arXiv subjects

Ziyan Wang

Publications and source records attributed to Ziyan Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

VAM: Verbalized Action Masking for Controllable Exploration in RL Post-Training -- A Chess Case Study

Exploration remains a key bottleneck for reinforcement learning (RL) post-training of large language models (LLMs), where sparse feedback and large action spaces can lead to premature collapse into repetitive behaviors. We propose Verbalized Action Masking (VAM), which verbalizes an action mask in the prompt and enforces that the model outputs an action from the masked set. Building on this interface, we introduce iterative action-space pruning: if the target action is not sampled, we remove valid sampled actions from the mask and resample under the reduced candidate set, repeating until the target is sampled or a fixed budget is exhausted. We study VAM in chess and evaluate it under two training regimes: an engine-play regime that generates states via play against an engine opponent and a fixed-dataset regime that trains from a fixed dataset of positions with verifier scores. Across held-out chess puzzles and full-game play measured by average centipawn loss (ACPL), VAM improves learning efficiency and final performance over strong baselines, highlighting verbalized masking as a practical mechanism for controllable exploration in LLM RL post-training.

cs.LG↗

arXiv2026

Graph Domain Adaptation via Homophily-Agnostic Reconstructing Structure

Graph Domain Adaptation (GDA) transfers knowledge from labeled source graphs to unlabeled target graphs, addressing the challenge of label scarcity. However, existing GDA methods typically assume that both source and target graphs exhibit homophily, leading existing methods to perform poorly when heterophily is present. Furthermore, the lack of labels in the target graph makes it impossible to assess its homophily level beforehand. To address this challenge, we propose a novel homophily-agnostic approach that effectively transfers knowledge between graphs with varying degrees of homophily. Specifically, we adopt a divide-and-conquer strategy that first separately reconstructs highly homophilic and heterophilic variants of both the source and target graphs, and then performs knowledge alignment separately between corresponding graph variants. Extensive experiments conducted on five benchmark datasets demonstrate the superior performance of our approach, particularly highlighting its substantial advantages on heterophilic graphs.

cs.SI↗

arXiv2025

DGH: Dynamic Gaussian Hair

The creation of photorealistic dynamic hair remains a major challenge in digital human modeling because of the complex motions, occlusions, and light scattering. Existing methods often resort to static capture and physics-based models that do not scale as they require manual parameter fine-tuning to handle the diversity of hairstyles and motions, and heavy computation to obtain high-quality appearance. In this paper, we present Dynamic Gaussian Hair (DGH), a novel framework that efficiently learns hair dynamics and appearance. We propose: (1) a coarse-to-fine model that learns temporally coherent hair motion dynamics across diverse hairstyles; (2) a strand-guided optimization module that learns a dynamic 3D Gaussian representation for hair appearance with support for differentiable rendering, enabling gradient-based learning of view-consistent appearance under motion. Unlike prior simulation-based pipelines, our approach is fully data-driven, scales with training data, and generalizes across various hairstyles and head motion sequences. Additionally, DGH can be seamlessly integrated into a 3D Gaussian avatar framework, enabling realistic, animatable hair for high-fidelity avatar representation. DGH achieves promising geometry and appearance results, providing a scalable, data-driven alternative to physics-based simulation and rendering.

cs.CV↗

arXiv2025

Evaluating Generalization Capabilities of LLM-Based Agents in Mixed-Motive Scenarios Using Concordia

Large Language Model (LLM) agents have demonstrated impressive capabilities for social interaction and are increasingly being deployed in situations where they might engage with both human and artificial agents. These interactions represent a critical frontier for LLM-based agents, yet existing evaluation methods fail to measure how well these capabilities generalize to novel social situations. In this paper, we introduce a method for evaluating the ability of LLM-based agents to cooperate in zero-shot, mixed-motive environments using Concordia, a natural language multi-agent simulation environment. Our method measures general cooperative intelligence by testing an agent's ability to identify and exploit opportunities for mutual gain across diverse partners and contexts. We present empirical results from the NeurIPS 2024 Concordia Contest, where agents were evaluated on their ability to achieve mutual gains across a suite of diverse scenarios ranging from negotiation to collective action problems. Our findings reveal significant gaps between current agent capabilities and the robust generalization required for reliable cooperation, particularly in scenarios demanding persuasion and norm enforcement.

cs.AI↗

arXiv2025

Think Before You Prune: Self-Reflective Structured Pruning for Reasoning Language Models

Reasoning LLMs (RLMs) such as OpenAI o1, DeepSeek-R1, and Qwen3 deliver strong multi-step reasoning through chain-of-thought generation, but their large model sizes and lengthy decode-time outputs make them costly to deploy and unsuitable for resource-constrained settings. To reduce computing and memory cost, pruning offers a promising solution by removing unimportant parameters. However, despite their success on standard LLMs, existing pruning methods severely damage RLMs, as even moderate sparsity (e.g., 20%) can collapse accuracy and completely disrupt the model's reasoning coherence. We begin by analyzing why existing pruning pipelines fail on reasoning LLMs and find that their brittleness largely stems from a mismatch between the calibration data, the pruning objective, and the model's decode-time reasoning behavior. Our study further shows that the most reliable calibration signal comes not from human-written labels but from the model's own self-generated reasoning traces, which more accurately reflect its inference distribution. Guided by these insights, we introduce RESP, a self-reflective structured pruning framework that aligns pruning decisions with the model's reasoning dynamics through self-generated calibration, decode-only gradient-based importance estimation, and progressive regeneration that maintains calibration fidelity as sparsity increases. Experiments on Qwen3-8B demonstrate that RESP markedly outperforms existing structured pruning methods on both GSM8K and MathQA, preserving near-dense accuracy at 20-30% sparsity and substantially mitigating performance collapse at higher sparsity levels. At 40% sparsity, RESP attains 81.3% accuracy on GSM8K and 59.6% on MathQA, surpassing the strongest baselines by 66.87% and 47%, respectively.

cs.CL↗

arXiv2025

WorldGen: From Text to Traversable and Interactive 3D Worlds

We introduce WorldGen, a system that enables the automatic creation of large-scale, interactive 3D worlds directly from text prompts. Our approach transforms natural language descriptions into traversable, fully textured environments that can be immediately explored or edited within standard game engines. By combining LLM-driven scene layout reasoning, procedural generation, diffusion-based 3D generation, and object-aware scene decomposition, WorldGen bridges the gap between creative intent and functional virtual spaces, allowing creators to design coherent, navigable worlds without manual modeling or specialized 3D expertise. The system is fully modular and supports fine-grained control over layout, scale, and style, producing worlds that are geometrically consistent, visually rich, and efficient to render in real time. This work represents a step towards accessible, generative world-building at scale, advancing the frontier of 3D generative AI for applications in gaming, simulation, and immersive social environments.

cs.CV↗

arXiv2025

Tipping Points and Cascading Transitions: Methods, Principles, and Evidences

This review synthesizes recent advancements in understanding tipping points and cascading transitions within the Earth system, framing them through the lens of nonlinear dynamics and complexity science. It outlines the fundamental concepts of tipping elements, large-scale subsystems like the Atlantic Meridional Overturning Circulation and the Amazon rainforest, and classifies tipping mechanisms into bifurcation-, noise-, and rate-induced types. The article critically evaluates methods for detecting early-warning signals, particularly those based on critical slowing down, while also acknowledging their limitations and the promise of non-conventional indicators. Furthermore, we explore the significant risk of cascading failures between interacting tipping elements, often modeled using conceptual network models. This shows that such interactions can substantially increase systemic risk under global warming. The review concludes by outlining key challenges related to data limitations and methodological robustness, and emphasizes the promising role of artificial intelligence and complex network science in advancing prediction and risk assessment of Earth system tipping dynamics.

physics.ao-ph↗

arXiv2025

ExplorAR: Assisting Older Adults to Learn Smartphone Apps through AR-powered Trial-and-Error with Interactive Guidance

Older adults tend to encounter challenges when learning to use new smartphone apps due to age-related cognitive and physical changes. Compared to traditional support methods such as video tutorials, trial-and-error allows older adults to learn to use smartphone apps by making and correcting mistakes. However, it remains unknown how trial-and-error should be designed to empower older adults to use smartphone apps and how well it would work for older adults. Informed by the guidelines derived from prior work, we designed and implemented ExplorAR, an AR-based trial-and-error system that offers real-time and situated visual guidance in the augmented space around the smartphone to empower older adults to explore and correct mistakes independently. We conducted a user study with 18 older adults to compare ExplorAR with traditional video tutorials and a simplified version of ExplorAR. Results show that the AR-supported trial-and-error method enhanced older adults' learning experience by fostering deeper cognitive engagement and improving confidence in exploring unknown operations.

cs.HC↗

arXiv2025

PKAG-DDI: Pairwise Knowledge-Augmented Language Model for Drug-Drug Interaction Event Text Generation

Drug-drug interactions (DDIs) arise when multiple drugs are administered concurrently. Accurately predicting the specific mechanisms underlying DDIs (named DDI events or DDIEs) is critical for the safe clinical use of drugs. DDIEs are typically represented as textual descriptions. However, most computational methods focus more on predicting the DDIE class label over generating human-readable natural language increasing clinicians' interpretation costs. Furthermore, current methods overlook the fact that each drug assumes distinct biological functions in a DDI, which, when used as input context, can enhance the understanding of the DDIE process and benefit DDIE generation by the language model (LM). In this work, we propose a novel pairwise knowledge-augmented generative method (termed PKAG-DDI) for DDIE text generation. It consists of a pairwise knowledge selector efficiently injecting structural information between drugs bidirectionally and simultaneously to select pairwise biological functions from the knowledge set, and a pairwise knowledge integration strategy that matches and integrates the selected biological functions into the LM. Experiments on two professional datasets show that PKAG-DDI outperforms existing methods in DDIE text generation, especially in challenging inductive scenarios, indicating its practicality and generalization.

q-bio.BM↗

arXiv2025

LLM-Driven Dual-Level Multi-Interest Modeling for Recommendation

Recently, much effort has been devoted to modeling users' multi-interests based on their behaviors or auxiliary signals. However, existing methods often rely on heuristic assumptions, e.g., co-occurring items indicate the same interest of users, failing to capture user multi-interests aligning with real-world scenarios. While large language models (LLMs) show significant potential for multi-interest analysis due to their extensive knowledge and powerful reasoning capabilities, two key challenges remain. First, the granularity of LLM-driven multi-interests is agnostic, possibly leading to overly fine or coarse interest grouping. Second, individual user analysis provides limited insights due to the data sparsity issue. In this paper, we propose an LLM-driven dual-level multi-interest modeling framework for more effective recommendation. At the user-individual level, we exploit LLMs to flexibly allocate items engaged by users into different semantic clusters, indicating their diverse and distinct interests. To alleviate the agnostic generation of LLMs, we adaptively assign these semantic clusters to users' collaborative multi-interests learned from global user-item interactions, allowing the granularity to be automatically adjusted according to the user's behaviors using an alignment module. To alleviate the limited insights derived from individual users' behaviors, at the user-crowd level, we propose aggregating user cliques into synthesized users with rich behaviors for more comprehensive LLM-driven multi-interest analysis. We formulate a max covering problem to ensure the compactness and representativeness of synthesized users' behaviors, and then conduct contrastive learning based on their LLM-driven multi-interests to disentangle item representations among different interests. Experiments on real-world datasets show the superiority of our approach against state-of-the-art methods.

cs.IR↗

arXiv2025

RemVerse: Supporting Reminiscence Activities for Older Adults through AI-Assisted Virtual Reality

Reminiscence activities, which involve recalling and sharing past experiences, have proven beneficial for improving cognitive function, mood, and overall well-being. However, urbanization has led to the disappearance of familiar environments, removing visual and audio cues for effective reminiscence. While old photos can serve as visual cues to aid reminiscence, it is challenging for people to reconstruct the reminisced content and environment that are not in the photos. Virtual reality (VR) and artificial intelligence (AI) offer the ability to reconstruct an immersive environment with dynamic content and to converse with people to help them gradually reminisce. We designed RemVerse, an AI-empowered VR prototype aimed to support reminiscence activities. Integrating generative models and AI agent into a VR environment, RemVerse helps older adults reminisce with AI-generated visual cues and interactive dialogues. Our user study with 14 older adults showed that RemVerse effectively supported reminiscence activities by triggering, concretizing, and deepening personal memories, while fostering increased engagement and autonomy among older adults. Based on our findings, we proposed design implications to make reminiscence activities in AI-assisted VR more accessible and engaging for older adults.

cs.HC↗

arXiv2025

Learning Instruction-Following Policies through Open-Ended Instruction Relabeling with Large Language Models

Developing effective instruction-following policies in reinforcement learning remains challenging due to the reliance on extensive human-labeled instruction datasets and the difficulty of learning from sparse rewards. In this paper, we propose a novel approach that leverages the capabilities of large language models (LLMs) to automatically generate open-ended instructions retrospectively from previously collected agent trajectories. Our core idea is to employ LLMs to relabel unsuccessful trajectories by identifying meaningful subtasks the agent has implicitly accomplished, thereby enriching the agent's training data and substantially alleviating reliance on human annotations. Through this open-ended instruction relabeling, we efficiently learn a unified instruction-following policy capable of handling diverse tasks within a single policy. We empirically evaluate our proposed method in the challenging Craftax environment, demonstrating clear improvements in sample efficiency, instruction coverage, and overall policy performance compared to state-of-the-art baselines. Our results highlight the effectiveness of utilizing LLM-guided open-ended instruction relabeling to enhance instruction-following reinforcement learning.

cs.LG↗

arXiv2025

PoGDiff: Product-of-Gaussians Diffusion Models for Imbalanced Text-to-Image Generation

Diffusion models have made significant advancements in recent years. However, their performance often deteriorates when trained or fine-tuned on imbalanced datasets. This degradation is largely due to the disproportionate representation of majority and minority data in image-text pairs. In this paper, we propose a general fine-tuning approach, dubbed PoGDiff, to address this challenge. Rather than directly minimizing the KL divergence between the predicted and ground-truth distributions, PoGDiff replaces the ground-truth distribution with a Product of Gaussians (PoG), which is constructed by combining the original ground-truth targets with the predicted distribution conditioned on a neighboring text embedding. Experiments on real-world datasets demonstrate that our method effectively addresses the imbalance problem in diffusion models, improving both generation accuracy and quality.

cs.LG↗

arXiv2025

SRLAgent: Enhancing Self-Regulated Learning Skills through Gamification and LLM Assistance

Self-regulated learning (SRL) is crucial for college students navigating increased academic demands and independence. Insufficient SRL skills can lead to disorganized study habits, low motivation, and poor time management, undermining learners ability to thrive in challenging environments. Through a formative study involving 59 college students, we identified key challenges students face in developing SRL skills, including difficulties with goal-setting, time management, and reflective learning. To address these challenges, we introduce SRLAgent, an LLM-assisted system that fosters SRL skills through gamification and adaptive support from large language models (LLMs). Grounded in Zimmermans three-phase SRL framework, SRLAgent enables students to engage in goal-setting, strategy execution, and self-reflection within an interactive game-based environment. The system offers real-time feedback and scaffolding powered by LLMs to support students independent study efforts. We evaluated SRLAgent using a between-subjects design, comparing it to a baseline system (SRL without Agent features) and a traditional multimedia learning condition. Results showed significant improvements in SRL skills within the SRLAgent group (p < .001, Cohens d = 0.234) and higher engagement compared to the baselines. This work highlights the value of embedding SRL scaffolding and real-time AI support within gamified environments, offering design implications for educational technologies that aim to promote deeper learning and metacognitive skill development.

cs.HC↗

arXiv2025

M3HF: Multi-agent Reinforcement Learning from Multi-phase Human Feedback of Mixed Quality

Designing effective reward functions in multi-agent reinforcement learning (MARL) is a significant challenge, often leading to suboptimal or misaligned behaviors in complex, coordinated environments. We introduce Multi-agent Reinforcement Learning from Multi-phase Human Feedback of Mixed Quality ($\text{M}^3\text{HF}$), a novel framework that integrates multi-phase human feedback of mixed quality into the MARL training process. By involving humans with diverse expertise levels to provide iterative guidance, $\text{M}^3\text{HF}$ leverages both expert and non-expert feedback to continuously refine agents' policies. During training, we strategically pause agent learning for human evaluation, parse feedback using large language models to assign it appropriately and update reward functions through predefined templates and adaptive weights by using weight decay and performance-based adjustments. Our approach enables the integration of nuanced human insights across various levels of quality, enhancing the interpretability and robustness of multi-agent cooperation. Empirical results in challenging environments demonstrate that $\text{M}^3\text{HF}$ significantly outperforms state-of-the-art methods, effectively addressing the complexities of reward design in MARL and enabling broader human participation in the training process.

cs.MA↗

arXiv2025

Towards Domain Adaptive Neural Contextual Bandits

Contextual bandit algorithms are essential for solving real-world decision making problems. In practice, collecting a contextual bandit's feedback from different domains may involve different costs. For example, measuring drug reaction from mice (as a source domain) and humans (as a target domain). Unfortunately, adapting a contextual bandit algorithm from a source domain to a target domain with distribution shift still remains a major challenge and largely unexplored. In this paper, we introduce the first general domain adaptation method for contextual bandits. Our approach learns a bandit model for the target domain by collecting feedback from the source domain. Our theoretical analysis shows that our algorithm maintains a sub-linear regret bound even adapting across domains. Empirical results show that our approach outperforms the state-of-the-art contextual bandit algorithms on real-world datasets.

cs.LG↗

arXiv2025

FRESA: Feedforward Reconstruction of Personalized Skinned Avatars from Few Images

We present a novel method for reconstructing personalized 3D human avatars with realistic animation from only a few images. Due to the large variations in body shapes, poses, and cloth types, existing methods mostly require hours of per-subject optimization during inference, which limits their practical applications. In contrast, we learn a universal prior from over a thousand clothed humans to achieve instant feedforward generation and zero-shot generalization. Specifically, instead of rigging the avatar with shared skinning weights, we jointly infer personalized avatar shape, skinning weights, and pose-dependent deformations, which effectively improves overall geometric fidelity and reduces deformation artifacts. Moreover, to normalize pose variations and resolve coupled ambiguity between canonical shapes and skinning weights, we design a 3D canonicalization process to produce pixel-aligned initial conditions, which helps to reconstruct fine-grained geometric details. We then propose a multi-frame feature aggregation to robustly reduce artifacts introduced in canonicalization and fuse a plausible avatar preserving person-specific identities. Finally, we train the model in an end-to-end framework on a large-scale capture dataset, which contains diverse human subjects paired with high-quality 3D scans. Extensive experiments show that our method generates more authentic reconstruction and animation than state-of-the-arts, and can be directly generalized to inputs from casually taken phone photos. Project page and code is available at https://github.com/rongakowang/FRESA.

cs.CV↗

arXiv2025

Probe Pruning: Accelerating LLMs through Dynamic Pruning via Model-Probing

We introduce Probe Pruning (PP), a novel framework for online, dynamic, structured pruning of Large Language Models (LLMs) applied in a batch-wise manner. PP leverages the insight that not all samples and tokens contribute equally to the model's output, and probing a small portion of each batch effectively identifies crucial weights, enabling tailored dynamic pruning for different batches. It comprises three main stages: probing, history-informed pruning, and full inference. In the probing stage, PP selects a small yet crucial set of hidden states, based on residual importance, to run a few model layers ahead. During the history-informed pruning stage, PP strategically integrates the probing states with historical states. Subsequently, it structurally prunes weights based on the integrated states and the PP importance score, a metric developed specifically to assess the importance of each weight channel in maintaining performance. In the final stage, full inference is conducted on the remaining weights. A major advantage of PP is its compatibility with existing models, as it operates without requiring additional neural network modules or fine-tuning. Comprehensive evaluations of PP on LLaMA-2/3 and OPT models reveal that even minimal probing-using just 1.5% of FLOPs-can substantially enhance the efficiency of structured pruning of LLMs. For instance, when evaluated on LLaMA-2-7B with WikiText2, PP achieves a 2.56 times lower ratio of performance degradation per unit of runtime reduction compared to the state-of-the-art method at a 40% pruning ratio. Our code is available at https://github.com/Qi-Le1/Probe_Pruning.

cs.CL↗