arXiv subjects

Xin Liu

Publications and source records attributed to Xin Liu.

At least 307 records · Page 17Linked to original sources

arXiv2025

UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning

The development of autonomous agents for graphical user interfaces (GUIs) presents major challenges in artificial intelligence. While recent advances in native agent models have shown promise by unifying perception, reasoning, action, and memory through end-to-end learning, open problems remain in data scalability, multi-turn reinforcement learning (RL), the limitations of GUI-only operation, and environment stability. In this technical report, we present UI-TARS-2, a native GUI-centered agent model that addresses these challenges through a systematic training methodology: a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment that integrates file systems and terminals, and a unified sandbox platform for large-scale rollouts. Empirical evaluation demonstrates that UI-TARS-2 achieves significant improvements over its predecessor UI-TARS-1.5. On GUI benchmarks, it reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld, outperforming strong baselines such as Claude and OpenAI agents. In game environments, it attains a mean normalized score of 59.8 across a 15-game suite-roughly 60% of human-level performance-and remains competitive with frontier proprietary models (e.g., OpenAI o3) on LMGame-Bench. Additionally, the model can generalize to long-horizon information-seeking tasks and software engineering benchmarks, highlighting its robustness across diverse agent tasks. Detailed analyses of training dynamics further provide insights into achieving stability and efficiency in large-scale agent RL. These results underscore UI-TARS-2's potential to advance the state of GUI agents and exhibit strong generalization to real-world interactive scenarios.

cs.AI↗

arXiv2025

LiquidGEMM: Hardware-Efficient W4A8 GEMM Kernel for High-Performance LLM Serving

Quantization is a critical technique for accelerating LLM inference by reducing memory footprint and improving computational efficiency. Among various schemes, 4-bit weight and 8-bit activation quantization (W4A8) offers a strong balance between accuracy and performance. However, existing W4A8 GEMM kernels fall short in practice due to inefficient dequantization on CUDA Cores, which cannot keep pace with the high throughput of Tensor Cores. In this paper, we present LiquidGEMM, a hardware-efficient W4A8 GEMM kernel for efficient LLM serving. LiquidGEMM designs two key techniques: LiquidQuant, a hardware-efficient quantization method that enables fast, overflow-safe dequantization using just two arithmetic instructions per four elements; and an implicit fine-grained pipeline that fully overlaps weight loading, dequantization, and MMA across warp groups without software synchronization or redundant memory traffic. Experimental results show that LiquidGEMM achieves up to 2.90x speedup over state-of-the-art W4A8 kernels and up to 4.94x end-to-end system-level speedup. Compared to various quantized GEMM kernels in NVIDIA TensorRT-LLM, LiquidGEMM delivers 1.12-1.63x performance gains, and achieves up to 1.63x system-level speedup.

cs.DC↗

arXiv2025

Balanced Actor Initialization: Stable RLHF Training of Distillation-Based Reasoning Models

The development of alignment and reasoning capabilities in large language models has seen remarkable progress through two paradigms: instruction tuning and reinforcement learning from human feedback (RLHF) alignment paradigm, and distillation-based reasoning fine-tuning paradigm. While both approaches prove effective independently, the third paradigm of applying RLHF to distillation-trained models presents significant challenges. Our investigation reveals two critical phenomena that emerge in this paradigm: Sequence Length Collapse, where language generation dramatically reduces during early RLHF training, and the Reward Hockey Stick Curve, featuring severe reward score drops followed by gradual recovery. These instabilities fundamentally compromise the model's alignment and reasoning capabilities. To address these challenges, we propose Balanced Actor Initialization (BAI), a two-stage weighted model merging approach. BAI first merges instruction-following and distillation-based reasoning fine-tuned models, then further combines this intermediate model with the pretrained model to preserve foundational knowledge. Through comprehensive experiments across diverse benchmarks and detailed analysis of training experiments, we demonstrate that BAI resolves Sequence Length Collapse, mitigates the Reward Hockey Stick Curve, and enables continuous sequence length improvement during training. Additionally, our analysis reveals that balanced merging ratios achieve optimal trade-offs between training stability and reasoning capability preservation. Our work provides the effective solution for stable training in this third paradigm, enabling more capable reasoning models that combine distillation efficiency with RLHF alignment.

cs.CL↗

arXiv2025

Edge dependent Josephson Diode effect in WTe$_{2}$-Based Josephson junction

The Josephson diode effect (JDE), a nonreciprocal supercurrent, is a cornerstone for future dissipationless electronics, yet achieving high efficiency in a simple device architecture remains a significant challenge. Here, we theoretically investigate the JDE in a junction based on monolayer 1T'-WTe$_2$. We first establish that different edge terminations of a WTe$_2$ nanoribbon lead to diverse electronic band structures, some of which host asymmetric edge states even with crystallographically equivalent terminations. This intrinsic asymmetry provides a natural platform for realizing the JDE. With a WTe$_2$-based Josephson junction, we demonstrate a significant JDE arising purely from these asymmetric edges when time-reversal symmetry is broken by a magnetic flux. While the efficiency of this edge-state-driven JDE is inherently limited, we discover a crucial mechanism for its enhancement: by tuning the chemical potential into the bulk bands, the interplay between edge and bulk transport channels boosts the maximum diode efficiency more than $50\%$. Furthermore, we show that this enhanced JDE is robust against moderate edge disorder. Our findings not only propose a novel route to achieve a highly efficient JDE using intrinsic material properties but also highlight the potential of engineered WTe$_2$ systems for developing advanced superconducting quantum devices.

cond-mat.supr-con↗

arXiv2025

Dynamic Collaboration of Multi-Language Models based on Minimal Complete Semantic Units

This paper investigates the enhancement of reasoning capabilities in language models through token-level multi-model collaboration. Our approach selects the optimal tokens from the next token distributions provided by multiple models to perform autoregressive reasoning. Contrary to the assumption that more models yield better results, we introduce a distribution distance-based dynamic selection strategy (DDS) to optimize the multi-model collaboration process. To address the critical challenge of vocabulary misalignment in multi-model collaboration, we propose the concept of minimal complete semantic units (MCSU), which is simple yet enables multiple language models to achieve natural alignment within the linguistic space. Experimental results across various benchmarks demonstrate the superiority of our method. The code will be available at https://github.com/Fanye12/DDS.

cs.AI↗

arXiv2025

TaiBai: A fully programmable brain-inspired processor with topology-aware efficiency

Brain-inspired computing has emerged as a promising paradigm to overcome the energy-efficiency limitations of conventional intelligent systems by emulating the brain's partitioned architecture and event-driven sparse computation. However, existing brain-inspired chips often suffer from rigid network topology constraints and limited neuronal programmability, hindering their adaptability. To address these challenges, we present TaiBai, an event-driven, programmable many-core brain-inspired processor that leverages temporal and spatial spike sparsity to minimize bandwidth and computational overhead. TaiBai chip contains three key features: First, a brain-inspired hierarchical topology encoding scheme is designed to flexibly support arbitrary network architectures while slashing storage overhead for large-scale networks; Second, a multi-granularity instruction set enables programmability of brain-like spiking neuron or synapses with various dynamics and on-chip learning rules; Third, a co-designed compiler stack optimizes task mapping and resource allocation. After evaluating across various tasks, such as speech recognition, ECG classification, and cross-day brain-computer interface decoding, we found spiking neural networks embedded on the TaiBai chip could achieve more than 200 times higher energy efficiency than a standard NVIDIA RTX 3090 GPU at a comparable accuracy. These results demonstrated its high potentiation as a scalable, programmable, and ultra-efficient solution for both multi-scale brain simulation and brain-inspired computation.

cs.AR↗

arXiv2025

Segregation-driven cross-slip mechanism of shockley partials in the gamma prime phase of CoNi-based superalloys

In general, the cross-slip of superdislocations (a/2<011>) from {111} planes to {001} planes has been frequently observed in superalloys, accompanied by the formation of an antiphase boundary (APB) and driven by thermal activation. However, no prior studies have evidenced the occurrence of Shockley partial dislocation (a/6<112>) cross-slip within the gamma prime phase of superalloys. In this work, we present a newly observed cross-slip phenomenon: the Shockley partial dislocations cross-slip from one {111} plane to another {111} conjugate plane, facilitated by the formation of a stair-rod dislocation in the ordered gamma prime phase of a CoNi-based superalloy. Compression tests were conducted at 850 degrees Celsius with a strain rate of 10^-4 s^-1. Defects such as stacking faults and dislocations, along with the associated chemical fluctuations, were characterized using high-resolution scanning transmission electron microscopy (HRSTEM) and energy-dispersive X-ray spectroscopy (EDS). Elemental segregation was found to reduce the activation energy required for cross-slip by decreasing the energies of stacking faults and dislocations. In addition to elemental segregation, local stress concentrations, arising from the combined effects of applied stress, shearing dislocations within the gamma prime phase, and dislocation pile-ups, also play a critical role in triggering cross-slip. The formation of sessile stair-rod dislocations via this newly identified Shockley partial cross-slip in the gamma prime phase is beneficial for enhancing the high-temperature deformation resistance of the alloy by increasing the critical resolved shear stress required for further plastic deformation.

cond-mat.mtrl-sci↗

arXiv2025

On the Role of Entity and Event Level Conceptualization in Generalizable Reasoning: A Survey of Tasks, Methods, Applications, and Future Directions

Conceptualization, a fundamental element of human cognition, plays a pivotal role in human generalizable reasoning. Generally speaking, it refers to the process of sequentially abstracting specific instances into higher-level concepts and then forming abstract knowledge that can be applied in unfamiliar or novel situations. This enhances models' inferential capabilities and supports the effective transfer of knowledge across various domains. Despite its significance, the broad nature of this term has led to inconsistencies in understanding conceptualization across various works, as there exists different types of instances that can be abstracted in a wide variety of ways. There is also a lack of a systematic overview that comprehensively examines existing works on the definition, execution, and application of conceptualization to enhance reasoning tasks. In this paper, we address these gaps by first proposing a categorization of different types of conceptualizations into four levels based on the types of instances being conceptualized, in order to clarify the term and define the scope of our work. Then, we present the first comprehensive survey of over 150 papers, surveying various definitions, resources, methods, and downstream applications related to conceptualization into a unified taxonomy, with a focus on the entity and event levels. Furthermore, we shed light on potential future directions in this field and hope to garner more attention from the community.

cs.CL↗

arXiv2025

CU-JADE: A Method for Traversing Extinction Jumps along the Line of Sight

Although interstellar dust extinction serves as a powerful distance estimator, the solar system's location within the Galactic plane complicates distance determinations, especially for molecular clouds (MCs) at varying distances along the line of sight (LoS). The presence of complex extinction patterns along the LoS introduces degeneracies, resulting in less accurate distance measurements to overlapping MCs in crowded regions of the Galactic plane. In this study, we develop the CUSUM-based Jump-point Analysis for Distance Estimation (CU-JADE), a novel method designed to help mitigate these observational challenges. The key strengths of CU-JADE include: (1) sensitivity to detect abrupt jumps in Distance-$A_λ$ ($D$-$A$) datasets, (2) minimal systematic errors as demonstrated on both mock and observed data, and (3) the ability to combine CUSUM analysis with multiwavelength data to improve the completeness of distance measurements for nearby gas structures, even for extinction values as low as $ΔA_{V} \gtrsim 0.15$ mag. By combining CO survey data with a large sample of stars characterized by high-precision parallaxes and extinctions, we uncovered the multilayered molecular gas distribution in the high-latitude Cepheus region. We also determined accurate distances to MCs beyond the Cygnus Rift by analyzing the intricate structure of gas and extinction within the Galactic plane. Additionally, we constructed a full-sky 3D extinction map extending to 4 kpc, which provides critical insights into dense interstellar medium components dominated by molecular hydrogen. These results advance our understanding of the spatial distribution and physical properties of MCs across the Milky Way.

astro-ph.GA↗

arXiv2025

Study of $B_c \to χ_{cJ}\ (P, V)$ decays in the improved perturbative QCD formalism

Motivated by the recent LHCb measurements of the ratios between the branching ratios (BRs), $R_{χ_{c2}/J/ψ} \equiv {\rm BR}(B_c^+ \to χ_{c2} π^+)/{\rm BR}(B_c^+ \to J/ψπ^+)$ and $R_{χ_{c1}/χ_{c2}} \equiv {\rm BR}(B_c^+ \to χ_{c1} π^+)/{\rm BR}(B_c^+ \to χ_{c2} π^+)$, we analyze the decays $B_c^+ \to χ_{cJ} (P, V)^+$ in the improved perturbative QCD (iPQCD) formalism, where $χ_{cJ}$ denotes the $p$-wave charmonia with $J=0, 1, 2$, and $P (V)$ denotes the pseudoscalars $π$ and $K$ (the vectors $ρ$ and $K^*$). Our results $R_{χ_{c2}/J/ψ} = 0.32 \pm 0.05$ and $R_{χ_{c1}/χ_{c2}}= 0.11 \pm 0.01$ at leading order in the strong coupling $α_s$ agree well with the data within uncertainties. The result $R_{χ_{c0}/J/ψ} \equiv {\rm BR}(B_c^+ \to χ_{c0} π^+)/{\rm BR}(B_c^+ \to J/ψπ^+) = 0.17 \pm 0.02$ is lower than $1.41^{+0.50}_{-0.45}$ inferred by the $B_c^+ \to χ_{c0} π^+$ measurement. The large longitudinal polarization fractions imply that these components dominate the $B_c^+ \to χ_{c1,c2} ρ^+$ BRs. A peculiar constructive (destructive) interference between the twist-2 and twist-3 contributions is identified in the $B_c^+\to χ_{c1} ρ^+$ ($B_c^+\to χ_{c1} π^+$) mode, but not in the corresponding ones with $χ_{c0,c2}$ mesons. Inputting the BRs of the strong decays $χ_{c0,c2} \to π^+ π^- / K^+ K^-$ and $χ_{cJ} \to π^+π^-(π^+ π^-/K^+K^-)$ reported by the BESIII Collaboration, we obtain those of $B_c^+ \to χ_{cJ} (ρ, π)^+$ followed by secondary decay chains under the narrow-width approximation. All the above iPQCD predictions can be confronted by more precise experiments in the future to deepen our understanding of QCD dynamics in $B_c\toχ_{cJ}$ transitions.

hep-ph↗

arXiv2025

QuarkMed Medical Foundation Model Technical Report

Recent advancements in large language models have significantly accelerated their adoption in healthcare applications, including AI-powered medical consultations, diagnostic report assistance, and medical search tools. However, medical tasks often demand highly specialized knowledge, professional accuracy, and customization capabilities, necessitating a robust and reliable foundation model. QuarkMed addresses these needs by leveraging curated medical data processing, medical-content Retrieval-Augmented Generation (RAG), and a large-scale, verifiable reinforcement learning pipeline to develop a high-performance medical foundation model. The model achieved 70% accuracy on the Chinese Medical Licensing Examination, demonstrating strong generalization across diverse medical benchmarks. QuarkMed offers a powerful yet versatile personal medical AI solution, already serving over millions of users at ai.quark.cn.

cs.AI↗

arXiv2025

Distribution-Specific Learning for Joint Salient and Camouflaged Object Detection

Salient object detection (SOD) and camouflaged object detection (COD) are two closely related but distinct computer vision tasks. Although both are class-agnostic segmentation tasks that map from RGB space to binary space, the former aims to identify the most salient objects in the image, while the latter focuses on detecting perfectly camouflaged objects that blend into the background in the image. These two tasks exhibit strong contradictory attributes. Previous works have mostly believed that joint learning of these two tasks would confuse the network, reducing its performance on both tasks. However, here we present an opposite perspective: with the correct approach to learning, the network can simultaneously possess the capability to find both salient and camouflaged objects, allowing both tasks to benefit from joint learning. We propose SCJoint, a joint learning scheme for SOD and COD tasks, assuming that the decoding processes of SOD and COD have different distribution characteristics. The key to our method is to learn the respective means and variances of the decoding processes for both tasks by inserting a minimal amount of task-specific learnable parameters within a fully shared network structure, thereby decoupling the contradictory attributes of the two tasks at a minimal cost. Furthermore, we propose a saliency-based sampling strategy (SBSS) to sample the training set of the SOD task to balance the training set sizes of the two tasks. In addition, SBSS improves the training set quality and shortens the training time. Based on the proposed SCJoint and SBSS, we train a powerful generalist network, named JoNet, which has the ability to simultaneously capture both ``salient" and ``camouflaged". Extensive experiments demonstrate the competitive performance and effectiveness of our proposed method. The code is available at https://github.com/linuxsino/JoNet.

cs.CV↗

arXiv2025

VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

Recent advances in large language models (LLMs) have driven impressive progress in omni-modal understanding and generation. However, training omni-modal LLMs remains a significant challenge due to the heterogeneous model architectures required to process diverse modalities, necessitating sophisticated system design for efficient large-scale training. Existing frameworks typically entangle model definition with parallel logic, incurring limited scalability and substantial engineering overhead for end-to-end omni-modal training. We present VeOmni, a modular and efficient training framework to accelerate the development of omni-modal LLMs. VeOmni introduces model-centric distributed recipes that decouples communication from computation, enabling efficient 3D parallelism on omni-modal LLMs. VeOmni also features a flexible configuration interface supporting seamless integration of new modalities with minimal code change. Using VeOmni, a omni-modal mixture-of-experts (MoE) model with 30B parameters can be trained with over 2,800 tokens/sec/GPU throughput and scale to 160K context lengths via 3D parallelism on 128 GPUs, showcasing its superior efficiency and scalability for training large omni-modal LLMs.

cs.CL↗

arXiv2025

EgoTrigger: Toward Audio-Driven Image Capture for Human Memory Enhancement in All-Day Energy-Efficient Smart Glasses

All-day smart glasses are likely to emerge as platforms capable of continuous contextual sensing, uniquely positioning them for unprecedented assistance in our daily lives. Integrating the multi-modal AI agents required for human memory enhancement while performing continuous sensing, however, presents a major energy efficiency challenge for all-day usage. Achieving this balance requires intelligent, context-aware sensor management. Our approach, EgoTrigger, leverages audio cues from the microphone to selectively activate power-intensive cameras, enabling efficient sensing while preserving substantial utility for human memory enhancement. EgoTrigger uses a lightweight audio model (YAMNet) and a custom classification head to trigger image capture from hand-object interaction (HOI) audio cues, such as the sound of a drawer opening or a medication bottle being opened. In addition to evaluating on the QA-Ego4D dataset, we introduce and evaluate on the Human Memory Enhancement Question-Answer (HME-QA) dataset. Our dataset contains 340 human-annotated first-person QA pairs from full-length Ego4D videos that were curated to ensure that they contained audio, focusing on HOI moments critical for contextual understanding and memory. Our results show EgoTrigger can use 54% fewer frames on average, significantly saving energy in both power-hungry sensing components (e.g., cameras) and downstream operations (e.g., wireless transmission), while achieving comparable performance on datasets for an episodic memory task. We believe this context-aware triggering strategy represents a promising direction for enabling energy-efficient, functional smart glasses capable of all-day use -- supporting applications like helping users recall where they placed their keys or information about their routine activities (e.g., taking medications).

cs.CV↗

arXiv2025

Oldie but Goodie: Re-illuminating Label Propagation on Graphs with Partially Observed Features

In real-world graphs, we often encounter missing feature situations where a few or the majority of node features, e.g., sensitive information, are missed. In such scenarios, directly utilizing Graph Neural Networks (GNNs) would yield sub-optimal results in downstream tasks such as node classification. Despite the emergence of a few GNN-based methods attempting to mitigate its missing situation, when only a few features are available, they rather perform worse than traditional structure-based models. To this end, we propose a novel framework that further illuminates the potential of classical Label Propagation (Oldie), taking advantage of Feature Propagation, especially when only a partial feature is available. Now called by GOODIE, it takes a hybrid approach to obtain embeddings from the Label Propagation branch and Feature Propagation branch. To do so, we first design a GNN-based decoder that enables the Label Propagation branch to output hidden embeddings that align with those of the FP branch. Then, GOODIE automatically captures the significance of structure and feature information thanks to the newly designed Structure-Feature Attention. Followed by a novel Pseudo-Label contrastive learning that differentiates the contribution of each positive pair within pseudo-labels originating from the LP branch, GOODIE outputs the final prediction for the unlabeled nodes. Through extensive experiments, we demonstrate that our proposed model, GOODIE, outperforms the existing state-of-the-art methods not only when only a few features are available but also in abundantly available situations. Source code of GOODIE is available at: https://github.com/SukwonYun/GOODIE.

cs.LG↗

arXiv2025

Contact Sensors to Remote Cameras: Quantifying Cardiorespiratory Coupling in High-Altitude Exercise Recovery

Cardiorespiratory coupling (CRC) captures the dynamic interaction between the cardiac and respiratory systems--an interaction strengthened by physical exercise and linked to improved physiological function. We examined CRC at high altitude in two states, rest and post-exercise recovery, and found significant differences (p < 0.05). Quantitative analysis revealed that recovery involved more frequent yet less stable episodes of synchronization between respiration and pulse. Furthermore, we explored the feasibility of non-contact CRC measurement with remote photoplethysmography (rPPG), observing a strong correlation with oximeter-based metrics (Pearson r = 0.96). These findings highlight the potential of CRC as a sensitive marker for autonomic regulation and its future application in contactless monitoring. Source code is available at GitHub: https://github.com/McJackTang/CRC.

cs.CE↗

arXiv2025

Latent Knowledge Scalpel: Precise and Massive Knowledge Editing for Large Language Models

Large Language Models (LLMs) often retain inaccurate or outdated information from pre-training, leading to incorrect predictions or biased outputs during inference. While existing model editing methods can address this challenge, they struggle with editing large amounts of factual information simultaneously and may compromise the general capabilities of the models. In this paper, our empirical study demonstrates that it is feasible to edit the internal representations of LLMs and replace the entities in a manner similar to editing natural language inputs. Based on this insight, we introduce the Latent Knowledge Scalpel (LKS), an LLM editor that manipulates the latent knowledge of specific entities via a lightweight hypernetwork to enable precise and large-scale editing. Experiments conducted on Llama-2 and Mistral show even with the number of simultaneous edits reaching 10,000, LKS effectively performs knowledge editing while preserving the general abilities of the edited LLMs. Code is available at: https://github.com/Linuxin-xxx/LKS.

cs.LG↗

arXiv2025

Learned Off-aperture Encoding for Wide Field-of-view RGBD Imaging

End-to-end (E2E) designed imaging systems integrate coded optical designs with decoding algorithms to enhance imaging fidelity for diverse visual tasks. However, existing E2E designs encounter significant challenges in maintaining high image fidelity at wide fields of view, due to high computational complexity, as well as difficulties in modeling off-axis wave propagation while accounting for off-axis aberrations. In particular, the common approach of placing the encoding element into the aperture or pupil plane results in only a global control of the wavefront. To overcome these limitations, this work explores an additional design choice by positioning a DOE off-aperture, enabling a spatial unmixing of the degrees of freedom and providing local control over the wavefront over the image plane. Our approach further leverages hybrid refractive-diffractive optical systems by linking differentiable ray and wave optics modeling, thereby optimizing depth imaging quality and demonstrating system versatility. Experimental results reveal that the off-aperture DOE enhances the imaging quality by over 5 dB in PSNR at a FoV of approximately $45^\circ$ when paired with a simple thin lens, outperforming traditional on-aperture systems. Furthermore, we successfully recover color and depth information at nearly $28^\circ$ FoV using off-aperture DOE configurations with compound optics. Physical prototypes for both applications validate the effectiveness and versatility of the proposed method.

eess.IV↗