Tianxin Wei

Education

University of Illinois Urbana-Champaign

Ph.D. · 2021 ~ 2026 (expected)

Advisor: Prof. Jingrui He

University of Science and Technology of China

B.S. · 2016 ~ 2020

School of the Gifted Young · Advisor: Prof. Xiangnan He

Work Experience

Google DeepMind GenAI

May ~ Oct 2025 · Mountain View, CA & Urbana, IL

Research Intern · LLM Agent Memory and Personalization

Meta Ranking AI

Jan ~ May 2025 · Remote

Research Intern · Generative Recommender Systems with LLMs

Amazon

2023 ~ 2026 · Palo Alto, CA & Urbana, IL

Amazon Search · May ~ Dec 2023
Multi-modal LLMs for Personalization → ICLR 2024, WWW 2024, ICML 2024, KDD 2025

Amazon Shopping FM · May ~ Oct 2024
Multi-modal Watermarking for Generative Models → TMLR

Amazon Ads FM · Feb ~ May 2026
LLM Agent Continual Learning and Self-improving

News

Publications

* equal contribution   ^ mentorship

• Selected Preprints
• Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agents
  Preprint
  Tianxin Wei, Zhan Shi, Minhua Lin, Bing He, Zewen Liu, Yisi Sang, Yuanchen Bei, Xuying Ning, Jiaru Zou, Ting-Wei Li, Xiao Lin, Yanjun Zhao, Chi Wang, Benoit Dumoulin, Dakuo Wang, Jingrui He, Hanqing Lu
  Key Insight: Compiles contextual experience into reusable harness skills for self-evolving agents.
  Paper Code
  AgentLLM
• Code as Agent Harness
  Preprint
  Xuying Ning*, Katherine Tieu*, Dongqi Fu*, Tianxin Wei*, Zihao Li*, Yuanchen Bei*, Jiaru Zou, Mengting Ai, Zhining Liu, Ting-Wei Li, Lingjie Chen, Yanjun Zhao, Ke Yang, Bingxuan Li, Cheng Qian, Gaotang Li, Xiao Lin, Zhichen Zeng, Ruizhong Qiu, Sirui Chen, Yifan Sun, Xiyuan Yang, Ruida Wang, Rui Pan, Chenyuan Yang, Dylan Zhang, Liri Fang, Zikun Cui, Yang Cao, Pan Chen, Dorothy Sun, Ren Chen, Mahesh Srinivasan, Nipun Mathur, Yinglong Xia, Hong Li, Hong Yan, Pan Lu, Lingming Zhang, Tong Zhang, Hanghang Tong, Jingrui He
  Key Insight: Treats executable code as a harness for agent deployment, evaluation, and improvement.
  Paper Project Code
  AgentLLM
• ReContext: Recursive Evidence Replay as LLM Harness for Long-Context Reasoning
  Preprint
  Yanjun Zhao*, Ruizhong Qiu*, Tianxin Wei*, Yuanchen Bei, Zhining Liu, Lingjie Chen, Ismini Lourentzou, Hanghang Tong, Jingrui He
  Key Insight: Recursively replays evidence to build an LLM harness for long-context reasoning.
  Paper Code
  AgentLLMRetrieval
• EvoSelect: Data-Efficient LLM Evolution for Targeted Task Adaptation
  Preprint
  Ting-Wei Li, Sirui Chen, Jiaru Zou, Yingbing Huang, Tianxin Wei, Jingrui He, Hanghang Tong
  Key Insight: Data-efficient LLM evolution for adapting models to targeted tasks.
  Paper Code
  AgentLLMEfficiency
• Adaptive Auto-Harness: Sustained Self-Improvement for Agentic System Deployment on Open-Ended Task Streams
  Preprint
  Zewen Liu, Zhan Shi, Yisi Sang, Bing He, Minhua Lin, Tianxin Wei, Dakuo Wang, Benoit Dumoulin, Wei Jin, Hanqing Lu
  Key Insight: Sustained self-improvement for agentic systems deployed on open-ended task streams.
  Paper Code
  AgentLLM
• Harness Updating Is Not Harness Benefit: Disentangling Evolution Capabilities in Self-Evolving LLM Agents
  Preprint
  Minhua Lin, Juncheng Wu, Zijun Wang, Zhan Shi, Yisi Sang, Bing He, Zewen Liu, Tianxin Wei, Zongyu Wu, Zhiwei Zhang, Dakuo Wang, Xiang Zhang, Benoit Dumoulin, Cihang Xie, Yuyin Zhou
  Key Insight: Separates harness updates from true harness benefit in self-evolving LLM agents.
  Paper Code
  AgentLLM
• RRCM: Ranking-Driven Retrieval over Collaborative and Meta Memories for LLM Recommendation
  Preprint
  Shijun Li, Wooseong Yang, Yu Wang, Tianxin Wei, Joydeep Ghosh
  Key Insight: Ranking-driven retrieval over collaborative and metadata memories for LLM recommendation.
  Paper Code
  LLMPersonalizationRetrieval
• Heterogeneous Scientific Foundation Model Collaboration
  Preprint
  Zihao Li, Jiaru Zou, Feihao Fang, Xuying Ning, Mengting Ai, Tianxin Wei, Sirui Chen, Xiyuan Yang, Jingrui He
  Key Insight: Heterogeneous agent collaboration between language models and specialized scientific foundation models.
  Paper Code
  LLMMulti-AgentApplication
• Agentic Reasoning for Large Language Models: A Survey
  Survey 2026
  Tianxin Wei, Ting-Wei Li, Zhining Liu, Xuying Ning, Ze Yang, Jiaru Zou, Zhichen Zeng, Ruizhong Qiu, Xiao Lin, Dongqi Fu, Zihao Li, Mengting Ai, Duo Zhou, Wenxuan Bao, Yunzhe Li, Gaotang Li, Cheng Qian, Yu Wang, Xiangru Tang, Yin Xiao, Liri Fang, Hui Liu, Xianfeng Tang, Yuji Zhang, Chi Wang, Jiaxuan You, Heng Ji, Hanghang Tong, Jingrui He
  Key Insight: Unified taxonomy of how LLM agents plan, use tools, and reflect to reason.
  Paper Project Huggingface
  AgentSurvey
• Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory
  arXiv 2025
  Tianxin Wei, Noveen Sachdeva, Benjamin Coleman, Zhankui He, Yuarchen Bei, Xuying Ning, Mengting Ai, Yunzhe Li, Jingrui He, Ed H. Chi, Chi Wang, Shuo Chen, Fernando Pereira, Wang-Cheng Kang, Derek Zhiyuan Cheng
  Key Insight: LLM agents that autonomously build and refine memory from experience at test time.
  Paper Code
  AgentPersonalization
• CoFiRec: Coarse-to-Fine Tokenization for Generative Recommendation
  arXiv 2025
  Tianxin Wei*, Xuying Ning*, Xuxing Chen, Ruizhong Qiu, Yupeng Hou, Yan Xie, Shuang Yang, Zhigang Hua, Jingrui He
  Key Insight: Hierarchical tokenization bridging semantics and collaboration for generative recommendation.
  Paper Code
  LLMPersonalization
• Rethinking Memory Mechanisms of Foundation Agents in the Second Half: A Survey
  arXiv 2026
  Joint work across 27 institutions.
  Key Insight: Comprehensive categorization of agent memory mechanisms and key design gaps.
  Paper Project
  AgentSurveyPersonalization
• Conference
• AdaFuse: Adaptive Ensemble Decoding with Test-Time Scaling for LLMs
  ACL 2026 Main
  Chengming Cui*, Tianxin Wei*^, Ziyi Chen, Ruizhong Qiu, Zhichen Zeng, Zhining Liu, Xuying Ning, Duo Zhou, Jingrui He
  Key Insight: Fuse multiple LLM outputs adaptively at decode time via test-time scaling.
  Paper Code
  LLMEfficiencyMulti-Agent
• PaperMind: Benchmarking Agentic Reasoning and Critique over Scientific Papers in Multimodal LLMs
  ACL 2026 Findings
  Yanjun Zhao*, Tianxin Wei*^, Jiaru Zou, Xuying Ning, Yuanchen Bei, Lingjie Chen, Simmi Rana, Wendy H. Yang, Hanghang Tong, Jingrui He
  Key Insight: Benchmark for evaluating multimodal LLM reasoning and critique over scientific papers.
  Paper Code
  AgentMultimodal
• Mem-Gallery: Benchmarking Multimodal Long-Term Conversational Memory for MLLM Agents
  ACL 2026 Main
  Yuanchen Bei, Tianxin Wei^, Xuying Ning, Yanjun Zhao, Zhining Liu, Xiao Lin, Yada Zhu, Hendrik Hamann, Jingrui He, Hanghang Tong
  Key Insight: Benchmark exposing MLLM agents' failure in long-term multimodal conversational memory.
  Paper Code
  AgentMultimodalPersonalization
• MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains
  ICLR 2026 Oral (Top 1%). Prelim version at NeurIPS 2025 Workshop.
  Xuying Ning*, Dongqi Fu*, Tianxin Wei*, Mengting Ai, Jiaru Zou, Ting-Wei Li, Hanghang Tong, Yada Zhu, Hendrik Hamann, Jingrui He
  Key Insight: Structured long reasoning chains for multimodal agentic search.
  Paper Code
  AgentMultimodalRobustness
• Seeing but Not Believing: Probing the Disconnect Between Visual Attention and Answer Correctness in VLMs
  ICLR 2026
  Zhining Liu, Ziyi Chen, Hui Liu, Chen Luo, Xianfeng Tang, Suhang Wang, Joy Zeng, Zhenwei Dai, Zhan Shi, Tianxin Wei, Benoit Dumoulin, Hanghang Tong
  Key Insight: Correct visual attention does not guarantee correct VLM reasoning.
  Paper Code
  MultimodalRobustness
• Harnessing Consistency for Robust Test-Time LLM Ensemble
  EACL 2026 Findings
  Zhichen Zeng, Qi Yu, Xiao Lin, Ruizhong Qiu, Xuying Ning, Tianxin Wei, Yuchen Yan, Jingrui He, Hanghang Tong
  Key Insight: Consistency-driven test-time LLM ensemble for robust multi-model agreement.
  Paper Code
  LLMMulti-Agent
• Panda: Test-Time Adaptation with Negative Data Augmentation
  AAAI 2026
  Ruxi Deng, Wenxuan Bao, Tianxin Wei, Jingrui He
  Key Insight: Negative data augmentation at test time for robust domain adaptation.
  Paper Code
  MultimodalTest-timePersonalizationRobustness
• CLIMB: Class-imbalanced Learning Benchmark on Tabular Data
  NeurIPS 2025
  Zhining Liu, Zihao Li, Ze Yang, Tianxin Wei, Jian Kang, Yada Zhu, Hendrik Hamann, Jingrui He, Hanghang Tong
  Key Insight: Unified benchmark for class-imbalanced tabular learning.
  Paper Code
  Robustness
• Latte: Collaborative Test-Time Adaptation of Vision-Language Models in Federated Learning
  ICCV 2025
  Wenxuan Bao, Ruxi Deng, Ruizhong Qiu, Tianxin Wei, Hanghang Tong, Jingrui He
  Key Insight: Collaborative test-time VLM adaptation across federated clients without sharing data.
  Paper Code
  MultimodalTest-timePersonalizationRobustness
• Learning to Instruct: Fine-Tuning a Task-Aware Instruction Optimizer for Black-Box LLMs
  EMNLP 2025 Findings
  Yunzhe Qi, Jinjin Tian, Tianci Liu, Ruirui Li, Tianxin Wei, Hui Liu, Xianfeng Tang, Monica Xiao Cheng, Jingrui He
  Key Insight: Auto-rewrite prompts for black-box LLMs via a learned instruction optimizer.
  Paper Code
  LLM
• SelfElicit: Your Language Model Secretly Knows Where is the Relevant Evidence
  ACL 2025 Main
  Zhining Liu, Rana Ali Amjad, Ravinarayana Adkathimar, Tianxin Wei, Hanghang Tong
  Key Insight: LLMs implicitly know where evidence is; self-eliciting beats external retrievers.
  Paper Code
  LLMRetrieval
• i²VAE: Interest Information Augmentation with Variational Regularizers for Cross-Domain Sequential Recommendation
  UAI 2025
  Xuying Ning, Wujiang Xu, Tianxin Wei, Xiaolei Liu
  Key Insight: Variational interest augmentation for cross-domain sequential recommendation.
  Paper Code
  PersonalizationRobustness
• Graph4MM: Weaving Multimodal Learning with Structural Information
  ICML 2025
  Xuying Ning, Dongqi Fu, Tianxin Wei, Wujiang Xu, Jingrui He
  Key Insight: Graph structure as inductive bias for richer multimodal representations.
  Paper Code
  LLMMultimodal
• Breaking Silos: Adaptive Model Fusion Unlocks Better Time Series Forecasting
  ICML 2025
  Zhining Liu, Ze Yang, Xiao Lin, Ruizhong Qiu, Tianxin Wei, Yada Zhu, Hendrik Hamann, Jingrui He, Hanghang Tong
  Key Insight: Adaptive fusion of heterogeneous time series forecasters breaks model silos.
  Paper Code
  ApplicationEfficiency
• Connecting Domains and Contrasting Samples: A Ladder for Domain Generalization
  KDD 2025 (Full Research, AR: 19%)
  Tianxin Wei*, Yifan Chen*, Wenxuan Bao, Jingrui He
  Key Insight: Progressive domain connection via contrastive alignment for domain generalization.
  Paper Code
  Robustness
• MLP Fusion: Towards Efficient Fine-tuning of Dense and Mixture-of-Experts Language Models
  v1 accepted to ICML 2023. Submitted to IEEE for possible publication.
  Mengting Ai*, Tianxin Wei*, Yifan Chen*, Zeming Guo, Jingrui He
  Key Insight: Lightweight MLP fusion for parameter-efficient fine-tuning of dense and MoE LLMs.
  Paper Code
  LLMEfficiency
• ResMoE: Space-efficient Compression of Mixture of Experts LLMs via Residual Restoration
  KDD 2025 (Full Research, AR: 19%)
  Mengting Ai*, Tianxin Wei*^, Yifan Chen*, Zhichen Zeng, Ritchie Zhao, Girish Varatkar, Bita Darvish Rouhani, Xianfeng Tang, Hanghang Tong, Jingrui He
  Key Insight: Residual restoration compresses MoE LLMs while preserving expert specialization.
  Paper Code
  LLMEfficiency
• Meta Clustering of Neural Bandits
  KDD 2024 (Full Research, AR: 20%)
  Yikun Ban, Yunzhe Qi, Tianxin Wei, Lihui Liu, Jingrui He
  Key Insight: Meta-learning discovers bandit arm clusters for efficient shared exploration.
  Paper Code
  Efficiency
• Graph Mixup on Approximate Gromov-Wasserstein Geodesics
  ICML 2024 (Full Research, AR: 27.5%)
  Zhichen Zeng, Ruizhong Qiu, Zhe Xu, Zhining Liu, Yuchen Yan, Tianxin Wei, Lei Ying, Jingrui He, Hanghang Tong
  Key Insight: Graph mixup along Gromov-Wasserstein geodesics for topology-aware augmentation.
  Paper Code
  Robustness
• Language Models as Semantic Indexers
  ICML 2024 (Full Research, AR: 27.5%)
  Bowen Jin, Hansi Zeng, Guoyin Wang, Xiusi Chen, Tianxin Wei, Ruirui Li, Zhengyang Wang, Zheng Li, Yang Li, Hanqing Lu, Suhang Wang, Jiawei Han, Xianfeng Tang
  Key Insight: Language models as end-to-end semantic indexers replacing traditional retrieval.
  Paper Code
  LLMRetrieval
• Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond
  ICLR 2024 (Full Research, AR: 31%)
  Tianxin Wei, Bowen Jin, Ruirui Li, Hansi Zeng, Zhengyang Wang, Jianhui Sun, Qingyu Yin, Hanqing Lu, Suhang Wang, Jingrui He, Xianfeng Tang
  Key Insight: One vision-language model for unified multi-modal generative recommendation.
  Paper Code
  LLMMultimodalPersonalization
• Scalable and Effective Generative Information Retrieval
  WWW 2024 (Oral) (Full Research, AR: 20.2%)
  Hansi Zeng, Chen Luo, Bowen Jin, Sheikh Muhammad Sarwar, Tianxin Wei, Hamed Zamani
  Key Insight: Scalable generative retrieval via residual-quantized prefix language modeling.
  Paper Code
  LLMRetrieval
• TAU: Trajectory Data Augmentation with Uncertainty for Next POI Recommendation
  AAAI 2024 (Full Research, AR: 24%)
  Zhuang Zhuang, Tianxin Wei^, Lingbo Liu, Heng Qi, Yanming Shen, Baocai Yin
  Key Insight: Uncertainty-aware trajectory augmentation for next POI recommendation.
  Paper Code
  RobustnessPersonalization
• Meta-Learning with Neural Bandit Scheduler
  NeurIPS 2023 (Full Research, AR: 26.1%)
  Yunzhe Qi, Yikun Ban, Tianxin Wei, Jiaru Zou, Huaxiu Yao, Jingrui He
  Key Insight: Neural bandit scheduling for adaptive task allocation in meta-learning.
  Paper Code
  Efficiency
• Adaptive Test-Time Personalization for Federated Learning
  NeurIPS 2023 (Full Research, AR: 26.1%)
  Wenxuan Bao*, Tianxin Wei*, Haohan Wang, Jingrui He
  Key Insight: Test-time personalization bridging global federated models to local distributions.
  Paper Code
  RobustnessTest-timePersonalization
• NTK-approximating MLP Fusion for Efficient Language Model Fine-tuning
  ICML 2023 (Full Research, AR: 27.9%)
  Tianxin Wei*^, Zeming Guo*, Yifan Chen*, Jingrui He
  Key Insight: NTK-principled MLP fusion for theoretically grounded efficient LLM fine-tuning.
  Paper Code
  LLMEfficiency
• Augmentations in Hypergraph Contrastive Learning: Fabricated and Generative
  NeurIPS 2022 (Full Research, AR: 25.6%)
  Tianxin Wei*, Yuning You*, Tianlong Chen, Yang Shen, Jingrui He, Zhangyang Wang
  Key Insight: Fabricated and generative augmentations for hypergraph contrastive learning.
  Paper Code Appendix
  Robustness
• Comprehensive Fair Meta-learned Recommender System
  KDD 2022 (Full Research, AR: 15.0%)
  Tianxin Wei, Jingrui He
  Key Insight: Meta-learned recommendation with built-in fairness across user demographics.
  Paper Code
  FairnessPersonalization
• Model-Agnostic Counterfactual Reasoning for Eliminating Popularity Bias in Recommender System
  KDD 2021 (Full Research, AR: 15.4%)
  Tianxin Wei, Fuli Feng, Jiawei Chen, Ziwei Wu, Jinfeng Yi, Xiangnan He
  Key Insight: Counterfactual reasoning removes popularity bias from any recommender.
  Paper Code
  BiasPersonalization
• Causal Intervention for Leveraging Popularity Bias in Recommendation
  SIGIR 2021 Best Paper Honorable Mention (1 out of All) (Full Research, AR: 21%)
  Yang Zhang, Fuli Feng, Xiangnan He, Tianxin Wei, Chonggang Song, Guohui Ling, Yongdong Zhang
  Key Insight: Causal intervention that leverages popularity signals while suppressing recommendation bias.
  Paper Code
  BiasPersonalization
• Fast Adaptation for Cold-start Collaborative Filtering with Meta-learning
  ICDM 2020 (Full Research, AR: 9.8%)
  Tianxin Wei, Ziwei Wu, Ruirui Li, Ziniu Hu, Fuli Feng, Xiangnan He, Yizhou Sun, Wei Wang
  Key Insight: Meta-learning for rapid cold-start adaptation from few user interactions.
  Paper Code
  TransferabilityPersonalization
• Unpaired Multimodal Neural Machine Translation via Reinforcement Learning
  DASFAA 2021 (Full Research, AR: 20%)
  Yijun Wang*, Tianxin Wei*, Qi Liu, Enhong Chen
  Key Insight: RL-based alignment of unpaired images and text for multimodal translation.
  Paper
  Multimodal
• AR-Stock: Deep Augmented Relational Stock Prediction
  AAAI 2021 Workshop on Knowledge Discovery from Unstructured Data (Oral)
  Tianxin Wei, Yuning You, Tianlong Chen
  Key Insight: Augmented relational graph learning capturing inter-stock market dynamics.
  Paper Code
  Application
• Load more publications
• Journal
• DiffKGW: Stealthy and Robust Diffusion Model Watermarking
  TMLR
  Tianxin Wei, Ruizhong Qiu, Yifan Chen, Yunzhe Qi, Jiacheng Lin, Wenxuan Bao, Wenju Xu, Sreyashi Nag, Ruirui Li, Hanqing Lu, Zhengyang Wang, Chen Luo, Hui Liu, Suhang Wang, Jingrui He, Qi He, Xianfeng Tang
  Key Insight: Stealthy robust watermarking for diffusion models via generation time intervention analogous to LLMs.
  Paper Code
  DiffusionRobustness
• Pave Your Own Path: Graph Gradual Domain Adaptation on Fused Gromov-Wasserstein Geodesics
  TMLR
  Zhichen Zeng, Ruizhong Qiu, Wenxuan Bao, Tianxin Wei, Xiao Lin, Yuchen Yan, Tarek F. Abdelzaher, Jiawei Han, Hanghang Tong
  Key Insight: Gradual graph domain adaptation along fused Gromov-Wasserstein geodesics.
  Paper Code
  Robustness
• Flow Matching Meets Biology and Life Science: A Survey
  npj Artificial Intelligence
  Zihao Li, Zhichen Zeng, Xiao Lin, Feihao Fang, Yanru Qu, Zhe Xu, Zhining Liu, Xuying Ning, Tianxin Wei, Ge Liu, Hanghang Tong, Jingrui He
  Key Insight: Survey connecting flow matching to molecular and biological modeling.
  Paper Code
  SurveyApplication
• Graph Contrastive Learning: An Odyssey towards Generalizable, Scalable and Principled Representation Learning on Graphs
  IEEE Data Engineering Bulletin
  Yan Han, Yuning You, Wenqing Zheng, Scott Hoang, Tianxin Wei, Majdi Hassan, Tianlong Chen, Ying Ding, Yang Shen, Zhangyang Wang
  Key Insight: Roadmap from augmentation design to principled graph contrastive learning.
  Paper
  Survey

Awards & Service

Gallery