arXiv subjects

Qirui Li

Publications and source records attributed to Qirui Li.

At least 19 recordsLinked to original sources

arXiv2026

Role of the $\delta$ Meson in Softening the Symmetry Energy within the DDRHF Model

We investigate the effects of the isovector-scalar $\delta$ meson on the density dependence of the symmetry energy within the density-dependent relativistic Hartree--Fock (DDRHF) framework. As a baseline, we generate $1006$ accepted DDRHF parametrizations including the $\sigma$, $\omega$, $\rho$, and $\pi$ mesons by imposing empirical constraints on the saturation properties of nuclear matter. The resulting symmetry-energy slope parameters are confined to relatively large values, $L\simeq65$--$110~\mathrm{MeV}$. Two representative parametrizations, denoted RHF-NK1 and RHF-NK2, are randomly selected from this ensemble. Starting from these two parametrizations, we introduce the $\delta$ meson and readjust the meson--nucleon couplings under the same saturation-property constraints. The numerical optimization shows that small values of $L$ are obtained most efficiently when the $\delta$ coupling is taken to be constant. In this case, $L$ is reduced from approximately $73$ to $32~\mathrm{MeV}$, while the binding energy per nucleon, saturation density, symmetry energy, and incompressibility coefficient remain nearly unchanged. A channel-by-channel decomposition shows that the softening is not caused by the direct $\delta$-meson contribution alone, but by a redistribution among the $\delta$, $\rho$, and $\pi$ mesons together with the isoscalar Fock contributions. The resulting neutron-star mass--radius relations shift toward smaller radii, indicating that the $\delta$ meson provides an efficient additional degree of freedom for controlling the isovector properties of DDRHF models.

nucl-th↗

arXiv2026

SCOPE: Subspace Clustering with Online Per-Head Top-K Estimation for Sparse Video Attention

Diffusion Transformers (DiTs) incur quadratic self-attention cost over spatiotemporal tokens. Existing training-free sparse attention methods often construct sparse masks from block-level or cluster-level proxy scores, which can obscure fine-grained differences among keys and miss high contribution keys under aggressive sparsity. Moreover, such proxy scores may yield overly concentrated softmax distributions, causing Top-$p$ to retain too few keys for some query clusters. Although a fixed Top-$k$ minimum alleviates this failure mode, a shared value cannot adapt to variations across heads and inputs. To address both limitations, we propose SCOPE, a training-free sparse attention framework that combines 3D-RoPE-aligned key subspace clustering with online per-head Top-$k$ estimation for efficient video-DiT inference. SCOPE partitions post-RoPE keys into temporal, height, and width subspaces, clusters them independently, and aggregates the corresponding centroid scores through lookup tables to obtain per key proxy scores for each query cluster. Building on existing hybrid Top-$p$/fixed Top-$k$ selection, SCOPE derives a head-specific Top-$k$ value online by averaging the initial retained key counts within each head, weighted by query cluster size, and selects additional keys only for query clusters whose initial retained key counts fall below this value. Sparse attention is then computed over the selected original keys and values. Across six model--task configurations, SCOPE consistently outperforms existing training-free baselines in both fidelity and latency, achieving up to a $1.99\times$ end-to-end speedup on 720p HunyuanVideo with $28.46$ dB PSNR relative to dense attention.

cs.CV↗

arXiv2026

Text Template Tokens Are Implicit Semantic Registers in Diffusion Transformers

Modern text-to-image diffusion transformers (DiTs) generate images through joint attention, in which text and image tokens interact directly within a single sequence. In large-scale DiTs, the conditioning input contains not only the user prompt but also chat-template tokens introduced by LLM-based text encoders. Yet how these tokens participate in the denoising computation remains poorly understood. To probe this, we introduce a causal interpretability framework. Using it to separate prompt-content tokens from chat-template tokens, we find that the template tokens carry little prompt-specific information at the encoder output. Yet surprisingly, they emerge as dominant image-to-text attention sinks and causally maintain object identity inside the DiT, acting as implicit semantic registers. We show that they acquire this identity indirectly. Rather than reading the prompt tokens, they draw the identity from the image latents into which the prompt semantics have already been injected at the very first layer. We further reveal a division of labor across heads and depth in DiTs, where distinct heads route semantics or render visual structure, and identity is committed in early blocks, carried by middle blocks, and refined in late ones. As a practical payoff, this analysis yields a training-free pruning rule that removes the causally inert prompt-reading heads and cuts $20\%$ of joint-attention FLOPs at a $1.4$-point cost in GenEval accuracy. Overall, our work not only reveals that the tokens encoding semantics at the input need not be those that maintain them during generation, but also provides a causal view of internal mechanisms in diffusion transformers.

cs.CV↗

arXiv2026

PhysAgent: Reflective Agentic Physics Control for Physically Plausible Video Generation

Recent advances in physics-grounded video generation leverage physics simulation as a physical prior to guide video synthesis toward physically plausible outcomes. The simulation process is controlled by physical specifications, which are typically generated by a vision-language model in a single pass. Such one-shot prediction often fails to accurately translate user intent into executable simulations, particularly for fine-grained object dynamics, complex motion trajectories, and temporally structured interactions. In this paper, we propose PhysAgent, a reflective agentic framework that closes the loop among physical program generation, physics simulation, stage-specific verification, and targeted program repair. Beyond improving the control of coupled physical parameters, our framework enables the agent to progressively realize complex trajectories, multi-stage interactions, and precise event outcomes by treating each physical program as an executable hypothesis. In addition, we design a set of physics-control APIs to support more stable and complex motion behaviors. Extensive experiments demonstrate that PhysAgent produces more physically plausible videos, achieves better prompt alignment, and generalizes more effectively across diverse physical scenarios.

cs.CV↗

arXiv2026

Refusal Lives Downstream of Persona in Chat Models

Linear directions in activation space have been identified for both refusal and persona traits in instruction-tuned chat models, but the two have been studied as separate mechanisms. We show they interact: a compliant persona gates refusal. In Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct, we extract a compliant model-persona direction and a refusal direction and intervene on both. Compliant persona steering suppresses refusal -- in Llama, the refusal rate falls from 97% to 2%. Reintroducing the refusal direction partially restores refusal at late layers but not at early ones. Projecting out the persona direction in a late-layer window restores it to baseline; projecting out a random direction does not. Refusal is therefore gated at the late-layer expression stage, downstream of where it is computed. Treating refusal as a single isolated direction misses its dependence on persona.

cs.AI↗

arXiv2026

Rethinking Cross-Layer Information Routing in Diffusion Transformers

Diffusion Transformers (DiTs) have become a de facto backbone of modern visual generation, and nearly every major axis of their design -- tokenization, attention, conditioning, objectives, and latent autoencoders -- has been extensively revisited. The residual stream that governs how information accumulates across layers, however, has been directly inherited from the original Transformer. In this paper, we present a systematic empirical analysis of cross-layer information flow in DiTs, jointly along depth and denoising timestep, and identify three concrete symptoms of traditional residual addition, namely monotonic forward magnitude inflation, sharp backward gradient decay, and pronounced block-wise redundancy. Motivated by this diagnosis, we propose Diffusion-Adaptive Routing (\textsc{DAR}), a drop-in residual replacement that performs \emph{learnable, timestep-adaptive, and non-incremental} aggregation over the history of sublayer outputs. Moreover, the proposed \textsc{DAR} is compatible with many modern Transformer enhancement methods, such as REPA. On ImageNet $256\times256$, \textsc{DAR} improves SiT-XL/2 by $2.11$ FID ($7.56$ vs.\ $9.67$) and matches the baseline's converged quality with $8.75\times$ fewer training iterations. Stacked on top of REPA, it yields a $2\times$ training acceleration in the early stage, suggesting cross-layer information routing as an underexplored design axis in diffusion modeling, one that operates orthogonally to existing representation-alignment objectives. Beyond pretraining, \textsc{DAR} can also be applied during the fine-tuning stage of large-scale T2I models and preserves high-frequency details during Distribution Matching Distillation.

cs.CV↗

arXiv2025

InpaintDPO: Mitigating Spatial Relationship Hallucinations in Foreground-conditioned Inpainting via Diverse Preference Optimization

Foreground-conditioned inpainting, which aims at generating a harmonious background for a given foreground subject based on the text prompt, is an important subfield in controllable image generation. A common challenge in current methods, however, is the occurrence of Spatial Relationship Hallucinations between the foreground subject and the generated background, including inappropriate scale, positional relationships, and viewpoints. Critically, the subjective nature of spatial rationality makes it challenging to quantify, hindering the use of traditional reward-based RLHF methods. To address this issue, we propose InpaintDPO, the first Direct Preference Optimization (DPO) based framework dedicated to spatial rationality in foreground-conditioned inpainting, ensuring plausible spatial relationships between foreground and background elements. To resolve the gradient conflicts in standard DPO caused by identical foreground in win-lose pairs, we propose MaskDPO, which confines preference optimization exclusively to the background to enhance background spatial relationships, while retaining the inpainting loss in the foreground region for robust foreground preservation. To enhance coherence at the foreground-background boundary, we propose Conditional Asymmetric Preference Optimization, which samples pairs with differentiated cropping operations and applies global preference optimization to promote contextual awareness and enhance boundary coherence. Finally, based on the observation that winning samples share a commonality in plausible spatial relationships, we propose Shared Commonality Preference Optimization to enhance the model's understanding of spatial commonality across high-quality winning samples, further promoting shared spatial rationality.

cs.CV↗

arXiv2025

Compact Attention: Exploiting Structured Spatio-Temporal Sparsity for Fast Video Generation

The computational demands of self-attention mechanisms pose a critical challenge for transformer-based video generation, particularly in synthesizing ultra-long sequences. Current approaches, such as factorized attention and fixed sparse patterns, fail to fully exploit the inherent spatio-temporal redundancies in video data. Through systematic analysis of video diffusion transformers (DiT), we uncover a key insight: Attention matrices exhibit structured, yet heterogeneous sparsity patterns, where specialized heads dynamically attend to distinct spatiotemporal regions (e.g., local pattern, cross-shaped pattern, or global pattern). Existing sparse attention methods either impose rigid constraints or introduce significant overhead, limiting their effectiveness. To address this, we propose Compact Attention, a hardware-aware acceleration framework featuring three innovations: 1) Adaptive tiling strategies that approximate diverse spatial interaction patterns via dynamic tile grouping, 2) Temporally varying windows that adjust sparsity levels based on frame proximity, and 3) An automated configuration search algorithm that optimizes sparse patterns while preserving critical attention pathways. Our method achieves 1.6~2.5x acceleration in attention computation on single-GPU setups while maintaining comparable visual quality with full-attention baselines. This work provides a principled approach to unlocking efficient long-form video generation through structured sparsity exploitation. Project Page: https://yo-ava.github.io/Compact-Attention.github.io/

cs.CV↗

arXiv2025

VRBench: A Benchmark for Multi-Step Reasoning in Long Narrative Videos

We present VRBench, the first long narrative video benchmark crafted for evaluating large models' multi-step reasoning capabilities, addressing limitations in existing evaluations that overlook temporal reasoning and procedural validity. It comprises 960 long videos (with an average duration of 1.6 hours), along with 8,243 human-labeled multi-step question-answering pairs and 25,106 reasoning steps with timestamps. These videos are curated via a multi-stage filtering process including expert inter-rater reviewing to prioritize plot coherence. We develop a human-AI collaborative framework that generates coherent reasoning chains, each requiring multiple temporally grounded steps, spanning seven types (e.g., event attribution, implicit inference). VRBench designs a multi-phase evaluation pipeline that assesses models at both the outcome and process levels. Apart from the MCQs for the final results, we propose a progress-level LLM-guided scoring metric to evaluate the quality of the reasoning chain from multiple dimensions comprehensively. Through extensive evaluations of 12 LLMs and 19 VLMs on VRBench, we undertake a thorough analysis and provide valuable insights that advance the field of multi-step reasoning.

cs.CV↗

arXiv2025

A Proof of the Biquadratic Linear AFL for GL(4)

We prove both the biquadratic Guo--Jacquet Fundamental Lemma (FL) and the biquadratic linear Arithmetic Fundamental Lemma (AFL) for GL(4) with the unit test function. Our approach relies on a detailed study of pairs of quadratic embeddings, which ultimately enables a reduction from the biquadratic case of GL(4) to the coquadratic case of GL(2). We further identify conditions under which the biquadratic case can be derived from the coquadratic case, and show that this reduction allows us to establish the conjectures for all orbits in GL(4). As an additional consequence, we also prove the biquadratic FL for the identity test function in certain special families of orbits in GL(2n). All results hold over both p-adic fields and local fields of positive characteristic.

math.NT↗

arXiv2025

CAT: A GPU-Accelerated FHE Framework with Its Application to High-Precision Private Dataset Query

We introduce an open-source GPU-accelerated fully homomorphic encryption (FHE) framework CAT, which surpasses existing solutions in functionality and efficiency. \emph{CAT} features a three-layer architecture: a foundation of core math, a bridge of pre-computed elements and combined operations, and an API-accessible layer of FHE operators. It utilizes techniques such as parallel executed operations, well-defined layout patterns of cipher data, kernel fusion/segmentation, and dual GPU pools to enhance the overall execution efficiency. In addition, a memory management mechanism ensures server-side suitability and prevents data leakage. Based on our framework, we implement three widely used FHE schemes: CKKS, BFV, and BGV. The results show that our implementation on Nvidia 4090 can achieve up to 2173$\times$ speedup over CPU implementation and 1.25$\times$ over state-of-the-art GPU acceleration work for specific operations. What's more, we offer a scenario validation with CKKS-based Privacy Database Queries, achieving a 33$\times$ speedup over its CPU counterpart. All query tasks can handle datasets up to $10^3$ rows on a single GPU within 1 second, using 2-5 GB storage. Our implementation has undergone extensive stability testing and can be easily deployed on commercial GPUs. We hope that our work will significantly advance the integration of state-of-the-art FHE algorithms into diverse real-world systems by providing a robust, industry-ready, and open-source tool.

cs.CR↗

arXiv2024

On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection

Large numbers of synthesized videos from diffusion models pose threats to information security and authenticity, leading to an increasing demand for generated content detection. However, existing video-level detection algorithms primarily focus on detecting facial forgeries and often fail to identify diffusion-generated content with a diverse range of semantics. To advance the field of video forensics, we propose an innovative algorithm named Multi-Modal Detection(MM-Det) for detecting diffusion-generated videos. MM-Det utilizes the profound perceptual and comprehensive abilities of Large Multi-modal Models (LMMs) by generating a Multi-Modal Forgery Representation (MMFR) from LMM's multi-modal space, enhancing its ability to detect unseen forgery content. Besides, MM-Det leverages an In-and-Across Frame Attention (IAFA) mechanism for feature augmentation in the spatio-temporal domain. A dynamic fusion strategy helps refine forgery representations for the fusion. Moreover, we construct a comprehensive diffusion video dataset, called Diffusion Video Forensics (DVF), across a wide range of forgery videos. MM-Det achieves state-of-the-art performance in DVF, demonstrating the effectiveness of our algorithm. Both source code and DVF are available at https://github.com/SparkleXFantasy/MM-Det.

cs.CV↗

arXiv2023

MetaScript: Few-Shot Handwritten Chinese Content Generation via Generative Adversarial Networks

In this work, we propose MetaScript, a novel Chinese content generation system designed to address the diminishing presence of personal handwriting styles in the digital representation of Chinese characters. Our approach harnesses the power of few-shot learning to generate Chinese characters that not only retain the individual's unique handwriting style but also maintain the efficiency of digital typing. Trained on a diverse dataset of handwritten styles, MetaScript is adept at producing high-quality stylistic imitations from minimal style references and standard fonts. Our work demonstrates a practical solution to the challenges of digital typography in preserving the personal touch in written communication, particularly in the context of Chinese script. Notably, our system has demonstrated superior performance in various evaluations, including recognition accuracy, inception score, and Frechet inception distance. At the same time, the training conditions of our model are easy to meet and facilitate generalization to real applications.

cs.CV↗

arXiv2023

Arithmetic transfer for inner forms of $GL_{2n}$

We formulate Guo--Jacquet type fundamental lemma conjectures and arithmetic transfer conjectures for inner forms of $GL_{2n}$. Our main results confirm these conjectures for division algebras of invariant $1/4$ and $3/4$.

math.NT↗

arXiv2022

On the Linear AFL: The Non-Basic Case

The linear Arithmetic Fundamental Lemma (AFL) conjecture compares intersection numbers on Lubin--Tate deformation spaces with derivatives of orbital integrals. It has been introduced for elliptic orbits in arXiv:1803.07553 and arXiv:2010.07365. In these cases, the relevant intersection problem is formulated for the basic isogeny class. In the present article, we extend the theory to all orbits and all isogeny classes. Our main result is a reduction of the non-basic cases of the AFL to the basic ones, which is achieved by exploiting the connected-\'etale sequence. Our theory will be relevant in the global setting, where also locally non-elliptic orbits may contribute in a non-trivial way.

math.AG↗

arXiv2021

An intersection number formula for CM cycles in Lubin-Tate towers

We give an explicit formula for the arithmetic intersection number of CM cycles on Lubin-Tate spaces for all levels. We prove our formula by formulating the intersection number on the infinite level. Our CM cycles are constructed by choosing two separable quadratic extensions $K_1,K_2/F$ of non-Archimedean local fields $F$. Our formula works for all cases, $K_1$ and $K_2$ can be either the same or different, ramify or unramified. As applications, this formula translate the linear Arithmetic Fundamental Lemma (linear AFL) into a comparison of integrals. This formula can also be used to recover Gross and Keating's result on lifting endomorphism of formal modules.

math.NT↗

arXiv2020

A computational proof of the linear Arithmetic Fundamental Lemma of GL$_4$

Let $K/F$ be an unramified quadratic extension of non-Archimedian local fields with residue character not equals to 2. We prove the linear Arithmetic Fundamental Lemma for GL$_4$ with the unit element in the spherical Hecke Algebra. In this article, all measures are normalized by its hyperspecial subgroup.

math.NT↗

arXiv2020

Learning by Sampling and Compressing: Efficient Graph Representation Learning with Extremely Limited Annotations

Graph convolution network (GCN) attracts intensive research interest with broad applications. While existing work mainly focused on designing novel GCN architectures for better performance, few of them studied a practical yet challenging problem: How to learn GCNs from data with extremely limited annotation? In this paper, we propose a new learning method by sampling strategy and model compression to overcome this challenge. Our approach has multifold advantages: 1) the adaptive sampling strategy largely suppresses the GCN training deviation over uniform sampling; 2) compressed GCN-based methods with a smaller scale of parameters need fewer labeled data to train; 3) the smaller scale of training data is beneficial to reduce the human resource cost to label them. We choose six popular GCN baselines and conduct extensive experiments on three real-world datasets. The results show that by applying our method, all GCN baselines cut down the annotation requirement by as much as 90$\%$ and compress the scale of parameters more than 6$\times$ without sacrificing their strong performance. It verifies that the training method could extend the existing semi-supervised GCN-based methods to the scenarios with the extremely small scale of labeled data.

cs.LG↗