arXiv subjects

Yuxiang Wei

Publications and source records attributed to Yuxiang Wei.

At least 37 records · Page 2Linked to original sources

arXiv2025

SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution

The recent DeepSeek-R1 release has demonstrated the immense potential of reinforcement learning (RL) in enhancing the general reasoning capabilities of large language models (LLMs). While DeepSeek-R1 and other follow-up work primarily focus on applying RL to competitive coding and math problems, this paper introduces SWE-RL, the first approach to scale RL-based LLM reasoning for real-world software engineering. Leveraging a lightweight rule-based reward (e.g., the similarity score between ground-truth and LLM-generated solutions), SWE-RL enables LLMs to autonomously recover a developer's reasoning processes and solutions by learning from extensive open-source software evolution data -- the record of a software's entire lifecycle, including its code snapshots, code changes, and events such as issues and pull requests. Trained on top of Llama 3, our resulting reasoning model, Llama3-SWE-RL-70B, achieves a 41.0% solve rate on SWE-bench Verified -- a human-verified collection of real-world GitHub issues. To our knowledge, this is the best performance reported for medium-sized (<100B) LLMs to date, even comparable to leading proprietary LLMs like GPT-4o. Surprisingly, despite performing RL solely on software evolution data, Llama3-SWE-RL has even emerged with generalized reasoning skills. For example, it shows improved results on five out-of-domain tasks, namely, function coding, library use, code reasoning, mathematics, and general language understanding, whereas a supervised-finetuning baseline even leads to performance degradation on average. Overall, SWE-RL opens up a new direction to improve the reasoning capabilities of LLMs through reinforcement learning on massive software engineering data.

cs.SE↗

arXiv2025

Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly?

Large Language Models (LLMs) are reshaping almost all industries, including software engineering. In recent years, a number of LLM agents have been proposed to solve real-world software problems. Such software agents are typically equipped with a suite of coding tools and can autonomously decide the next actions to form complete trajectories to solve end-to-end software tasks. While promising, they typically require dedicated design and may still be suboptimal, since it can be extremely challenging and costly to exhaust the entire agent scaffold design space. Recognizing that software agents are inherently software themselves that can be further refined/modified, researchers have proposed a number of self-improving software agents recently, including the Darwin-Gödel Machine (DGM). Meanwhile, such self-improving agents require costly offline training on specific benchmarks and may not generalize well across different LLMs or benchmarks. In this paper, we propose Live-SWE-agent, the first live software agent that can autonomously and continuously evolve itself on-the-fly during runtime when solving real-world software problems. More specifically, Live-SWE-agent starts with the most basic agent scaffold with only access to bash tools (e.g., mini-SWE-agent), and autonomously evolves its own scaffold implementation while solving real-world software problems. Our evaluation on the widely studied SWE-bench Verified benchmark shows that LIVE-SWE-AGENT can achieve an impressive solve rate of 77.4% without test-time scaling, outperforming all existing software agents, including the best proprietary solution. Moreover, Live-SWE-agent outperforms state-of-the-art manually crafted software agents on the recent SWE-Bench Pro benchmark, achieving the best-known solve rate of 45.8%.

cs.SE↗

arXiv2025

PurpCode: Reasoning for Safer Code Generation

We introduce PurpCode, the first post-training recipe for training safe code reasoning models towards generating secure code and defending against malicious cyberactivities. PurpCode trains a reasoning model in two stages: (i) Rule Learning, which explicitly teaches the model to reference cybersafety rules to generate vulnerability-free code and to avoid facilitating malicious cyberactivities; and (ii) Reinforcement Learning, which optimizes model safety and preserves model utility through diverse, multi-objective reward mechanisms. To empower the training pipelines with comprehensive cybersafety data, we conduct internal red-teaming to synthesize comprehensive and high-coverage prompts based on real-world tasks for inducing unsafe cyberactivities in the model. Based on PurpCode, we develop a reasoning-based coding model, namely PurpCode-32B, which demonstrates state-of-the-art cybersafety, outperforming various frontier models. Meanwhile, our alignment method decreases the model overrefusal rates in both general and cybersafety-specific scenarios, while preserving model utility in both code generation and common security knowledge.

cs.CR↗

arXiv2025

MoRE-Brain: Routed Mixture of Experts for Interpretable and Generalizable Cross-Subject fMRI Visual Decoding

Decoding visual experiences from fMRI offers a powerful avenue to understand human perception and develop advanced brain-computer interfaces. However, current progress often prioritizes maximizing reconstruction fidelity while overlooking interpretability, an essential aspect for deriving neuroscientific insight. To address this gap, we propose MoRE-Brain, a neuro-inspired framework designed for high-fidelity, adaptable, and interpretable visual reconstruction. MoRE-Brain uniquely employs a hierarchical Mixture-of-Experts architecture where distinct experts process fMRI signals from functionally related voxel groups, mimicking specialized brain networks. The experts are first trained to encode fMRI into the frozen CLIP space. A finetuned diffusion model then synthesizes images, guided by expert outputs through a novel dual-stage routing mechanism that dynamically weighs expert contributions across the diffusion process. MoRE-Brain offers three main advancements: First, it introduces a novel Mixture-of-Experts architecture grounded in brain network principles for neuro-decoding. Second, it achieves efficient cross-subject generalization by sharing core expert networks while adapting only subject-specific routers. Third, it provides enhanced mechanistic insight, as the explicit routing reveals precisely how different modeled brain regions shape the semantic and spatial attributes of the reconstructed image. Extensive experiments validate MoRE-Brain's high reconstruction fidelity, with bottleneck analyses further demonstrating its effective utilization of fMRI signals, distinguishing genuine neural decoding from over-reliance on generative priors. Consequently, MoRE-Brain marks a substantial advance towards more generalizable and interpretable fMRI-based visual decoding. Code will be publicly available soon: https://github.com/yuxiangwei0808/MoRE-Brain.

cs.LG↗

arXiv2025

CWM: An Open-Weights LLM for Research on Code Generation with World Models

We release Code World Model (CWM), a 32-billion-parameter open-weights LLM, to advance research on code generation with world models. To improve code understanding beyond what can be learned from training on static code alone, we mid-train CWM on a large amount of observation-action trajectories from Python interpreter and agentic Docker environments, and perform extensive multi-task reasoning RL in verifiable coding, math, and multi-turn software engineering environments. With CWM, we provide a strong testbed for researchers to explore the opportunities world modeling affords for improving code generation with reasoning and planning in computational environments. We present first steps of how world models can benefit agentic coding, enable step-by-step simulation of Python code execution, and show early results of how reasoning can benefit from the latter. CWM is a dense, decoder-only LLM trained with a context size of up to 131k tokens. Independent of its world modeling capabilities, CWM offers strong performance on general coding and math tasks: it reaches pass@1 scores of 65.8% on SWE-bench Verified (with test-time scaling), 68.6% on LiveCodeBench, 96.6% on Math-500, and 76.0% on AIME 2024. To support further research on code world modeling, we release model checkpoints after mid-training, SFT, and RL.

cs.SE↗

arXiv2025

Visual Instance-aware Prompt Tuning

Visual Prompt Tuning (VPT) has emerged as a parameter-efficient fine-tuning paradigm for vision transformers, with conventional approaches utilizing dataset-level prompts that remain the same across all input instances. We observe that this strategy results in sub-optimal performance due to high variance in downstream datasets. To address this challenge, we propose Visual Instance-aware Prompt Tuning (ViaPT), which generates instance-aware prompts based on each individual input and fuses them with dataset-level prompts, leveraging Principal Component Analysis (PCA) to retain important prompting information. Moreover, we reveal that VPT-Deep and VPT-Shallow represent two corner cases based on a conceptual understanding, in which they fail to effectively capture instance-specific information, while random dimension reduction on prompts only yields performance between the two extremes. Instead, ViaPT overcomes these limitations by balancing dataset-level and instance-level knowledge, while reducing the amount of learnable parameters compared to VPT-Deep. Extensive experiments across 34 diverse datasets demonstrate that our method consistently outperforms state-of-the-art baselines, establishing a new paradigm for analyzing and optimizing visual prompts for vision transformers.

cs.CV↗

arXiv2025

Next-Generation Travel Demand Modeling with a Generative Framework for Household Activity Coordination

Travel demand models are critical tools for planning, policy, and mobility system design. Traditional activity-based models (ABMs), although grounded in behavioral theories, often rely on simplified rules and assumptions, and are costly to develop and difficult to adapt across different regions. This paper presents a learning-based travel demand modeling framework that synthesizes household-coordinated daily activity patterns based on a household's socio-demographic profiles. The whole framework integrates population synthesis, coordinated activity generation, location assignment, and large-scale microscopic traffic simulation into a unified system. It is fully generative, data-driven, scalable, and transferable to other regions. A full-pipeline implementation is conducted in Los Angeles with a 10 million population. Comprehensive validation shows that the model closely replicates real-world mobility patterns and matches the performance of legacy ABMs with significantly reduced modeling cost and greater scalability. With respect to the SCAG ABM benchmark, the origin-destination matrix achieves a cosine similarity of 0.97, and the daily vehicle miles traveled (VMT) in the network yields a 0.006 Jensen-Shannon Divergence (JSD) and a 9.8% mean absolute percentage error (MAPE). When compared to real-world observations from Caltrans PeMS, the evaluation on corridor-level traffic speed and volume reaches a 0.001 JSD and a 6.11% MAPE.

cs.LG↗

arXiv2025

Bi-LSTM based Multi-Agent DRL with Computation-aware Pruning for Agent Twins Migration in Vehicular Embodied AI Networks

With the advancement of large language models and embodied Artificial Intelligence (AI) in the intelligent transportation scenarios, the combination of them in intelligent transportation spawns the Vehicular Embodied AI Network (VEANs). In VEANs, Autonomous Vehicles (AVs) are typical agents whose local advanced AI applications are defined as vehicular embodied AI agents, enabling capabilities such as environment perception and multi-agent collaboration. Due to computation latency and resource constraints, the local AI applications and services running on vehicular embodied AI agents need to be migrated, and subsequently referred to as vehicular embodied AI agent twins, which drive the advancement of vehicular embodied AI networks to offload intensive tasks to Roadside Units (RSUs), mitigating latency problems while maintaining service quality. Recognizing workload imbalance among RSUs in traditional approaches, we model AV-RSU interactions as a Stackelberg game to optimize bandwidth resource allocation for efficient migration. A Tiny Multi-Agent Bidirectional LSTM Proximal Policy Optimization (TMABLPPO) algorithm is designed to approximate the Stackelberg equilibrium through decentralized coordination. Furthermore, a personalized neural network pruning algorithm based on Path eXclusion (PX) dynamically adapts to heterogeneous AV computation capabilities by identifying task-critical parameters in trained models, reducing model complexity with less performance degradation. Experimental validation confirms the algorithm's effectiveness in balancing system load and minimizing delays, demonstrating significant improvements in vehicular embodied AI agent deployment.

cs.GT↗

arXiv2025

TinyMA-IEI-PPO: Exploration Incentive-Driven Multi-Agent DRL with Self-Adaptive Pruning for Vehicular Embodied AI Agent Twins Migration

Embodied Artificial Intelligence (EAI) addresses autonomous driving challenges in Vehicular Embodied AI Networks (VEANETs) through multi-modal perception, adaptive decision-making, and hardware-software co-scheduling. However, the computational demands of virtual services and the inherent mobility of autonomous vehicles (AVs) necessitate real-time migration of Vehicular Embodied Agent AI Twins (VEAATs) between resource-constrained Roadside Units (RSUs). This paper proposes a novel framework for efficient VEAAT migration in VEANETs, combining a multi-leader multi-follower (MLMF) Stackelberg game-theoretic incentive mechanism with a tiny multi-agent deep reinforcement learning (MADRL) algorithm. First, We propose an virtual immersive experience-driven utility model that captures AV-RSU dynamic interactions by integrating AVs' social influence, service complementarity and substitutability, and RSUs' resource allocation strategies to optimize VEAAT migration decisions. Second, to enhance training efficiency and enable efficient deployment on computation-constrained AVs while preserving exploration-exploitation performance, we propose TinyMA-IEI-PPO, a self-adaptive dynamic structured pruning algorithm that dynamically adjusts neuron importance based on agents' exploration incentives. Numerical results demonstrate that our approach achieves convergence comparable to baseline models and closely approximates the Stackelberg equilibrium.

cs.MA↗

arXiv2025

4D Multimodal Co-attention Fusion Network with Latent Contrastive Alignment for Alzheimer's Diagnosis

Multimodal neuroimaging provides complementary structural and functional insights into both human brain organization and disease-related dynamics. Recent studies demonstrate enhanced diagnostic sensitivity for Alzheimer's disease (AD) through synergistic integration of neuroimaging data (e.g., sMRI, fMRI) with behavioral cognitive scores tabular data biomarkers. However, the intrinsic heterogeneity across modalities (e.g., 4D spatiotemporal fMRI dynamics vs. 3D anatomical sMRI structure) presents critical challenges for discriminative feature fusion. To bridge this gap, we propose M2M-AlignNet: a geometry-aware multimodal co-attention network with latent alignment for early AD diagnosis using sMRI and fMRI. At the core of our approach is a multi-patch-to-multi-patch (M2M) contrastive loss function that quantifies and reduces representational discrepancies via geometry-weighted patch correspondence, explicitly aligning fMRI components across brain regions with their sMRI structural substrates without one-to-one constraints. Additionally, we propose a latent-as-query co-attention module to autonomously discover fusion patterns, circumventing modality prioritization biases while minimizing feature redundancy. We conduct extensive experiments to confirm the effectiveness of our method and highlight the correspondance between fMRI and sMRI as AD biomarkers.

cs.MM↗

arXiv2025

EfficientMT: Efficient Temporal Adaptation for Motion Transfer in Text-to-Video Diffusion Models

The progress on generative models has led to significant advances on text-to-video (T2V) generation, yet the motion controllability of generated videos remains limited. Existing motion transfer methods explored the motion representations of reference videos to guide generation. Nevertheless, these methods typically rely on sample-specific optimization strategy, resulting in high computational burdens. In this paper, we propose EfficientMT, a novel and efficient end-to-end framework for video motion transfer. By leveraging a small set of synthetic paired motion transfer samples, EfficientMT effectively adapts a pretrained T2V model into a general motion transfer framework that can accurately capture and reproduce diverse motion patterns. Specifically, we repurpose the backbone of the T2V model to extract temporal information from reference videos, and further propose a scaler module to distill motion-related information. Subsequently, we introduce a temporal integration mechanism that seamlessly incorporates reference motion features into the video generation process. After training on our self-collected synthetic paired samples, EfficientMT enables general video motion transfer without requiring test-time optimization. Extensive experiments demonstrate that our EfficientMT outperforms existing methods in efficiency while maintaining flexible motion controllability. Our code will be available https://github.com/PrototypeNx/EfficientMT.

cs.CV↗

arXiv2025

Hierarchical Spatio-Temporal State-Space Modeling for fMRI Analysis

Recent advances in deep learning structured state space models, especially the Mamba architecture, have demonstrated remarkable performance improvements while maintaining linear complexity. In this study, we introduce functional spatiotemporal Mamba (FST-Mamba), a Mamba-based model designed for discovering neurological biomarkers using functional magnetic resonance imaging (fMRI). We focus on dynamic functional network connectivity (dFNC) derived from fMRI and propose a hierarchical spatiotemporal Mamba-based network that processes spatial and temporal information separately using Mamba-based encoders. Leveraging the topological uniqueness of the FNC matrix, we introduce a component-wise varied-scale aggregation (CVA) mechanism to aggregate connectivity across individual components within brain networks, enabling the model to capture component-level and network-level information. Additionally, we propose symmetric rotary position encoding (SymRope) to encode the relative positions of each functional connection while considering the symmetric nature of the FNC matrix. Experimental results demonstrate significant improvements in the proposed FST-Mamba model on various brain-based classification and regression tasks. We further show brain connectivities and dynamics that are crucial for the prediction. Our work reveals the substantial potential of attention-free sequence modeling in brain discovery. The codes are publicly available here: https://github.com/yuxiangwei0808/FunctionalMamba/tree/main.

cs.LG↗

arXiv2025

Personalized Image Generation with Deep Generative Models: A Decade Survey

Recent advancements in generative models have significantly facilitated the development of personalized content creation. Given a small set of images with user-specific concept, personalized image generation allows to create images that incorporate the specified concept and adhere to provided text descriptions. Due to its wide applications in content creation, significant effort has been devoted to this field in recent years. Nonetheless, the technologies used for personalization have evolved alongside the development of generative models, with their distinct and interrelated components. In this survey, we present a comprehensive review of generalized personalized image generation across various generative models, including traditional GANs, contemporary text-to-image diffusion models, and emerging multi-model autoregressive models. We first define a unified framework that standardizes the personalization process across different generative models, encompassing three key components, i.e., inversion spaces, inversion methods, and personalization schemes. This unified framework offers a structured approach to dissecting and comparing personalization techniques across different generative architectures. Building upon this unified framework, we further provide an in-depth analysis of personalization techniques within each generative model, highlighting their unique contributions and innovations. Through comparative analysis, this survey elucidates the current landscape of personalized image generation, identifying commonalities and distinguishing features among existing methods. Finally, we discuss the open challenges in the field and propose potential directions for future research. We keep tracing related works at https://github.com/csyxwei/Awesome-Personalized-Image-Generation.

cs.CV↗

arXiv2025

ACE: Anti-Editing Concept Erasure in Text-to-Image Models

Recent advance in text-to-image diffusion models have significantly facilitated the generation of high-quality images, but also raising concerns about the illegal creation of harmful content, such as copyrighted images. Existing concept erasure methods achieve superior results in preventing the production of erased concept from prompts, but typically perform poorly in preventing undesired editing. To address this issue, we propose an Anti-Editing Concept Erasure (ACE) method, which not only erases the target concept during generation but also filters out it during editing. Specifically, we propose to inject the erasure guidance into both conditional and the unconditional noise prediction, enabling the model to effectively prevent the creation of erasure concepts during both editing and generation. Furthermore, a stochastic correction guidance is introduced during training to address the erosion of unrelated concepts. We conducted erasure editing experiments with representative editing methods (i.e., LEDITS++ and MasaCtrl) to erase IP characters, and the results indicate that our ACE effectively filters out target concepts in both types of edits. Additional experiments on erasing explicit concepts and artistic styles further demonstrate that our ACE performs favorably against state-of-the-art methods. Our code will be publicly available at https://github.com/120L020904/ACE.

cs.CV↗

arXiv2024

Practical Program Repair in the Era of Large Pre-trained Language Models

Automated Program Repair (APR) aims to help developers automatically patch software bugs. However, current state-of-the-art traditional and learning-based APR techniques face the problem of limited patch variety, failing to fix complicated bugs. This is mainly due to the reliance on bug-fixing datasets to craft fix templates or directly predict potential patches. Large Pre-Trained Language Models (PLMs), trained using billions of text/code tokens, can potentially help avoid this issue. Very recently, researchers have directly leveraged PLMs for APR without relying on any bug-fixing datasets. Meanwhile, such existing work either failed to include state-of-the-art PLMs or was not evaluated on realistic datasets. In this work, we perform the first extensive study on directly applying PLMs for APR. We select 9 recent state-of-the-art PLMs, including both generative and infilling models, ranging from 125M to 20B in size. We designed 3 different repair settings to evaluate the different ways we can use PLMs to generate patches. We apply the PLMs under these repair settings on 5 datasets across 3 different languages and compare different PLMs in the number of bugs fixed, generation speed and compilation rate. Our study demonstrates that directly applying state-of-the-art PLMs can already substantially outperform all existing APR techniques on all our datasets. Among the studied PLMs, the scaling effect exists for APR where larger models tend to achieve better performance. Also, we show for the first time that suffix code after the buggy line (adopted in infilling-style APR) is important in not only generating more fixes but more patches with higher compilation rate. Besides patch generation, the PLMs consider correct patches to be more natural than other ones, and can even be leveraged for effective patch ranking or patch correctness checking.

cs.SE↗

arXiv2024

SelfCodeAlign: Self-Alignment for Code Generation

Instruction tuning is a supervised fine-tuning approach that significantly improves the ability of large language models (LLMs) to follow human instructions. We propose SelfCodeAlign, the first fully transparent and permissive pipeline for self-aligning code LLMs without extensive human annotations or distillation. SelfCodeAlign employs the same base model for inference throughout the data generation process. It first extracts diverse coding concepts from high-quality seed snippets to generate new tasks. It then samples multiple responses per task, pairs each with test cases, and validates them in a sandbox environment. Finally, passing examples are selected for instruction tuning. In our primary experiments, we use SelfCodeAlign with CodeQwen1.5-7B to generate a dataset of 74k instruction-response pairs. Finetuning on this dataset leads to a model that achieves a 67.1 pass@1 on HumanEval+, surpassing CodeLlama-70B-Instruct despite being ten times smaller. Across all benchmarks, this finetuned model consistently outperforms the original version trained with OctoPack, the previous state-of-the-art method for instruction tuning without human annotations or distillation. Additionally, we show that SelfCodeAlign is effective across LLMs of various sizes, from 3B to 33B, and that the base models can benefit more from alignment with their own data distribution. We further validate each component's effectiveness in our pipeline, showing that SelfCodeAlign outperforms both direct distillation from GPT-4o and leading GPT-3.5-based distillation methods, such as OSS-Instruct and Evol-Instruct. SelfCodeAlign has also led to the creation of StarCoder2-Instruct, the first fully transparent, permissively licensed, and self-aligned code LLM that achieves state-of-the-art coding performance.

cs.CL↗

arXiv2024

Arctic-SnowCoder: Demystifying High-Quality Data in Code Pretraining

Recent studies have been increasingly demonstrating that high-quality data is crucial for effective pretraining of language models. However, the precise definition of "high-quality" remains underexplored. Focusing on the code domain, we introduce Arctic-SnowCoder-1.3B, a data-efficient base code model pretrained on 555B tokens through three phases of progressively refined data: (1) general pretraining with 500B standard-quality code tokens, preprocessed through basic filtering, deduplication, and decontamination, (2) continued pretraining with 50B high-quality tokens, selected from phase one by a BERT-style quality annotator trained to distinguish good code from random data, using positive examples drawn from high-quality code files, along with instruction data from Magicoder and StarCoder2-Instruct, and (3) enhanced pretraining with 5B synthetic data created by Llama-3.1-70B using phase two data as seeds, adapting the Magicoder approach for pretraining. Despite being trained on a limited dataset, Arctic-SnowCoder achieves state-of-the-art performance on BigCodeBench, a coding benchmark focusing on practical and challenging programming tasks, compared to similarly sized models trained on no more than 1T tokens, outperforming Phi-1.5-1.3B by 36%. Across all evaluated benchmarks, Arctic-SnowCoder-1.3B beats StarCoderBase-3B pretrained on 1T tokens. Additionally, it matches the performance of leading small base code models trained on trillions of tokens. For example, Arctic-SnowCoder-1.3B surpasses StarCoder2-3B, pretrained on over 3.3T tokens, on HumanEval+, a benchmark that evaluates function-level code generation, and remains competitive on BigCodeBench. Our evaluation presents a comprehensive analysis justifying various design choices for Arctic-SnowCoder. Most importantly, we find that the key to high-quality data is its alignment with the distribution of downstream applications.

cs.CL↗

arXiv2024

Multi-Treatment Multi-Task Uplift Modeling for Enhancing User Growth

As a key component in boosting online user growth, uplift modeling aims to measure individual user responses (e.g., whether to play the game) to various treatments, such as gaming bonuses, thereby enhancing business outcomes. However, previous research typically considers a single-task, single-treatment setting, where only one treatment exists and the overall treatment effect is measured by a single type of user response. In this paper, we propose a Multi-Treatment Multi-Task (MTMT) uplift network to estimate treatment effects in a multi-task scenario. We identify the multi-treatment problem as a causal inference problem with a tiered response, comprising a base effect (from offering a treatment) and an incremental effect (from offering a specific type of treatment), where the base effect can be numerically much larger than the incremental effect. Specifically, MTMT separately encodes user features and treatments. The user feature encoder uses a multi-gate mixture of experts (MMOE) network to encode relevant user features, explicitly learning inter-task relations. The resultant embeddings are used to measure natural responses per task. Furthermore, we introduce a treatment-user feature interaction module to model correlations between each treatment and user feature. Consequently, we separately measure the base and incremental treatment effect for each task based on the produced treatment-aware representations. Experimental results based on an offline public dataset and an online proprietary dataset demonstrate the effectiveness of MTMT in single/multi-treatment and single/multi-task settings. Additionally, MTMT has been deployed in our gaming platform to improve user experience.

cs.LG↗