Author name cluster

Jian Guan

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

13 papers

1 author row

AAAI Conference 2026 Conference Paper

MetaAct-RL: Training Language Models for Reasoning Through Meta-Action-Based Reinforcement Learning

Zhiheng Xi
Yuhui Wang
Yiwen Ding
Guanyu Li
Senjie Jin
Shichun Liu
Jixuan Huang
Dingwen Yang

Outcome-based reinforcement learning has made notable advances in training language models (LMs) for reasoning. However, without explicit incentives and controls, this paradigm has limitations and instability in eliciting high-quality reasoning trajectories with diverse actions—particularly for models whose pretraining lacked extensive reasoning-related data. To this end, we introduce MetaAct-RL, a new RL framework that frames LMs’ thinking as sequential decision making over meta-actions. In this framework, the model chooses and executes a high-level action at each step—such as forward reasoning, critique, or refinement—to gradually reach the correct answer. To encourage deeper exploration, richer action diversity, and to improve sampling efficiency in the RL optimization process, MetaAct-RL incorporates appropriate length-based reward and regularization, and a key-state restart mechanism. Extensive experiments across six benchmarks show that MetaAct-RL improves reasoning performance by 7.99 on Llama3.2-1B and 7.17 on Llama3.1-8B relative to vanilla RL method. Moreover, on the challenging AIME-2024, our method outperforms the vanilla RL by 7.5 with Qwen2.5-1.5B.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

DynaAct: Large Language Model Reasoning with Dynamic Action Spaces

Xueliang Zhao
Wei Wu
Jian Guan
Qintong Li
Lingpeng Kong

In modern sequential decision-making systems, the construction of an optimal candidate action space is critical to efficient inference. However, existing approaches either rely on manually defined action spaces that lack scalability or utilize unstructured spaces that render exhaustive search computationally prohibitive. In this paper, we propose a novel framework named \textsc{DynaAct} for automatically constructing a compact action space to enhance sequential reasoning in complex problem-solving scenarios. Our method first estimates a proxy for the complete action space by extracting general sketches observed in a corpus covering diverse complex reasoning problems using large language models. We then formulate a submodular function that jointly evaluates candidate actions based on their utility to the current state and their diversity, and employ a greedy algorithm to select an optimal candidate set. Extensive experiments on six diverse standard benchmarks demonstrate that our approach significantly improves overall performance, while maintaining efficient inference without introducing substantial latency. The implementation is available at \url{https: //github. com/zhaoxlpku/DynaAct}.

PDF Details

NeurIPS Conference 2025 Conference Paper

Reinforcing Spatial Reasoning in Vision-Language Models with Interwoven Thinking and Visual Drawing

Junfei Wu
Jian Guan
Kaituo Feng
Qiang Liu
Shu Wu
Liang Wang
Wei Wu
Tieniu Tan

As textual reasoning with large language models (LLMs) has advanced significant, there has been growing interest in enhancing the multimodal reasoning capabilities of large vision-language models (LVLMs). However, existing methods primarily approach multimodal reasoning in a straightforward, text-centric manner, where both reasoning and answer derivation are conducted purely through text, with the only difference being the presence of multimodal input. As a result, these methods often encounter fundamental limitations in spatial reasoning tasks that demand precise geometric understanding and continuous spatial tracking\textemdash capabilities that humans achieve through mental visualization and manipulation. To address the limitations, we propose drawing to reason in space, a novel paradigm that enables LVLMs to reason through elementary drawing operations in the visual space. By equipping models with basic drawing operations including annotating bounding boxes and drawing auxiliary lines, we empower them to express and analyze spatial relationships through direct visual manipulation, meanwhile avoiding the performance ceiling imposed by specialized perception tools in previous tool-integrated reasoning approaches. To cultivate this capability, we develop a three-stage training framework: cold-start training with synthetic data to establish basic drawing abilities, reflective rejection sampling to enhance self-reflection behaviors, and reinforcement learning to directly optimize for target rewards. Extensive experiments demonstrate that our model, named \textsc{Spark}, consistently outperforms existing methods across diverse spatial reasoning benchmarks involving maze navigation, static spatial reasoning, video-based reasoning and multi-view-based reasoning tasks, with an average improvement of 11. 5\%. Ablation studies reveal the critical role of each training stage, with reflective rejection sampling particularly enhancing the model's self-correction capabilities and reasoning potential.

PDF Details

NeurIPS Conference 2025 Conference Paper

Theoretical Benefit and Limitation of Diffusion Language Model

Guhao Feng
Yihan Geng
Jian Guan
Wei Wu
Liwei Wang
Di He

Diffusion language models have emerged as a new approach for text generation. By enabling the parallel sampling of multiple tokens in each diffusion step, they appear to offer a more efficient alternative to auto-regressive models. However, our observations show that current open-sourced diffusion language models require more sampling steps to achieve comparable accuracy on representative tasks--resulting in even higher inference costs than their auto-regressive counterparts. To investigate whether this is an inherent limitation, we conduct a rigorous theoretical analysis of a widely adopted variant: the Masked Diffusion Model (MDM). Surprisingly, our analysis reveals that the conclusion is highly sensitive to the choice of evaluation metric. Under mild conditions, we prove that when the target is near-optimal perplexity, MDMs can achieve this goal in a constant number of sampling steps, independent of sequence length. This result demonstrates that efficiency can, in principle, be attained without compromising generation quality. However, when targeting low sequence error rate--which is important for assessing the ``correctness" of a generated sequence, such as a reasoning chain--we show that in the worst case, the required sampling steps must scale linearly with sequence length, thereby eliminating the efficiency advantage. Our analysis establishes the first theoretical foundation for understanding the comparative strengths and limitations of MDMs, offering practical guidance on when to favor MDMs over the auto-regressive models and vice versa.

PDF Details

NeurIPS Conference 2024 Conference Paper

AMOR: A Recipe for Building Adaptable Modular Knowledge Agents Through Process Feedback

Jian Guan
Wei Wu
Zujie Wen
Peng Xu
Hongning Wang
Minlie Huang

The notable success of large language models (LLMs) has sparked an upsurge in building language agents to complete various complex tasks. We present AMOR, an agent framework based on open-source LLMs, which reasons with external knowledge bases and adapts to specific domains through human supervision to the reasoning process. AMOR builds reasoning logic over a finite state machine (FSM)that solves problems through autonomous executions and transitions over disentangled modules. This allows humans to provide direct feedback to the individual modules, and thus naturally forms process supervision. Based on this reasoning and feedback framework, we develop AMOR through two-stage fine-tuning: warm-up and adaptation. The former fine-tunes the LLM with examples automatically constructed from various public datasets, enabling AMOR to generalize across different knowledge environments, while the latter tailors AMOR to specific domains using process feedback. Extensive experiments across multiple domains demonstrate the advantage of AMOR to strong baselines, thanks to its FSM-based reasoning and process feedback mechanism. The code and data are publicly available athttps: //github. com/JianGuanTHU/AMOR.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

FastDrag: Manipulate Anything in One Step

Xuanjia Zhao
Jian Guan
Congyi Fan
Dongli Xu
Youtian Lin
Haiwei Pan
Pengming Feng

Drag-based image editing using generative models provides precise control over image contents, enabling users to manipulate anything in an image with a few clicks. However, prevailing methods typically adopt $n$-step iterations for latent semantic optimization to achieve drag-based image editing, which is time-consuming and limits practical applications. In this paper, we introduce a novel one-step drag-based image editing method, i. e. , FastDrag, to accelerate the editing process. Central to our approach is a latent warpage function (LWF), which simulates the behavior of a stretched material to adjust the location of individual pixels within the latent space. This innovation achieves one-step latent semantic optimization and hence significantly promotes editing speeds. Meanwhile, null regions emerging after applying LWF are addressed by our proposed bilateral nearest neighbor interpolation (BNNI) strategy. This strategy interpolates these regions using similar features from neighboring areas, thus enhancing semantic integrity. Additionally, a consistency-preserving strategy is introduced to maintain the consistency between the edited and original images by adopting semantic information from the original image, saved as key and value pairs in self-attention module during diffusion inversion, to guide the diffusion sampling. Our FastDrag is validated on the DragBench dataset, demonstrating substantial improvements in processing time over existing methods, while achieving enhanced editing performance.

PDF Details DOI

AAAI Conference 2024 Conference Paper

Preference Aware Dual Contrastive Learning for Item Cold-Start Recommendation

Wenbo Wang
Bingquan Liu
Lili Shan
Chengjie Sun
Ben Chen
Jian Guan

Existing cold-start recommendation methods often adopt item-level alignment strategies to align the content feature and the collaborative feature of warm items for model training, however, cold items in the test stage have no historical interactions with users to obtain the collaborative feature. These existing models ignore the aforementioned condition of cold items in the training stage, resulting in the performance limitation. In this paper, we propose a preference aware dual contrastive learning based recommendation model (PAD-CLRec), where the user preference is explored to take into account the condition of cold items for feature alignment. Here, the user preference is obtained by aggregating a group of collaborative feature of the warm items in the user's purchase records. Then, a group-level alignment between the user preference and the item's content feature can be realized via a proposed preference aware contrastive function for enhancing cold-item recommendation. In addition, a joint objective function is introduced to achieve a better trade-off between the recommendation performance of warm items and cold items from both item-level and group-level perspectives, yielding better overall recommendation performance. Extensive experiments are conducted to demonstrate the effectiveness of the proposed method, and the results show the superiority of our method, as compared with the state-of-the-arts.

PDF Details DOI

AAAI Conference 2023 Conference Paper

Generating Coherent Narratives by Learning Dynamic and Discrete Entity States with a Contrastive Framework

Jian Guan
Zhenyu Yang
Rongsheng Zhang
Zhipeng Hu
Minlie Huang

Despite advances in generating fluent texts, existing pretraining models tend to attach incoherent event sequences to involved entities when generating narratives such as stories and news. We conjecture that such issues result from representing entities as static embeddings of superficial words, while neglecting to model their ever-changing states, i.e., the information they carry, as the text unfolds. Therefore, we extend the Transformer model to dynamically conduct entity state updates and sentence realization for narrative generation. We propose a contrastive framework to learn the state representations in a discrete space, and insert additional attention layers into the decoder to better exploit these states. Experiments on two narrative datasets show that our model can generate more coherent and diverse narratives than strong baselines with the guidance of meaningful entity states.

PDF Details DOI

JBHI Journal 2023 Journal Article

Learning Common and Task-Specific Radiomic Features via Graph Regularized NMF for the Joint Prediction of Multiple Clinical Indicators in Breast Cancer

Jian Guan
Ming Fan
Tieyong Zeng
Lihua Li

Assessments of multiple clinical indicators based on radiomic analysis of magnetic resonance imaging (MRI) are beneficial to the diagnosis, prognosis and treatment of breast cancer patients. Many machine learning methods have been designed to jointly predict multiple indicators for more accurate assessments while using original clinical labels directly without considering the noisy and redundant information among them. To this end, we propose a multilabel learning method based on label space dimensionality reduction (LSDR), which learns common and task-specific features via graph regularized nonnegative matrix factorization (CTFGNMF) for the joint prediction of multiple indicators in breast cancer. A nonnegative matrix factorization (NMF) is adopted to map original clinical labels to a low-dimensional latent space. The latent labels are employed to exploit task correlations by using a least square loss function with ${{\bm{l}}}_{2, 1}$ -norm regularization to identify common features, which help to improve the generalization performance of correlated tasks. Furthermore, task-specific features were retained by a multitask regression formulation to increase the discrimination power for different tasks. Common and task-specific features are incorporated by dynamic graph Laplacian regularization into a unified model to learn complementary features. Then, a multilabel classification is built to predict multiple clinical indicators including human epidermal growth factor receptor 2 (HER2), Ki-67, and histological grade. Experimental results show that CTFGNMF achieves AUCs of 0. 823, 0. 691 and 0. 776 in the three indicator predictions, outperforming other counterparts that consider only task-independent features or common features. It indicates CTFGNMF is a promising application for multiple classification tasks in breast cancer.

Details DOI

NeurIPS Conference 2019 Conference Paper

A Model-Based Reinforcement Learning with Adversarial Training for Online Recommendation

Xueying Bai
Jian Guan
Hongning Wang

Reinforcement learning is effective in optimizing policies for recommender systems. Current solutions mostly focus on model-free approaches, which require frequent interactions with a real environment, and thus are expensive in model learning. Offline evaluation methods, such as importance sampling, can alleviate such limitations, but usually request a large amount of logged data and do not work well when the action space is large. In this work, we propose a model-based reinforcement learning solution which models the user-agent interaction for offline policy learning via a generative adversarial network. To reduce bias in the learnt policy, we use the discriminator to evaluate the quality of generated sequences and rescale the generated rewards. Our theoretical analysis and empirical evaluations demonstrate the effectiveness of our solution in identifying patterns from given offline data and learning policies based on the offline and generated data.

PDF Details

EAAI Journal 2019 Journal Article

Batch and data streaming classification models for detecting adverse events and understanding the influencing factors

Donghui Shi
Jozef Zurada
Waldemar Karwowski
Jian Guan
Erman Çakıt

Constructing effective models for detecting, reducing, and/or preventing adverse events is very important in domains such as aviation safety, healthcare, drug administration, and war theaters. This study presents batch and data streaming models to detecting adverse events using data from a war theater context. In all the previous studies, regression models and several machine learning techniques were used for predicting continuous values in an active theater of war, and the error values reported on the test sets were large. In order to overcome the shortcoming, this study investigates the effectiveness of batch and data streaming classification algorithms in detecting or classifying adverse events given infrastructure development spending data and other variables in an active theater of war in Afghanistan. By the feature selection, the valid input variables are obtained and their indexes show that the input variables are mainly the adverse events (t-1) at the previous month, the population densities and related project investments. From the country level, fewer of the 14 project investments affect the adverse events. From the region level, some projects with higher index values, such as Security in the South Western region, Energy and Emergency Assistance in the North Eastern region, and Education in the Eastern region are mainly affecting factors. Three batch classification methods and three data streaming classification methods were assessed for their ability to detect adverse events given infrastructure development data. The study uses cost-sensitive measures to address the very unbalanced nature of the data and it applies variable reduction techniques to identify significant variables. The three batch classification algorithms are C4. 5, k-nearest Neighbor, and Support Vector Machine. The three data streaming algorithms are Naïve Bayes, Hoeffding Tree, and Single Classifier Drift. In general, the performance of the cost-sensitive methods in the batch setting is comparable to those in the data stream setting. However, in the batch setting the cost matrix needs to be adjusted manually. In contrast the data stream setting allows one to adjust the models based on the analysis of the classifiers’ performance over time and changing data distribution. The Kappa values using Naïve Bayes are the highest in the three data stream algorithms in the whole country and its regions. The Naïve Bayes classifier has the best global performance. By the Kappa statistic curve, we can observe the concept drifts. In a region level, many models have a better performance including more investments related to project compared with those in a country level. In addition as data distribution becomes more balanced, the classifiers in the data stream setting outperform in terms of the overall classification rates in comparison to the classifiers in the batch setting. The results thus demonstrate the potential of data streaming algorithms to significantly outperform when the data become less unbalanced, and can be used for detecting adverse events in similar areas.

Details DOI

AAAI Conference 2019 Conference Paper

Story Ending Generation with Incremental Encoding and Commonsense Knowledge

Jian Guan
Yansen Wang
Minlie Huang

Generating a reasonable ending for a given story context, i. e. , story ending generation, is a strong indication of story comprehension. This task requires not only to understand the context clues which play an important role in planning the plot, but also to handle implicit knowledge to make a reasonable, coherent story. In this paper, we devise a novel model for story ending generation. The model adopts an incremental encoding scheme to represent context clues which are spanning in the story context. In addition, commonsense knowledge is applied through multi-source attention to facilitate story comprehension, and thus to help generate coherent and reasonable endings. Through building context clues and using implicit knowledge, the model is able to produce reasonable story endings. Automatic and manual evaluation shows that our model can generate more reasonable story endings than state-of-the-art baselines1.

PDF Details

EAAI Journal 1993 Journal Article

A hybrid diagnostic system with learning capabilities

James H. Graham
Jian Guan
Suraj M. Alexander

Fast and accurate diagnosis of faults in computer integrated manufacturing systems is essential in order to avoid excessive equipment downtime, and to take full advantage of these systems. Traditional approaches to diagnosis have yielded to artificial intelligence approaches over recent years, as system complexity has increased; but results have been mixed. Symptom-based approaches have been too limited, while structural-based approaches have required excessive computational resources. This paper presents a hybrid model for diagnostics that is computationally efficient, and at the same time incorporates the potential to improve its performance with use through a learning scheme.

Details DOI