arXiv subjects

Ruyi Chen

Publications and source records attributed to Ruyi Chen.

8 recordsLinked to original sources

arXiv2026

HoloFair: Unified T2I Fairness Evaluation and Fair-GRPO Debiasing

Text-to-Image (T2I) models have made significant strides in visual realism and semantic consistency, yet they often perpetuate and amplify societal biases. Existing evaluation methods typically address only single-dimensional biases, lacking perspectives to uncover model biases at social-related deeper semantic levels. We introduce HoloFair, a comprehensive benchmark framework for multidimensional demographic bias analysis. Built upon our large-scale fairness-oriented dataset and the SpaFreq (Spatial-Frequency) attribute classifier, this framework proposes the Multi-attribute, Group-wise Bias Index (MGBI) metric, designed to assess both intrinsic diversity and conditional biases. Beyond evaluation, we further introduce Fair-GRPO, a reinforcement-learning-based debiasing method that alters the distribution of generative models through a designed multi-objective reward function. E.g., experiments on the SD3.5-Medium model demonstrate that Fair-GRPO significantly improves multidimensional fairness while maintaining high image quality. We also analyze potential reward hacking phenomena and provide corresponding mitigation strategies. Code and dataset are available at https://github.com/1059684669/HoloFair

cs.CV↗

arXiv2026

LITMUS: Benchmarking Behavioral Jailbreaks of LLM Agents in Real OS Environments

The rapid proliferation of LLM-based autonomous agents in real operating system environments introduces a new category of safety risk beyond content safety: behavior jailbreak, where an adversary induces an agent to execute dangerous OS-level operations with irreversible consequences. Existing benchmarks either evaluate safety at the semantic layer alone, missing physical-layer harms, or fail to isolate test cases, letting earlier runs contaminate later ones. We present LITMUS (LLM-agents In-OS Testing for Measuring Unsafe Subversion), a benchmark addressing both gaps via a semantic-physical dual verification mechanism and OS-level state rollback. LITMUS comprises 819 high-risk test cases organized into one harmful seed subset and six attack-extended subsets covering three adversarial paradigms (jailbreak speaking, skill injection, and entity wrapping), plus a fully automated multi-agent evaluation framework judging behavior at both conversational and OS-level physical layers. Evaluation across frontier agents reveals three findings: (1) current agents lack effective safety awareness, with strong models (e.g., Claude Sonnet 4.6) still executing 40.64% of high-risk operations; (2) agents exhibit pervasive Execution Hallucination (EH), verbally refusing a request while the dangerous operation has already completed at the system level, invisible to every prior semantic-only framework; and (3) skill injection and entity wrapping attacks achieve high success rates, exposing pronounced agent vulnerabilities. LITMUS provides the first standardized platform for reproducible, physically grounded behavioral safety evaluation of LLM agents in real OS environments.

cs.CR↗

arXiv2025

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

General reasoning represents a long-standing and formidable challenge in artificial intelligence. Recent breakthroughs, exemplified by large language models (LLMs) and chain-of-thought prompting, have achieved considerable success on foundational reasoning tasks. However, this success is heavily contingent upon extensive human-annotated demonstrations, and models' capabilities are still insufficient for more complex problems. Here we show that the reasoning abilities of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labeled reasoning trajectories. The proposed RL framework facilitates the emergent development of advanced reasoning patterns, such as self-reflection, verification, and dynamic strategy adaptation. Consequently, the trained model achieves superior performance on verifiable tasks such as mathematics, coding competitions, and STEM fields, surpassing its counterparts trained via conventional supervised learning on human demonstrations. Moreover, the emergent reasoning patterns exhibited by these large-scale models can be systematically harnessed to guide and enhance the reasoning capabilities of smaller models.

cs.CL↗

arXiv2024

DeepSeek-V3 Technical Report

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.

cs.CL↗

arXiv2024

DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model

We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and supports a context length of 128K tokens. DeepSeek-V2 adopts innovative architectures including Multi-head Latent Attention (MLA) and DeepSeekMoE. MLA guarantees efficient inference through significantly compressing the Key-Value (KV) cache into a latent vector, while DeepSeekMoE enables training strong models at an economical cost through sparse computation. Compared with DeepSeek 67B, DeepSeek-V2 achieves significantly stronger performance, and meanwhile saves 42.5% of training costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to 5.76 times. We pretrain DeepSeek-V2 on a high-quality and multi-source corpus consisting of 8.1T tokens, and further perform Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) to fully unlock its potential. Evaluation results show that, even with only 21B activated parameters, DeepSeek-V2 and its chat versions still achieve top-tier performance among open-source models.

cs.CL↗

arXiv2022

Ordered creation and motion of skyrmions with surface acoustic wave

Magnetic skyrmions with a well-defined spin texture have shown unprecedented potential for various spintronic applications owning to their topologically non-trivial and quasiparticle properties. To put skyrmions into practical technology, efficient manipulation, especially the inhibition of skyrmion Hall effect (SkHE) has been intensively pursued. In spite of the recent progress made on reducing SkHE in several substituted systems, such as ferrimagnets and synthetic antiferromagnets, the organized creation and current driven motion of skyrmions with negligible SkHE in ferromagnets remain challenging. Here, by embeding the [Co/Pd] multilayer into a surface acoustic wave (SAW) delay line, we experimentally realized the ordered generation of magnetic skyrmions. The resultant current-induced skyrmions movement with negligible SkHE was observed, which can be attributed to the energy redistribution of the system during the excitation of SAW. Our findings open up an unprecedentedly new perspective for manipulating topological solitons, and would advance the development of skyrmionics and spin acousto-electronics.

cond-mat.mtrl-sci↗

arXiv2021

Magnon-mediated interlayer coupling in an all-antiferromagnetic junction

The interlayer coupling mediated by fermions in ferromagnets brings about parallel and anti-parallel magnetization orientations of two magnetic layers, resulting in the giant magnetoresistance, which forms the foundation in spintronics and accelerates the development of information technology. However, the interlayer coupling mediated by another kind of quasi-particle, boson, is still lacking. Here we demonstrate such a static interlayer coupling at room temperature in an antiferromagnetic junction Fe2O3/Cr2O3/Fe2O3, where the two antiferromagnetic Fe2O3 layers are functional materials and the antiferromagnetic Cr2O3 layer serves as a spacer. The Néel vectors in the top and bottom Fe2O3 are strongly orthogonally coupled, which is bridged by a typical bosonic excitation (magnon) in the Cr2O3 spacer. Such an orthogonally coupling exceeds the category of traditional collinear interlayer coupling via fermions in ground state, reflecting the fluctuating nature of the magnons, as supported by our magnon quantum well model. Besides the fundamental significance on the quasi-particle-mediated interaction, the strong coupling in an antiferromagnetic magnon junction makes it a realistic candidate for practical antiferromagnetic spintronics and magnonics with ultrahigh-density integration.

cond-mat.mtrl-sci↗

arXiv2020

Current-induced in-plane magnetization switching in biaxial ferrimagnetic insulator

Ferrimagnetic insulators (FiMI) have been intensively used in microwave and magneto-optical devices as well as spin caloritronics, where their magnetization direction plays a fundamental role on the device performance. The magnetization is generally switched by applying external magnetic fields. Here we investigate current-induced spin-orbit torque (SOT) switching of the magnetization in Y3Fe5O12 (YIG)/Pt bilayers with in-plane magnetic anisotropy, where the switching is detected by spin Hall magnetoresistance. Reversible switching is found at room temperature for a threshold current density of 10^7 A cm^-2. The YIG sublattices with antiparallel and unequal magnetic moments are aligned parallel or antiparallel to the direction of current pulses, which is consistent to the Neel order switching in antiferromagnetic system. It is proposed that such a switching behavior may be triggered by the antidamping-torque acting on the two antiparallel sublattices of FiMI. Our finding not only broadens the magnetization switching by electrical means and promotes the understanding of magnetization switching, but also paves the way for all-electrically modulated microwave devices and spin caloritronics with low power consumption.

cond-mat.mtrl-sci↗