Gemini Embedding 2, a native multimodal embedding model that allows embedding video, audio, image, and text modalities in a unified representation space. We leverage the multimodal capabilities of Gemini to produce embeddings for arbitrary combinations of interleaved inputs across all these modalities that generalize well across a wide variety of tasks. Applying large-scale contrastive learning in a multi-task multi-stage training setup, we achieve state-of-the-art performance on key embedding benchmarks including unimodal, cross-modal, and multimodal retrieval spanning a diverse set of tasks. We show that our embedding model demonstrates strong performance (with a score of 62.9 R@1 on MSCOCO, 68.8 NDCG@10 on Vatex, 69.9 on MTEB multilingual and 84.0 on MTEB Code) across a variety of tasks surpassing the performance of specialized models. These unified capabilities make Gemini Embedding 2 a promising candidate for downstream use cases such as RAG, recommendation and search. Furthermore, its robust zero-shot performance across distinct fields - from astronomy and bioscience to fine arts and the culinary arts - establishes it as a highly reliable, out-of-the-box representation even for specialized domains.

cs.CV↗

Twisted de Rham theory for string double copy in AdS

This work is motivated by the recent evidence for a double-copy relationship between open- and closed-string amplitudes in Anti-de Sitter (AdS) space. At present, the evidence has the form of a double-copy relation for string-amplitude building blocks, which are combined using the multiple-polylogarithm (MPL) generating functions. These generate MPLs relevant for all-order AdS curvature corrections of four-point string amplitudes. In this paper, we prove this building-block double copy using a new, noncommutative version of twisted de Rham theory. In flat space, the usual twisted de Rham theory is already known to be a natural framework to describe the Kawai-Lewellen-Tye (KLT) double-copy map from open- to closed-string amplitudes, in which the KLT kernel can be computed from the intersections of the open-string amplitude integration contours. We formulate twisted de Rham theory for noncommutative-ring-valued differential forms on complex manifolds and use it to derive the intersection number of two open-string contours, which are closed in the noncommutative twisted homology sense. The inverse of this intersection number is precisely the AdS double-copy kernel for the four-point open- and closed-string generating functions.

hep-th↗

Spectral-Aware Text-to-Time Series Generation with Billion-Scale Multimodal Meteorological Data

Text-to-time-series generation is particularly important in meteorology, where natural language offers intuitive control over complex, multi-scale atmospheric dynamics. Existing approaches are constrained by the lack of large-scale, physically grounded multimodal datasets and by architectures that overlook the spectral-temporal structure of weather signals. We address these challenges with a unified framework for text-guided meteorological time-series generation. First, we introduce MeteoCap-3B, a billion-scale weather dataset paired with expert-level captions constructed via a Multi-agent Collaborative Captioning (MACC) pipeline, yielding information-dense and physically consistent annotations. Building on this dataset, we propose MTransformer, a diffusion-based model that enables precise semantic control by mapping textual descriptions into multi-band spectral priors through a Spectral Prompt Generator, which guides generation via frequency-aware attention. Extensive experiments on real-world benchmarks demonstrate state-of-the-art generation quality, accurate cross-modal alignment, strong semantic controllability, and substantial gains in downstream forecasting under data-sparse and zero-shot settings. Additional results on general time-series benchmarks indicate that the proposed framework generalizes beyond meteorology.

cs.LG↗

Multi-Agent Causal Reasoning for Suicide Ideation Detection Through Online Conversations

Suicide remains a pressing global public health concern. While social media platforms offer opportunities for early risk detection through online conversation trees, existing approaches face two major limitations: (1) They rely on predefined rules (e.g., quotes or relies) to log conversations that capture only a narrow spectrum of user interactions, and (2) They overlook hidden influences such as user conformity and suicide copycat behavior, which can significantly affect suicidal expression and propagation in online communities. To address these limitations, we propose a Multi-Agent Causal Reasoning (MACR) framework that collaboratively employs a Reasoning Agent to scale user interactions and a Bias-aware Decision-Making Agent to mitigate harmful biases arising from hidden influences. The Reasoning Agent integrates cognitive appraisal theory to generate counterfactual user reactions to posts, thereby scaling user interactions. It analyses these reactions through structured dimensions, i.e., cognitive, emotional, and behavioral patterns, with a dedicated sub-agent responsible for each dimension. The Bias-aware Decision-Making Agent mitigates hidden biases through a front-door adjustment strategy, leveraging the counterfactual user reactions produced by the Reasoning Agent. Through the collaboration of reasoning and bias-aware decision making, the proposed MACR framework not only alleviates hidden biases, but also enriches contextual information of user interactions with counterfactual knowledge. Extensive experiments on real-world conversational datasets demonstrate the effectiveness and robustness of MACR in identifying suicide risk.

cs.CL↗

Know What You Know: Metacognitive Entropy Calibration for Verifiable RL Reasoning

Large reasoning models (LRMs) have emerged as a powerful paradigm for solving complex real-world tasks. In practice, these models are predominantly trained via Reinforcement Learning with Verifiable Rewards (RLVR), yet most existing outcome-only RLVR pipelines rely almost exclusively on a binary correctness signal and largely ignore the model's intrinsic uncertainty. We term this discrepancy the uncertainty-reward mismatch, under which high- and low-uncertainty solutions are treated equivalently, preventing the policy from "Know What You Know" and impeding the shift from optimizing for correct answers to optimizing effective reasoning paths. This limitation is especially critical in reasoning-centric tasks such as mathematics and question answering, where performance hinges on the quality of the model's internal reasoning process rather than mere memorization of final answers. To address this, we propose EGPO, a metacognitive entropy calibration framework that explicitly integrates intrinsic uncertainty into RLVR for enhancing LRMs. EGPO estimates per-sample uncertainty using a zero-overhead entropy proxy derived from token-level likelihoods and aligns it with extrinsic correctness through an asymmetric calibration mechanism that preserves correct reasoning while selectively regulating overconfident failures, thereby enabling stable and uncertainty-aware policy optimization. Moreover, EGPO recovers informative learning signals from otherwise degenerate group-based rollouts without modifying the verifier or reward definition. Extensive experiments across multiple benchmarks demonstrate that the proposed EGPO leads to substantial and consistent improvements in reasoning performance, establishing a principled path for advancing LRMs through metacognitive entropy calibration.

cs.AI↗

Explore-on-Graph: Incentivizing Autonomous Exploration of Large Language Models on Knowledge Graphs with Path-refined Reward Modeling

The reasoning process of Large Language Models (LLMs) is often plagued by hallucinations and missing facts in question-answering tasks. A promising solution is to ground LLMs' answers in verifiable knowledge sources, such as Knowledge Graphs (KGs). Prevailing KG-enhanced methods typically constrained LLM reasoning either by enforcing rules during generation or by imitating paths from a fixed set of demonstrations. However, they naturally confined the reasoning patterns of LLMs within the scope of prior experience or fine-tuning data, limiting their generalizability to out-of-distribution graph reasoning problems. To tackle this problem, in this paper, we propose Explore-on-Graph (EoG), a novel framework that encourages LLMs to autonomously explore a more diverse reasoning space on KGs. To incentivize exploration and discovery of novel reasoning paths, we propose to introduce reinforcement learning during training, whose reward is the correctness of the reasoning paths' final answers. To enhance the efficiency and meaningfulness of the exploration, we propose to incorporate path information as additional reward signals to refine the exploration process and reduce futile efforts. Extensive experiments on five KGQA benchmark datasets demonstrate that, to the best of our knowledge, our method achieves state-of-the-art performance, outperforming not only open-source but also even closed-source LLMs.

cs.CL↗