arXiv subjects

Haoran Wang

Publications and source records attributed to Haoran Wang.

At least 127 records · Page 7Linked to original sources

arXiv2025

SWE-Dev: Building Software Engineering Agents with Training and Inference Scaling

Large language models (LLMs) have advanced rapidly from conversational problem solving to addressing real-world tasks involving tool use, such as software engineering (SWE). Recent LLM-powered toolkits, such as OpenAI Codex and Cursor, have offered end-to-end automation of the software development process. However, building effective SWE agents remains challenging due to the lack of high-quality training data and effective test cases. To address this issue, we present SWE-Dev, an SWE agent built upon open-source LLMs. First, we develop a robust pipeline to synthesize test cases for patch evaluation. Second, we scale up agent trajectories to construct the training data for building SWE-Dev. Experiments on the SWE-bench-Verified benchmark show that the SWE-Dev models can achieve top performance among all open SWE agents. Specifically, the success rates of the SWE-Dev 7B and 32B parameter models reach 23.4% and 36.6%, respectively, outperforming state-of-the-art open-source models. All code, models, and datasets are publicly available at https://github.com/THUDM/SWE-Dev.

cs.AI↗

arXiv2025

RealDriveSim: A Realistic Multi-Modal Multi-Task Synthetic Dataset for Autonomous Driving

As perception models continue to develop, the need for large-scale datasets increases. However, data annotation remains far too expensive to effectively scale and meet the demand. Synthetic datasets provide a solution to boost model performance with substantially reduced costs. However, current synthetic datasets remain limited in their scope, realism, and are designed for specific tasks and applications. In this work, we present RealDriveSim, a realistic multi-modal synthetic dataset for autonomous driving that not only supports popular 2D computer vision applications but also their LiDAR counterparts, providing fine-grained annotations for up to 64 classes. We extensively evaluate our dataset for a wide range of applications and domains, demonstrating state-of-the-art results compared to existing synthetic benchmarks. The dataset is publicly available at https://realdrivesim.github.io/.

cs.CV↗

arXiv2025

E2MPL:An Enduring and Efficient Meta Prompt Learning Framework for Few-shot Unsupervised Domain Adaptation

Few-shot unsupervised domain adaptation (FS-UDA) leverages a limited amount of labeled data from a source domain to enable accurate classification in an unlabeled target domain. Despite recent advancements, current approaches of FS-UDA continue to confront a major challenge: models often demonstrate instability when adapted to new FS-UDA tasks and necessitate considerable time investment. To address these challenges, we put forward a novel framework called Enduring and Efficient Meta-Prompt Learning (E2MPL) for FS-UDA. Within this framework, we utilize the pre-trained CLIP model as the backbone of feature learning. Firstly, we design domain-shared prompts, consisting of virtual tokens, which primarily capture meta-knowledge from a wide range of meta-tasks to mitigate the domain gaps. Secondly, we develop a task prompt learning network that adaptively learns task-specific specific prompts with the goal of achieving fast and stable task generalization. Thirdly, we formulate the meta-prompt learning process as a bilevel optimization problem, consisting of (outer) meta-prompt learner and (inner) task-specific classifier and domain adapter. Also, the inner objective of each meta-task has the closed-form solution, which enables efficient prompt learning and adaptation to new tasks in a single step. Extensive experimental studies demonstrate the promising performance of our framework in a domain adaptation benchmark dataset DomainNet. Compared with state-of-the-art methods, our method has improved accuracy by at least 15.4% and reduced the time by 68.5% on average in 5-way 1-shot tasks, and improved accuracy by 8.7% and reduced the time by 74.1% on average in 5-way 5-shot tasks. Moreover, our approach exhibits more enduring performance than the other methods, i.e., being more stable across 3600 test tasks.

cs.CV↗

arXiv2025

Balancing Efficiency and Equity in Classroom Assignment under Endogenous Peer Effects

This paper presents a three-step empirical framework for optimizing classroom assignments under endogenous peer effects, using data from the China Education Panel Survey (CEPS). We design \textit{PeerNN}, a neural network that mimics endogenous network formation as a discrete choice model, generating a friendship-intensity matrix ($Ω$) that captures student popularity. \textbf{Step 2: Estimating Peer Effects.} We measure the peer effect friends' average 6th-grade class rank weighted by $Ω$ on 8th-grade cognitive test score. Incorporating $Ω$ into the linear-in-means model induces endogeneity. Using quasi-random classroom assignments, we instrument friends' average 6th-grade class rank with the average classmates' 6th-grade class rank (unweighted by $Ω$). Our main regression result shows that a 10\% improvement in friends' 6th-grade class rank raises 8th-grade cognitive test scores by 0.13 SD. Positive $β$ implies maximizing (minimizing) the popularity of high (low) achievers optimizes outcomes. \textbf{Step 3: Simulating Policy Trade-offs.} We use estimates from Step 1 and Step 2 to simulate optimal classroom assignments. We first implement a genetic algorithm (GA) to maximize average peer effect and observe a 1.9\% improvement. However, serious inequity issues arise: low-achieving students are hurt the most in the pursuit of the higher average peer effect. We propose an \textit{Algorithmically Fair GA} (AFGA), achieving a 1.2\% gain while ensuring more equitable educational outcomes. These results underscore that efficiency-focused classroom assignment policies can exacerbate inequality. We recommend incorporating fairness considerations when designing classroom assignment policies that account for endogenous spillovers.

econ.GN↗

arXiv2025

RenderBender: A Survey on Adversarial Attacks Using Differentiable Rendering

Differentiable rendering techniques like Gaussian Splatting and Neural Radiance Fields have become powerful tools for generating high-fidelity models of 3D objects and scenes. Their ability to produce both physically plausible and differentiable models of scenes are key ingredient needed to produce physically plausible adversarial attacks on DNNs. However, the adversarial machine learning community has yet to fully explore these capabilities, partly due to differing attack goals (e.g., misclassification, misdetection) and a wide range of possible scene manipulations used to achieve them (e.g., alter texture, mesh). This survey contributes the first framework that unifies diverse goals and tasks, facilitating easy comparison of existing work, identifying research gaps, and highlighting future directions - ranging from expanding attack goals and tasks to account for new modalities, state-of-the-art models, tools, and pipelines, to underscoring the importance of studying real-world threats in complex scenes.

cs.LG↗

arXiv2025

3D Gaussian Splat Vulnerabilities

With 3D Gaussian Splatting (3DGS) being increasingly used in safety-critical applications, how can an adversary manipulate the scene to cause harm? We introduce CLOAK, the first attack that leverages view-dependent Gaussian appearances - colors and textures that change with viewing angle - to embed adversarial content visible only from specific viewpoints. We further demonstrate DAGGER, a targeted adversarial attack directly perturbing 3D Gaussians without access to underlying training data, deceiving multi-stage object detectors e.g., Faster R-CNN, through established methods such as projected gradient descent. These attacks highlight underexplored vulnerabilities in 3DGS, introducing a new potential threat to robotic learning for autonomous navigation and other safety-critical 3DGS applications.

cs.CR↗

arXiv2025

Accelerating the Evolution of Personalized Automated Lane Change through Lesson Learning

Personalization is crucial for the widespread adoption of advanced driver assistance system. To match up with each user's preference, the online evolution capability is a must. However, conventional evolution methods learn from naturalistic driving data, which requires a lot computing power and cannot be applied online. To address this challenge, this paper proposes a lesson learning approach: learning from driver's takeover interventions. By leveraging online takeover data, the driving zone is generated to ensure perceived safety using Gaussian discriminant analysis. Real-time corrections to trajectory planning rewards are enacted through apprenticeship learning. Guided by the objective of optimizing rewards within the constraints of the driving zone, this approach employs model predictive control for trajectory planning. This lesson learning framework is highlighted for its faster evolution capability, adeptness at experience accumulating, assurance of perceived safety, and computational efficiency. Simulation results demonstrate that the proposed system consistently achieves a successful customization without further takeover interventions. Accumulated experience yields a 24% enhancement in evolution efficiency. The average number of learning iterations is only 13.8. The average computation time is 0.08 seconds.

cs.LG↗

arXiv2025

Towards General Discrete Speech Codec for Complex Acoustic Environments: A Study of Reconstruction and Downstream Task Consistency

Neural speech codecs excel in reconstructing clean speech signals; however, their efficacy in complex acoustic environments and downstream signal processing tasks remains underexplored. In this study, we introduce a novel benchmark named Environment-Resilient Speech Codec Benchmark (ERSB) to systematically evaluate whether neural speech codecs are environment-resilient. Specifically, we assess two key capabilities: (1) robust reconstruction, which measures the preservation of both speech and non-speech acoustic details, and (2) downstream task consistency, which ensures minimal deviation in downstream signal processing tasks when using reconstructed speech instead of the original. Our comprehensive experiments reveal that complex acoustic environments significantly degrade signal reconstruction and downstream task consistency. This work highlights the limitations of current speech codecs and raises a future direction that improves them for greater environmental resilience.

cs.SD↗

arXiv2025

A Unified Gradient-based Framework for Task-agnostic Continual Learning-Unlearning

Recent advancements in deep models have highlighted the need for intelligent systems that combine continual learning (CL) for knowledge acquisition with machine unlearning (MU) for data removal, forming the Continual Learning-Unlearning (CLU) paradigm. While existing work treats CL and MU as separate processes, we reveal their intrinsic connection through a unified optimization framework based on Kullback-Leibler divergence minimization. This framework decomposes gradient updates for approximate CLU into four components: learning new knowledge, unlearning targeted data, preserving existing knowledge, and modulation via weight saliency. A critical challenge lies in balancing knowledge update and retention during sequential learning-unlearning cycles. To resolve this stability-plasticity dilemma, we introduce a remain-preserved manifold constraint to induce a remaining Hessian compensation for CLU iterations. A fast-slow weight adaptation mechanism is designed to efficiently approximate the second-order optimization direction, combined with adaptive weighting coefficients and a balanced weight saliency mask, proposing a unified implementation framework for gradient-based CLU. Furthermore, we pioneer task-agnostic CLU scenarios that support fine-grained unlearning at the cross-task category and random sample levels beyond the traditional task-aware setups. Experiments demonstrate that the proposed UG-CLU framework effectively coordinates incremental learning, precise unlearning, and knowledge stability across multiple datasets and model architectures, providing a theoretical foundation and methodological support for dynamic, compliant intelligent systems.

cs.LG↗

arXiv2025

HG2P: Hippocampus-inspired High-reward Graph and Model-Free Q-Gradient Penalty for Path Planning and Motion Control

Goal-conditioned hierarchical reinforcement learning (HRL) decomposes complex reaching tasks into a sequence of simple subgoal-conditioned tasks, showing significant promise for addressing long-horizon planning in large-scale environments. This paper bridges the goal-conditioned HRL based on graph-based planning to brain mechanisms, proposing a hippocampus-striatum-like dual-controller hypothesis. Inspired by the brain mechanisms of organisms (i.e., the high-reward preferences observed in hippocampal replay) and instance-based theory, we propose a high-return sampling strategy for constructing memory graphs, improving sample efficiency. Additionally, we derive a model-free lower-level Q-function gradient penalty to resolve the model dependency issues present in prior work, improving the generalization of Lipschitz constraints in applications. Finally, we integrate these two extensions, High-reward Graph and model-free Gradient Penalty (HG2P), into the state-of-the-art framework ACLG, proposing a novel goal-conditioned HRL framework, HG2P+ACLG. Experimentally, the results demonstrate that our method outperforms state-of-the-art goal-conditioned HRL algorithms on a variety of long-horizon navigation tasks and robotic manipulation tasks.

cs.LG↗

arXiv2025

Safety-Aware Human-Lead Vehicle Platooning by Proactively Reacting to Uncertain Human Behaving

Human-Lead Cooperative Adaptive Cruise Control (HL-CACC) is regarded as a promising vehicle platooning technology in real-world implementation. By utilizing a Human-driven Vehicle (HV) as the platoon leader, HL-CACC reduces the cost and enhances the reliability of perception and decision-making. However, state-of-the-art HL-CACC technology still has a great limitation on driving safety due to the lack of considering the leading human driver's uncertain behavior. In this study, a HL-CACC controller is designed based on Stochastic Model Predictive Control (SMPC). It is enabled to predict the driving intention of the leading Connected Human-Driven Vehicle (CHV). The proposed controller has the following features: i) enhanced perceived safety in oscillating traffic; ii) guaranteed safety against hard brakes; iii) computational efficiency for real-time implementation. The proposed controller is evaluated on a PreScan&Simulink simulation platform. Real vehicle trajectory data is collected for the calibration of the simulation. Results reveal that the proposed controller: i) improves perceived safety by 19.17% in oscillating traffic; ii) enhances actual safety by 7.76% against hard brakes; iii) is confirmed with string stability. The computation time is approximately 3.2 milliseconds when running on a laptop equipped with an Intel i5-13500H CPU. This indicates the proposed controller is ready for real-time implementation.

cs.RO↗

arXiv2025

Mirroring the Parking Target: An Optimal-Control-Based Parking Motion Planner with Strengthened Parking Reliability and Faster Parking Completion

Automated Parking Assist (APA) systems are now facing great challenges of low adoption in applications, due to users' concerns about parking capability, reliability, and completion efficiency. To upgrade the conventional APA planners and enhance user's acceptance, this research proposes an optimal-control-based parking motion planner. Its highlight lies in its control logic: planning trajectories by mirroring the parking target. This method enables: i) parking capability in narrow spaces; ii) better parking reliability by expanding Operation Design Domain (ODD); iii) faster completion of parking process; iv) enhanced computational efficiency; v) universal to all types of parking. A comprehensive evaluation is conducted. Results demonstrate the proposed planner does enhance parking success rate by 40.6%, improve parking completion efficiency by 18.0%, and expand ODD by 86.1%. It shows its superiority in difficult parking cases, such as the parallel parking scenario and narrow spaces. Moreover, the average computation time of the proposed planner is 74 milliseconds. Results indicate that the proposed planner is ready for real-time commercial applications.

cs.RO↗

arXiv2025

LLGS: Unsupervised Gaussian Splatting for Image Enhancement and Reconstruction in Pure Dark Environment

3D Gaussian Splatting has shown remarkable capabilities in novel view rendering tasks and exhibits significant potential for multi-view optimization.However, the original 3D Gaussian Splatting lacks color representation for inputs in low-light environments. Simply using enhanced images as inputs would lead to issues with multi-view consistency, and current single-view enhancement systems rely on pre-trained data, lacking scene generalization. These problems limit the application of 3D Gaussian Splatting in low-light conditions in the field of robotics, including high-fidelity modeling and feature matching. To address these challenges, we propose an unsupervised multi-view stereoscopic system based on Gaussian Splatting, called Low-Light Gaussian Splatting (LLGS). This system aims to enhance images in low-light environments while reconstructing the scene. Our method introduces a decomposable Gaussian representation called M-Color, which separately characterizes color information for targeted enhancement. Furthermore, we propose an unsupervised optimization method with zero-knowledge priors, using direction-based enhancement to ensure multi-view consistency. Experiments conducted on real-world datasets demonstrate that our system outperforms state-of-the-art methods in both low-light enhancement and 3D Gaussian Splatting.

cs.CV↗

arXiv2025

Understanding the Generalization of In-Context Learning in Transformers: An Empirical Study

Large language models (LLMs) like GPT-4 and LLaMA-3 utilize the powerful in-context learning (ICL) capability of Transformer architecture to learn on the fly from limited examples. While ICL underpins many LLM applications, its full potential remains hindered by a limited understanding of its generalization boundaries and vulnerabilities. We present a systematic investigation of transformers' generalization capability with ICL relative to training data coverage by defining a task-centric framework along three dimensions: inter-problem, intra-problem, and intra-task generalization. Through extensive simulation and real-world experiments, encompassing tasks such as function fitting, API calling, and translation, we find that transformers lack inter-problem generalization with ICL, but excel in intra-task and intra-problem generalization. When the training data includes a greater variety of mixed tasks, it significantly enhances the generalization ability of ICL on unseen tasks and even on known simple tasks. This guides us in designing training data to maximize the diversity of tasks covered and to combine different tasks whenever possible, rather than solely focusing on the target task for testing.

cs.LG↗

arXiv2025

Structural and Statistical Texture Knowledge Distillation for Semantic Segmentation

Existing knowledge distillation works for semantic segmentation mainly focus on transferring high-level contextual knowledge from teacher to student. However, low-level texture knowledge is also of vital importance for characterizing the local structural pattern and global statistical property, such as boundary, smoothness, regularity and color contrast, which may not be well addressed by high-level deep features. In this paper, we are intended to take full advantage of both structural and statistical texture knowledge and propose a novel Structural and Statistical Texture Knowledge Distillation (SSTKD) framework for semantic segmentation. Specifically, for structural texture knowledge, we introduce a Contourlet Decomposition Module (CDM) that decomposes low-level features with iterative Laplacian pyramid and directional filter bank to mine the structural texture knowledge. For statistical knowledge, we propose a Denoised Texture Intensity Equalization Module (DTIEM) to adaptively extract and enhance statistical texture knowledge through heuristics iterative quantization and denoised operation. Finally, each knowledge learning is supervised by an individual loss function, forcing the student network to mimic the teacher better from a broader perspective. Experiments show that the proposed method achieves state-of-the-art performance on Cityscapes, Pascal VOC 2012 and ADE20K datasets.

cs.CV↗

arXiv2025

Understanding Driver Cognition and Decision-Making Behaviors in High-Risk Scenarios: A Drift Diffusion Perspective

Ensuring safe interactions between autonomous vehicles (AVs) and human drivers in mixed traffic systems remains a major challenge, particularly in complex, high-risk scenarios. This paper presents a cognition-decision framework that integrates individual variability and commonalities in driver behavior to quantify risk cognition and model dynamic decision-making. First, a risk sensitivity model based on a multivariate Gaussian distribution is developed to characterize individual differences in risk cognition. Then, a cognitive decision-making model based on the drift diffusion model (DDM) is introduced to capture common decision-making mechanisms in high-risk environments. The DDM dynamically adjusts decision thresholds by integrating initial bias, drift rate, and boundary parameters, adapting to variations in speed, relative distance, and risk sensitivity to reflect diverse driving styles and risk preferences. By simulating high-risk scenarios with lateral, longitudinal, and multidimensional risk sources in a driving simulator, the proposed model accurately predicts cognitive responses and decision behaviors during emergency maneuvers. Specifically, by incorporating driver-specific risk sensitivity, the model enables dynamic adjustments of key DDM parameters, allowing for personalized decision-making representations in diverse scenarios. Comparative analysis with IDM, Gipps, and MOBIL demonstrates that DDM more precisely captures human cognitive processes and adaptive decision-making in high-risk scenarios. These findings provide a theoretical basis for modeling human driving behavior and offer critical insights for enhancing AV-human interaction in real-world traffic environments.

cs.AI↗

arXiv2025

WeakMedSAM: Weakly-Supervised Medical Image Segmentation via SAM with Sub-Class Exploration and Prompt Affinity Mining

We have witnessed remarkable progress in foundation models in vision tasks. Currently, several recent works have utilized the segmenting anything model (SAM) to boost the segmentation performance in medical images, where most of them focus on training an adaptor for fine-tuning a large amount of pixel-wise annotated medical images following a fully supervised manner. In this paper, to reduce the labeling cost, we investigate a novel weakly-supervised SAM-based segmentation model, namely WeakMedSAM. Specifically, our proposed WeakMedSAM contains two modules: 1) to mitigate severe co-occurrence in medical images, a sub-class exploration module is introduced to learn accurate feature representations. 2) to improve the quality of the class activation maps, our prompt affinity mining module utilizes the prompt capability of SAM to obtain an affinity map for random-walk refinement. Our method can be applied to any SAM-like backbone, and we conduct experiments with SAMUS and EfficientSAM. The experimental results on three popularly-used benchmark datasets, i.e., BraTS 2019, AbdomenCT-1K, and MSD Cardiac dataset, show the promising results of our proposed WeakMedSAM. Our code is available at https://github.com/wanghr64/WeakMedSAM.

cs.CV↗

arXiv2025

Improved high-index saddle dynamics for finding saddle points and solution landscape

We present an improved high-index saddle dynamics (iHiSD) for finding saddle points and constructing solution landscapes, which is a crossover dynamics from gradient flow to traditional HiSD such that the Morse theory for gradient flow could be involved. We propose analysis for the reflection manifold in iHiSD, and then prove its stable and nonlocal convergence from outside of the region of attraction to the saddle point, which resolves the dependence of the convergence of HiSD on the initial value. We then present and analyze a discretized iHiSD that inherits these convergence properties. Furthermore, based on the Morse theory, we prove that any two saddle points could be connected by a sequence of trajectories of iHiSD. Theoretically, this implies that a solution landscape with a finite number of stationary points could be completely constructed by means of iHiSD, which partly answers the completeness issue of the solution landscape for the first time and indicates the necessity of integrating the gradient flow in HiSD. Different methods are compared by numerical experiments to substantiate the effectiveness of the iHiSD method.

math.NA↗