Author name cluster

Yijun Wang

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.

15 papers

2 author rows

AAAI Conference 2026 Conference Paper

MedEyes: Learning Dynamic Visual Focus for Medical Progressive Diagnosis

Chunzheng Zhu
Yangfang Lin
Shen Chen
Yijun Wang
Jianxin Lin

Accurate medical diagnosis often involves progressive visual focusing and iterative reasoning, characteristics commonly observed in clinical workflows. While recent vision-language models demonstrate promising chain-of-thought (CoT) reasoning capabilities via reinforcement learning with verifiable rewards (RLVR), their purely on-policy learning paradigm tends to reinforce superficially coherent but clinically inaccurate reasoning paths. We propose MedEyes, a novel reinforcement learning framework that dynamically models clinician-style diagnostic reasoning by progressively attending to and interpreting relevant medical image regions. By incorporating off-policy expert guidance, MedEyes converts expert visual search trajectories into structured external behavioral signals, guiding the model toward clinically aligned visual reasoning. We design the Gaze-guided Reasoning Navigator (GRN) to emulate the diagnostic process through a dual-mode exploration strategy, scanning for systematic abnormality localization and drilling for detailed regional analysis. To balance expert imitation and autonomous discovery, we introduce the Confidence Value Sampler (CVS), which employs nucleus sampling and adaptive termination to create diverse yet credible exploration paths. Finally, the dual-stream GRPO optimization framework decouples on-policy and off-policy learning signals, mitigating reward assimilation and entropy collapse. Experiments demonstrate that MedEyes achieves an average performance improvement of +8.5% across multiple medical VQA benchmarks, validating MedEyes's potential in building trustworthy medical AI systems.

PDF Details DOI

ICLR Conference 2025 Conference Paper

B-STaR: Monitoring and Balancing Exploration and Exploitation in Self-Taught Reasoners

Weihao Zeng
Yuzhen Huang 0002
Lulu Zhao
Yijun Wang
Zifei Shan
Junxian He

In the absence of extensive human-annotated data for complex reasoning tasks, self-improvement -- where models are trained on their own outputs -- has emerged as a primary method for enhancing performance. Recently, the approach to self-improvement has shifted toward a more dynamic, online fashion through iterative training processes. However, the critical factors underlying the mechanism of these self-improving methods remain poorly understood, such as under what conditions self-improvement is effective, and what are the bottlenecks in the current iterations. In this work, we identify and propose methods to monitor two pivotal factors in this iterative process: (1) the model's ability to explore and generate high-quality responses among multiple candidates (exploration); and (2) the reliability of external rewards in selecting the best responses from the generated outputs (exploitation). These factors are inherently moving targets throughout the self-improvement cycles, yet their dynamics are rarely discussed in prior research -- It remains unclear what impedes continual model enhancement after only a few iterations. Using mathematical reasoning as a case study, we begin with a quantitative analysis to track the dynamics of exploration and exploitation, discovering that a model's exploratory capabilities rapidly deteriorate over iterations, and the effectiveness of exploiting external rewards diminishes as well due to shifts in distribution from the original policy. Motivated by these findings, we introduce B-STaR, a Self-Taught Reasoning framework that autonomously adjusts configurations across iterations to Balance exploration and exploitation, thereby optimizing the self-teaching effectiveness based on the current policy model and available rewards. Our experiments in mathematical reasoning demonstrate that B-STaR not only enhances the model's exploratory capabilities throughout training but also achieves a more effective balance between exploration and exploitation, leading to superior performance. Crucially, this work deconstructs the opaque nature of self-training algorithms, elucidating the interpretable dynamics throughout the process and highlighting current limitations for future research to address.

Details

AAAI Conference 2025 Conference Paper

Debiased All-in-one Image Restoration with Task Uncertainty Regularization

Gang Wu
Junjun Jiang
Yijun Wang
Kui Jiang
Xianming Liu

All-in-one image restoration is a fundamental low-level vision task with significant real-world applications. The primary challenge lies in addressing diverse degradations within a single model. While current methods primarily exploit task prior information to guide the restoration models, they typically employ uniform multi-task learning, overlooking the heterogeneity in model optimization across different degradation tasks. To eliminate the bias, we propose a task-aware optimization strategy, that introduces adaptive task-specific regularization for multi-task image restoration learning. Specifically, our method dynamically weights and balances losses for different restoration tasks during training, encouraging the implementation of the most reasonable optimization route. In this way, we can achieve more robust and effective model training. Notably, our approach can serve as a plug-and-play strategy to enhance existing models without requiring modifications during inference. Extensive experiments in diverse all-in-one restoration settings demonstrate the superiority and generalization of our approach. For example, AirNet retrained with TUR achieves average improvements of 1.16 dB on three distinct tasks and 1.81 dB on five distinct all-in-one tasks. These results underscore TUR's effectiveness in advancing the SOTAs in all-in-one image restoration, paving the way for more robust and versatile image restoration.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

HARDMath2: A Benchmark for Applied Mathematics Built by Students as Part of a Graduate Class

James Roggeveen
Erik Wang
David Ettel
Will Flintoft
Peter Donets
Raglan Ward
Ahmed Roman
Anton Graf

Large language models (LLMs) have shown remarkable progress in mathematical problem-solving, but evaluation has largely focused on problems that have exact analytical solutions or involve formal proofs, often overlooking approximation-based problems ubiquitous in applied science and engineering. To fill this gap, we build on prior work and present $\textbf{HARDMath2}$, a dataset of 211 original problems covering the core topics in an introductory graduate applied math class, including boundary-layer analysis, WKB methods, asymptotic solutions of nonlinear partial differential equations, and the asymptotics of oscillatory integrals. This dataset was designed and verified by the students and instructors of a core graduate applied mathematics course at Harvard. We build the dataset through a novel collaborative environment that challenges students to write and refine difficult problems consistent with the class syllabus, peer-validate solutions, test different models, and automatically check LLM-generated solutions against their own answers and numerical ground truths. Evaluation results show that leading frontier models still struggle with many of the problems in the dataset, highlighting a gap in the mathematical reasoning skills of current LLMs. Importantly, students identified strategies to create increasingly difficult problems by interacting with the models and exploiting common failure modes. This back-and-forth with the models not only resulted in a richer and more challenging benchmark but also led to qualitative improvements in the students' understanding of the course material, which is increasingly important as we enter an age where state-of-the-art language models can solve many challenging problems across a wide domain of fields.

PDF Details

EAAI Journal 2025 Journal Article

SAFE-Net: Multi-head attention enhanced framework for defect detection in anti-corrosion coatings on steel structures

Yue Yu
Shouchao Jiang
Yijun Wang
Shaojun Zhu

Steel structures exhibit poor corrosion resistance and require anti-corrosion coatings for protection. Coating aging can lead to reduced or even ineffective anti-corrosion effects, so maintenance of their anti-corrosion coatings is crucial. Computer vision offers efficient and accurate analysis of visual data, but detecting defects in anti-corrosion coatings remains challenging due to low contrast, subtle background differences, varied defect types, image noise, and environmental interference. Existing networks often struggle to address these complexities effectively. In this context, this study proposes a You Only Look Once version 8 (YOLOv8)-driven SAFE-Net (Steel Anti Corruption Fault Evaluation Network) based on multiple attention mechanisms for detecting defects in anti-corrosion coatings on steel structures, and constructs a dataset for defect detection in in-service steel structure anti-corrosion coatings for model validation. The proposed framework integrates a multi-scale deformable attention mechanism into the Concatenate-to-Fuse module of the Backbone, aiming to improve the detection performance for small-scale targets. The Head adopts a Dynamic Head that integrates the self-attention mechanism to enhance scale perception, spatial perception, and task perception. The classification loss function is optimized by introducing the Slide Loss to address the challenge of imbalanced samples in the dataset, while Shape Intersection over Union refines the bounding box regression for better multi-scale detection. SAFE-Net supports multi-defect detection and demonstrates strong generalization. Numerical case studies indicate that, compared with the YOLOv8 baseline model, the improved algorithm improves the accuracy by 5. 82 %, recall by 5. 01 %, and mean Average Precision index by 4. 78 %, respectively.

Details DOI

IJCAI Conference 2024 Conference Paper

An LLM-enhanced Agent-based Simulation Tool for Information Propagation

Yuxuan Hu
Gemju Sherpa
Lan Zhang
Weihua Li
Quan Bai
Yijun Wang
Xiaodan Wang

Influence diffusion models are used for simulating information propagation in social networks. While most existing influence diffusion models are probabilistic, the emergence of Large Language Model (LLM) sheds light on the language-level inferences and interactions of user agents. This paper presents an LLM-enhanced Agent-based Influence Diffusion model (LAID), and a web-based visualization tool, LAIDSim, for simulating the information propagation in social networks.

PDF Details DOI

YNIMG Journal 2024 Journal Article

Estimating and approaching the maximum information rate of noninvasive visual brain-computer interface

Nanlin Shi
Yining Miao
Changxing Huang
Xiang Li
Yonghao Song
Xiaogang Chen
Yijun Wang
Xiaorong Gao

An essential priority of visual brain-computer interfaces (BCIs) is to enhance the information transfer rate (ITR) to achieve high-speed communication. Despite notable progress, noninvasive visual BCIs have encountered a plateau in ITRs, leaving it uncertain whether higher ITRs are achievable. In this study, we used information theory to study the characteristics and capacity of the visual-evoked channel, which leads us to investigate whether and how we can decode higher information rates in a visual BCI system. Using information theory, we estimate the upper and lower bounds of the information rate with the white noise (WN) stimulus. Consequently, we found out that the information rate is determined by the signal-to-noise ratio (SNR) in the frequency domain, which reflects the spectrum resources of the channel. Based on this discovery, we propose a broadband WN BCI by implementing stimuli on a broader frequency band than the steady-state visual evoked potentials (SSVEPs)-based BCI. Through validation, the broadband BCI outperforms the SSVEP BCI by an impressive 7 bps, setting a record of 50 bps. The integration of information theory and the decoding analysis presented in this study offers valuable insights applicable to general sensory-evoked BCIs, providing a potential direction of next-generation human-machine interaction systems.

Details DOI

AAAI Conference 2023 Conference Paper

Yet Another Traffic Classifier: A Masked Autoencoder Based Traffic Transformer with Multi-Level Flow Representation

Ruijie Zhao
Mingwei Zhan
Xianwen Deng
Yanhao Wang
Yijun Wang
Guan Gui
Zhi Xue

Traffic classification is a critical task in network security and management. Recent research has demonstrated the effectiveness of the deep learning-based traffic classification method. However, the following limitations remain: (1) the traffic representation is simply generated from raw packet bytes, resulting in the absence of important information; (2) the model structure of directly applying deep learning algorithms does not take traffic characteristics into account; and (3) scenario-specific classifier training usually requires a labor-intensive and time-consuming process to label data. In this paper, we introduce a masked autoencoder (MAE) based traffic transformer with multi-level flow representation to tackle these problems. To model raw traffic data, we design a formatted traffic representation matrix with hierarchical flow information. After that, we develop an efficient Traffic Transformer, in which packet-level and flow-level attention mechanisms implement more efficient feature extraction with lower complexity. At last, we utilize the MAE paradigm to pre-train our classifier with a large amount of unlabeled data, and perform fine-tuning with a few labeled data for a series of traffic classification tasks. Experiment findings reveal that our method outperforms state-of-the-art methods on five real-world traffic datasets by a large margin. The code is available at https://github.com/NSSL-SJTU/YaTC.

PDF Details DOI

IJCAI Conference 2022 Conference Paper

3E-Solver: An Effortless, Easy-to-Update, and End-to-End Solver with Semi-Supervised Learning for Breaking Text-Based Captchas

Xianwen Deng
Ruijie Zhao
Yanhao Wang
Libo Chen
Yijun Wang
Zhi Xue

Text-based captchas are the most widely used security mechanism currently. Due to the limitations and specificity of the segmentation algorithm, the early segmentation-based attack method has been unable to deal with the current captchas with newly introduced security features (e. g. , occluding lines and overlapping). Recently, some works have designed captcha solvers based on deep learning methods with powerful feature extraction capabilities, which have greater generality and higher accuracy. However, these works still suffer from two main intrinsic limitations: (1) many labor costs are required to label the training data, and (2) the solver cannot be updated with unlabeled data to recognize captchas more accurately. In this paper, we present a novel solver using improved FixMatch for semi-supervised captcha recognition to tackle these problems. Specifically, we first build an end-to-end baseline model to effectively break text-based captchas by leveraging encoder-decoder architecture and attention mechanism. Then we construct our solver with a few labeled samples and many unlabeled samples by improved FixMatch, which introduces teacher forcing, adaptive batch normalization, and consistency loss to achieve more effective training. Experiment results show that our solver outperforms state-of-the-arts by a large margin on current captcha schemes. We hope that our work can help security experts to revisit the design and usability of text-based captchas. The source code of this work is available at https: //github. com/SJTU-dxw/3E-Solver-CAPTCHA.

PDF Details DOI

AAAI Conference 2021 Conference Paper

Cross-Oilfield Reservoir Classification via Multi-Scale Sensor Knowledge Transfer

Zhi Li
Zhefeng Wang
Zhicheng Wei
Xiangguang Zhou
Yijun Wang
Baoxing Huai
Qi Liu
Nicholas Jing Yuan

Reservoir classification is an essential step for the exploration and production process in the oil and gas industry. An appropriate automatic reservoir classification will not only reduce the manual workloads of experts, but also help petroleum companies to make optimal decisions efficiently, which in turn will dramatically reduce the costs. Existing methods mainly focused on generating reservoir classification in a single geological block but failed to work well on a new oilfield block. Indeed, how to transfer the subsurface characteristics and make accurate reservoir classification across the geological oilfields is a very important but challenging problem. To that end, in this paper, we present a focused study on the cross-oilfield reservoir classification task. Specifically, we first propose a Multi-scale Sensor Extraction (MSE) module to extract the multi-scale feature representations of geological characteristics from multivariate well logs. Furthermore, we design an encoder-decoder module, i. e. , Specific Feature Learning (SFL), to take advantage of specific information of both oilfields. Then, we develop a Knowledge- Attentive Transfer (KAT) module to learn the feature-invariant representation and transfer the geological knowledge from a source oilfield to a target oilfield. Finally, we evaluate our approaches by conducting extensive experiments with realworld industrial datasets. The experimental results clearly demonstrate the effectiveness of our proposed approaches to transfer the geological knowledge and generate the crossoilfield reservoir classifications.

PDF Details

AAAI Conference 2020 Conference Paper

Learning to Transfer: Unsupervised Domain Translation via Meta-Learning

Jianxin Lin
Yijun Wang
Zhibo Chen
Tianyu He

Unsupervised domain translation has recently achieved impressive performance with Generative Adversarial Network (GAN) and sufﬁcient (unpaired) training data. However, existing domain translation frameworks form in a disposable way where the learning experiences are ignored and the obtained model cannot be adapted to a new coming domain. In this work, we take on unsupervised domain translation problems from a meta-learning perspective. We propose a model called Meta-Translation GAN (MT-GAN) to ﬁnd good initialization of translation models. In the meta-training procedure, MT- GAN is explicitly trained with a primary translation task and a synthesized dual translation task. A cycle-consistency metaoptimization objective is designed to ensure the generalization ability. We demonstrate effectiveness of our model on ten diverse two-domain translation tasks and multiple face identity translation tasks. We show that our proposed approach signiﬁcantly outperforms the existing domain translation methods when each domain contains no more than ten training samples.

PDF Details

IJCAI Conference 2019 Conference Paper

Image-to-Image Translation with Multi-Path Consistency Regularization

Jianxin Lin
Yingce Xia
Yijun Wang
Tao Qin
Zhibo Chen

Image translation across different domains has attracted much attention in both machine learning and computer vision communities. Taking the translation from a source domain to a target domain as an example, existing algorithms mainly rely on two kinds of loss for training: One is the discrimination loss, which is used to differentiate images generated by the models and natural images; the other is the reconstruction loss, which measures the difference between an original image and the reconstructed version. In this work, we introduce a new kind of loss, multi-path consistency loss, which evaluates the differences between direct translation from source domain to target domain and indirect translation from source domain to an auxiliary domain to target domain, to regularize training. For multi-domain translation (at least, three) which focuses on building translation models between any two domains, at each training iteration, we randomly select three domains, set them respectively as the source, auxiliary and target domains, build the multi-path consistency loss and optimize the network. For two-domain translation, we need to introduce an additional auxiliary domain and construct the multi-path consistency loss. We conduct various experiments to demonstrate the effectiveness of our proposed methods, including face-to-face translation, paint-to-photo translation, and de-raining/de-noising translation.

PDF Details

AAAI Conference 2018 Conference Paper

Dual Transfer Learning for Neural Machine Translation with Marginal Distribution Regularization

Yijun Wang
Yingce Xia
Li Zhao
Jiang Bian
Tao Qin
Guiquan Liu
Tie-Yan Liu

Neural machine translation (NMT) heavily relies on parallel bilingual data for training. Since large-scale, high-quality parallel corpora are usually costly to collect, it is appealing to exploit monolingual corpora to improve NMT. Inspired by the law of total probability, which connects the probability of a given target-side monolingual sentence to the conditional probability of translating from a source sentence to the target one, we propose to explicitly exploit this connection to learn from and regularize the training of NMT models using monolingual data. The key technical challenge of this approach is that there are exponentially many source sentences for a target monolingual sentence while computing the sum of the conditional probability given each possible source sentence. We address this challenge by leveraging the dual translation model (target-to-source translation) to sample several mostly likely source-side sentences and avoid enumerating all possible candidate source sentences. That is, we transfer the knowledge contained in the dual model to boost the training of the primal model (source-to-target translation), and we call such an approach dual transfer learning. Experiment results on English→French and German→English tasks demonstrate that dual transfer learning achieves signiﬁcant improvement over several strong baselines and obtains new state-of-the-art results.

PDF Details

YNIMG Journal 2017 Journal Article

Cognitive emotion regulation modulates the balance of competing influences on ventral striatal aversive prediction error signals

Satja Mulej Bratec
Xiyao Xie
Yijun Wang
Leonhard Schilbach
Claus Zimmer
Afra M. Wohlschläger
Valentin Riedl
Christian Sorg

Cognitive emotion regulation (CER) is a critical human ability to face aversive emotional stimuli in a flexible way, via recruitment of specific prefrontal brain circuits. Animal research reveals a central role of ventral striatum in emotional behavior, for both aversive conditioning, with striatum signaling aversive prediction errors (aPE), and for integrating competing influences of distinct striatal inputs from regions such as the prefrontal cortex (PFC), amygdala, hippocampus and ventral tegmental area (VTA). Translating these ventral striatal findings from animal research to human CER, we hypothesized that successful CER would affect the balance of competing influences of striatal afferents on striatal aPE signals, in a way favoring PFC as opposed to ‘subcortical’ (i. e. , non-isocortical) striatal inputs. Using aversive Pavlovian conditioning with and without CER during fMRI, we found that during CER, superior regulators indeed reduced the modulatory impact of ‘subcortical’ striatal afferents (hippocampus, amygdala and VTA) on ventral striatal aPE signals, while keeping the PFC impact intact. In contrast, inferior regulators showed an opposite pattern. Our results demonstrate that ventral striatal aPE signals and associated competing modulatory inputs are critical mechanisms underlying successful cognitive regulation of aversive emotions in humans.

Details DOI

IJCAI Conference 2017 Conference Paper

Incremental Matrix Factorization: A Linear Feature Transformation Perspective

Xunpeng Huang
Le Wu
Enhong Chen
Hengshu Zhu
Qi Liu
Yijun Wang

Matrix Factorization (MF) is among the most widely used techniques for collaborative filtering based recommendation. Along this line, a critical demand is to incrementally refine the MF models when new ratings come in an online scenario. However, most of existing incremental MF algorithms are limited by specific MF models or strict use restrictions. In this paper, we propose a general incremental MF framework by designing a linear transformation of user and item latent vectors over time. This framework shows a relatively high accuracy with a computation and space efficient training process in an online scenario. Meanwhile, we explain the framework with a low-rank approximation perspective, and give an upper bound on the training error when this framework is used for incremental learning in some special cases. Finally, extensive experimental results on two real-world datasets clearly validate the effectiveness, efficiency and storage performance of the proposed framework.

PDF Details