Zixuan Wu

Sean Ye, Manisha Natarajan, Zixuan Wu et al.

The need for opponent modeling and tracking arises in several real-world scenarios, such as professional sports, video game design, and drug-trafficking interdiction. In this work, we present Graph based Adversarial Modeling with Mutal Information (GrAMMI) for modeling the behavior of an adversarial opponent agent. GrAMMI is a novel graph neural network (GNN) based approach that uses mutual information maximization as an auxiliary objective to predict the current and future states of an adversarial opponent with partial observability. To evaluate GrAMMI, we design two large-scale, pursuit-evasion domains inspired by real-world scenarios, where a team of heterogeneous agents is tasked with tracking and interdicting a single adversarial agent, and the adversarial agent must evade detection while achieving its own objectives. With the mutual information formulation, GrAMMI outperforms all baselines in both domains and achieves 31.68% higher log-likelihood on average for future adversarial state predictions across both domains.

Sean Ye, Manisha Natarajan, Zixuan Wu et al.

Target tracking plays a crucial role in real-world scenarios, particularly in drug-trafficking interdiction, where the knowledge of an adversarial target's location is often limited. Improving autonomous tracking systems will enable unmanned aerial, surface, and underwater vehicles to better assist in interdicting smugglers that use manned surface, semi-submersible, and aerial vessels. As unmanned drones proliferate, accurate autonomous target estimation is even more crucial for security and safety. This paper presents Constrained Agent-based Diffusion for Enhanced Multi-Agent Tracking (CADENCE), an approach aimed at generating comprehensive predictions of adversary locations by leveraging past sparse state information. To assess the effectiveness of this approach, we evaluate predictions on single-target and multi-target pursuit environments, employing Monte-Carlo sampling of the diffusion model to estimate the probability associated with each generated trajectory. We propose a novel cross-attention based diffusion model that utilizes constraint-based sampling to generate multimodal track hypotheses. Our single-target model surpasses the performance of all baseline methods on Average Displacement Error (ADE) for predictions across all time horizons.

Zixuan Wu, Sean Ye, Manisha Natarajan et al.

We study a search and tracking (S&T) problem where a team of dynamic search agents must collaborate to track an adversarial, evasive agent. The heterogeneous search team may only have access to a limited number of past adversary trajectories within a large search space. This problem is challenging for both model-based searching and reinforcement learning (RL) methods since the adversary exhibits reactionary and deceptive evasive behaviors in a large space leading to sparse detections for the search agents. To address this challenge, we propose a novel Multi-Agent RL (MARL) framework that leverages the estimated adversary location from our learnable filtering model. We show that our MARL architecture can outperform all baselines and achieves a 46% increase in detection rate.

Kin Man Lee, Sean Ye, Qingyu Xiao et al.

Advances in robot learning have enabled robots to generate skills for a variety of tasks. Yet, robot learning is typically sample inefficient, struggles to learn from data sources exhibiting varied behaviors, and does not naturally incorporate constraints. These properties are critical for fast, agile tasks such as playing table tennis. Modern techniques for learning from demonstration improve sample efficiency and scale to diverse data, but are rarely evaluated on agile tasks. In the case of reinforcement learning, achieving good performance requires training on high-fidelity simulators. To overcome these limitations, we develop a novel diffusion modeling approach that is offline, constraint-guided, and expressive of diverse agile behaviors. The key to our approach is a kinematic constraint gradient guidance (KCGG) technique that computes gradients through both the forward kinematics of the robot arm and the diffusion model to direct the sampling process. KCGG minimizes the cost of violating constraints while simultaneously keeping the sampled trajectory in-distribution of the training data. We demonstrate the effectiveness of our approach for time-critical robotic tasks by evaluating KCGG in two challenging domains: simulated air hockey and real table tennis. In simulated air hockey, we achieved a 25.4% increase in block rate, while in table tennis, we saw a 17.3% increase in success rate compared to imitation learning baselines.

Ye Wang, Zixuan Wu, Lifeng Shen et al.

Imbalanced data distribution remains a critical challenge in sequential learning, leading models to easily recognize frequent categories while failing to detect minority classes adequately. The Mixture-of-Experts model offers a scalable solution, yet its application is often hindered by parameter inefficiency, poor expert specialization, and difficulty in resolving prediction conflicts. To Master the Minority classes effectively, we propose the Uncertainty-based Multi-Expert fusion network (UME) framework. UME is designed with three core innovations: First, we employ Ensemble LoRA for parameter-efficient modeling, significantly reducing the trainable parameter count. Second, we introduce Sequential Specialization guided by Dempster-Shafer Theory (DST), which ensures effective specialization on the challenging-tailed classes. Finally, an Uncertainty-Guided Fusion mechanism uses DST's certainty measures to dynamically weigh expert opinions, resolving conflicts by prioritizing the most confident expert for reliable final predictions. Extensive experiments across four public hierarchical text classification datasets demonstrate that UME achieves state-of-the-art performance. We achieve a performance gain of up to 17.97\% over the best baseline on individual categories, while reducing trainable parameters by up to 10.32\%. The findings highlight that uncertainty-guided expert coordination is a principled strategy for addressing challenging-tailed sequence learning. Our code is available at https://github.com/CQUPTWZX/Multi-experts.

Wei Dai, Rui Liu, Zixuan Wu et al.

Early detection and accurate diagnosis can predict the risk of malignant disease transformation, thereby increasing the probability of effective treatment. Identifying mild syndrome with small pathological regions serves as an ominous warning and is fundamental in the early diagnosis of diseases. While deep learning algorithms, particularly convolutional neural networks (CNNs), have shown promise in segmenting medical objects, analyzing small areas in medical images remains challenging. This difficulty arises due to information losses and compression defects from convolution and pooling operations in CNNs, which become more pronounced as the network deepens, especially for small medical objects. To address these challenges, we propose a novel scale-variant attention-based network (SvANet) for accurately segmenting small-scale objects in medical images. The SvANet consists of scale-variant attention, cross-scale guidance, Monte Carlo attention, and vision transformer, which incorporates cross-scale features and alleviates compression artifacts for enhancing the discrimination of small medical objects. Quantitative experimental results demonstrate the superior performance of SvANet, achieving 96.12%, 96.11%, 89.79%, 84.15%, 80.25%, 73.05%, and 72.58% in mean Dice coefficient for segmenting kidney tumors, skin lesions, hepatic tumors, polyps, surgical excision cells, retinal vasculatures, and sperms, which occupy less than 1% of the image areas in KiTS23, ISIC 2018, ATLAS, PolypGen, TissueNet, FIVES, and SpermHealth datasets, respectively.

Zixuan Wu, Yoolim Kim, Carolyn Jane Anderson

Vision-Language Models (VLMs) building upon the foundation of powerful large language models have made rapid progress in reasoning across visual and textual data. While VLMs perform well on vision tasks that they are trained on, our results highlight key challenges in abstract pattern recognition. We present GlyphPattern, a 954 item dataset that pairs 318 human-written descriptions of visual patterns from 40 writing systems with three visual presentation styles. GlyphPattern evaluates abstract pattern recognition in VLMs, requiring models to understand and judge natural language descriptions of visual patterns. GlyphPattern patterns are drawn from a large-scale cognitive science investigation of human writing systems; as a result, they are rich in spatial reference and compositionality. Our experiments show that GlyphPattern is challenging for state-of-the-art VLMs (GPT-4o achieves only 55% accuracy), with marginal gains from few-shot prompting. Our detailed error analysis reveals challenges at multiple levels, including visual processing, natural language understanding, and pattern generalization.

Zixuan Wu, Sean Ye, Manisha Natarajan et al.

Reinforcement Learning (RL)-based motion planning has recently shown the potential to outperform traditional approaches from autonomous navigation to robot manipulation. In this work, we focus on a motion planning task for an evasive target in a partially observable multi-agent adversarial pursuit-evasion game (PEG). Pursuit-evasion problems are relevant to various applications, such as search and rescue operations and surveillance robots, where robots must effectively plan their actions to gather intelligence or accomplish mission tasks while avoiding detection or capture. We propose a hierarchical architecture that integrates a high-level diffusion model to plan global paths responsive to environment data, while a low-level RL policy reasons about evasive versus global path-following behavior. The benchmark results across different domains and different observability show that our approach outperforms baselines by 77.18% and 47.38% on detection and goal reaching rate, which leads to 51.4% increasing of the performance score on average. Additionally, our method improves interpretability, flexibility and efficiency of the learned policy.

Zixuan Wu, Francesca Lucchetti, Aleksander Boruch-Gruszecki et al.

Existing benchmarks for frontier models often test specialized, "PhD-level" knowledge that is difficult for non-experts to grasp. In contrast, we present a benchmark with 594 problems based on the NPR Sunday Puzzle Challenge that requires only general knowledge. Our benchmark is challenging for both humans and models; however correct solutions are easy to verify, and models' mistakes are easy to spot. As LLMs are more widely deployed in society, we believe it is useful to develop benchmarks for frontier models that humans can understand without the need for deep domain expertise. Our work reveals capability gaps that are not evident in existing benchmarks: OpenAI o1 significantly outperforms other reasoning models on our benchmark, despite being on par with other models when tested on benchmarks that test specialized knowledge. Furthermore, our analysis of reasoning outputs uncovers new kinds of failures. DeepSeek R1, for instance, often concedes with "I give up" before providing an answer that it knows is wrong. R1 can also be remarkably "uncertain" in its output and in rare cases, it does not "finish thinking," which suggests the need for techniques to "wrap up" before the context window limit is reached. We also quantify the effectiveness of reasoning longer to identify the point beyond which more reasoning is unlikely to improve accuracy on our benchmark.

Aleksander Boruch-Gruszecki, Yangtian Zi, Zixuan Wu et al.

Large language models (LLMs) already excel at writing code in high-resource languages such as Python and JavaScript, yet stumble on low-resource languages that remain essential to science and engineering. Besides the obvious shortage of pre-training data, post-training itself is a bottleneck: every new language seems to require new datasets, test harnesses, and reinforcement-learning (RL) infrastructure. We introduce Agnostics, a language-agnostic post-training pipeline that eliminates this per-language engineering. The key idea is to judge code solely by its externally observable behavior, so a single verifier can test solutions written in any language. Concretely, we (i) use an LLM to rewrite existing unit-test datasets into an I/O format, (ii) supply a short configuration that tells the verifier how to compile and run a target language, and (iii) apply reinforcement learning with verifiable rewards (RLVR) in a robust code execution environment. Applied to five low-resource languages--Lua, Julia, R, OCaml, and Fortran--Agnostics (1) improves Qwen-3 4B to performance that rivals other 16B-70B open-weight models; (2) scales cleanly to larger and diverse model families (Qwen-3 8B, DeepSeek Coder 6.7B Instruct, Phi 4 Mini); and (3) for ${\le} 16$B parameter models, sets new state-of-the-art pass@1 results on MultiPL-E and a new multi-language version LiveCodeBench that we introduce. We will release the language-agnostic training datasets (Ag-MBPP-X, Ag-Codeforces-X, Ag-LiveCodeBench-X), training code, and ready-to-use configurations, making RL post-training in any programming language as simple as editing a short YAML file.

Francesca Lucchetti, Zixuan Wu, Arjun Guha et al.

Although LLMs are increasing the productivity of professional programmers, existing work shows that beginners struggle to prompt LLMs to solve text-to-code tasks. Why is this the case? This paper explores two competing hypotheses about the cause of student-LLM miscommunication: (1) students simply lack the technical vocabulary needed to write good prompts, and (2) students do not understand the extent of information that LLMs need to solve code generation tasks. We study (1) with a causal intervention experiment on technical vocabulary and (2) by analyzing graphs that abstract how students edit prompts and the different failures that they encounter. We find that substance beats style: a poor grasp of technical vocabulary is merely correlated with prompt failure; that the information content of prompts predicts success; that students get stuck making trivial edits; and more. Our findings have implications for the use of LLMs in programming education, and for efforts to make computing more accessible with LLMs.

Yeo Jin Jung, Yating Liu, Zixuan Wu et al.

Conformal prediction provides distribution-free prediction sets with finite-sample conditional guarantees. We build upon the RKHS-based framework of Gibbs et al. (2023), which leverages families of covariate shifts to provide approximate conditional conformal prediction intervals, an approach with strong theoretical promise, but with prohibitive computational cost. To bridge this gap, we develop a stable and efficient algorithm that computes the full solution path of the regularized RKHS conformal optimization problem, at essentially the same cost as a single kernel quantile fit. Our path-tracing framework simultaneously tunes hyperparameters, providing smoothness control and data-adaptive calibration. To extend the method to high-dimensional settings, we further integrate our approach with low-rank latent embeddings that capture conditional validity in a data-driven latent space. Empirically, our method provides reliable conditional coverage across a variety of modern black-box predictors, improving the interval length of Gibbs et al. (2023) by 30%, while achieving a 40-fold speedup.

Zixuan Wu, Hengyuan Zhang, Ting-Hsuan Chen et al.

Parking is a critical pillar of driving safety. While recent end-to-end (E2E) approaches have achieved promising in-domain results, robustness under domain shifts (e.g., weather and lighting changes) remains a key challenge. Rather than relying on additional data, in this paper, we propose Dino-Diffusion Parking (DDP), a domain-agnostic autonomous parking pipeline that integrates visual foundation models with diffusion-based planning to enable generalized perception and robust motion planning under distribution shifts. We train our pipeline in CARLA at regular setting and transfer it to more adversarial settings in a zero-shot fashion. Our model consistently achieves a parking success rate above 90% across all tested out-of-distribution (OOD) scenarios, with ablation studies confirming that both the network architecture and algorithmic design significantly enhance cross-domain performance over existing baselines. Furthermore, testing in a 3D Gaussian splatting (3DGS) environment reconstructed from a real-world parking lot demonstrates promising sim-to-real transfer.

Yangtian Zi, Zixuan Wu, Aleksander Boruch-Gruszecki et al.

Fine-tuning large language models for code editing has typically relied on mining commits and pull requests. The working hypothesis has been that commit messages describe human intent in natural language, and patches to code describe the changes that implement that intent. However, much of the previously collected data is noisy: commit messages are terse, human-written commits commingle several unrelated edits, and many commits come from simple, rule-based bots. The recent adoption of software engineering agents changes this landscape. Code changes co-authored by humans and agents tend to be more narrowly scoped and focused on clearer goals. Their commit messages, generated by LLMs, articulate intent and rationale in much greater detail. Moreover, when these changes land in public repositories, they are implicitly filtered by humans: maintainers discard low-quality commits to their projects. We present AgentPack, a corpus of 1.3M code edits co-authored by Claude Code, OpenAI Codex, and Cursor Agent across public GitHub projects up to mid-August 2025. We describe the identification and curation pipeline, quantify adoption trends of these agents, and analyze the structural properties of the edits. Finally, we show that models fine-tuned on AgentPack can outperform models trained on prior human-only commit corpora, highlighting the potential of using public data from software engineering agents to train future code-editing models.

Zixuan Wu, So Won Jeong, Yating Liu et al.

With promising empirical performance across a wide range of applications, synthetic data augmentation appears a viable solution to data scarcity and the demands of increasingly data-intensive models. Its effectiveness lies in expanding the training set in a way that reduces estimator variance while introducing only minimal bias. Controlling this bias is therefore critical: effective data augmentation should generate diverse samples from the same underlying distribution as the training set, with minimal shifts. In this paper, we propose conformal data augmentation, a principled data filtering framework that leverages the power of conformal prediction to produce diverse synthetic data while filtering out poor-quality generations with provable risk control. Our method is simple to implement, requires no access to internal model logits, nor large-scale model retraining. We demonstrate the effectiveness of our approach across multiple tasks, including topic prediction, sentiment analysis, image classification, and fraud detection, showing consistent performance improvements of up to 40% in F1 score over unaugmented baselines, and 4% over other filtered augmentation baselines.

Shiyu Feng, Zixuan Wu, Yipu Zhao et al.

This paper describes a stereo image-based visual servoing system for trajectory tracking by a non-holonomic robot without externally derived pose information nor a known visual map of the environment. It is called trajectory servoing. The critical component is a feature-based, indirect Simultaneous Localization And Mapping (SLAM) method to provide a pool of available features with estimated depth, so that they may be propagated forward in time to generate image feature trajectories for visual servoing. Short and long distance experiments show the benefits of trajectory servoing for navigating unknown areas without absolute positioning. Empirically, trajectory servoing has better trajectory tracking performance than pose-based feedback when both rely on the same underlying SLAM system.