GraphRouter: A Graph-based Router for LLM Selections
Tao Feng*, Yanzhen Shen*, and Jiaxuan You
ICLR, 2025
code

We propose GraphRouter, a novel graph-based router that improves LLM selection for diverse tasks by leveraging an inductive graph framework to optimize performance-cost trade-offs and generalize across various models and scenarios.

GraphEval: A Lightweight Graph-Based LLM Framework for Idea Evaluation
Tao Feng*, Yihang Sun*, and Jiaxuan You (* Equal Contribution)
ICLR, 2025
code

GraphEval is a lightweight, graph-based LLM framework for idea evaluation, offering two methods: GraphEval-LP (training-free) and GraphEval-GNN (minimally trained).

Thought-Retriever: Don't Just Retrieve Raw Data, Retrieve Thoughts
Tao Feng*, Pengrui Han*, Guanyu Lin*, Ge Liu, Jiaxuan You (* Equal Contribution)
ICLR Workshop on How Far Are We From AGI, 2024

We introduce Thought-Retriever a novel model-agnostic algorithm that enables LLMs to effectively utilize external data without being limited by context length.

Paper Copilot: A Self-Evolving and Efficient LLM System for Personalized Academic Assistance
Guanyu Lin*, Tao Feng*, Pengrui Han*, Ge Liu, Jiaxuan You (* Equal Contribution)
System Demonstration Track of Empirical Methods in Natural Language Processing (EMNLP), 2024
Huggingface Live Demo: Link

We propose a light and efficient pipeline that enables both domain and non-domain experts to quickly generate synthetic debiasing data to mitigate specific or general bias in their models with parameter-efficient fine-tuning.

ILRoute: A Graph-based Imitation Learning Method to Unveil Riders' Routing Strategies in Food Delivery Service
Tao Feng, Huan Yan, Huandong Wang, Wenzhen Huang, Yuyang Han, Hongsen Liao, Jinghua Hao, Yong Li
KDD, 2023

We introduce ILRoute, a graph-based imitation learning approach for Pick-up and Delivery Route Prediction (PDRP) that effectively models complex, multi-source feature interactions and personalized rider preferences, demonstrating superior performance through both offline and online evaluations.

Precise mobility intervention for epidemic control using unobservable information via deep reinforcement learning
Tao Feng, Tong Xia, Xiaochen Fan, Huandong Wang, Zefang Zong, Yong Li
KDD, 2022

We introduce Vehicle, a variational hierarchical reinforcement learning framework that reconstructs unobservable epidemic risks and overcomes delayed rewards to enable precise individual mobility interventions for controlling COVID-19 spread.

GAT-MF: Graph Attention Mean Field for Very Large Scale Multi-Agent Reinforcement Learning
Qianyue Hao, Wenzhen Huang, Tao Feng, Jian Yuan, Yong Li
KDD, 2023
code

We introduce Graph Attention Mean Field (GAT-MF), a novel multi-agent reinforcement learning approach that reduces computational complexity and precisely models dynamic, diverse agent interactions, achieving significant performance and efficiency gains in experiments with over 3000 agents.

Causal Discovery and Inference towards Urban Elements and Associated Factors
Tao Feng, Yunke Zhang, Xiaochen Fan, Huandong Wang, Yong Li
arxiv, 2025

We propose a novel urban causal computing framework that uses reinforcement learning to discover causal relationships among urban factors and reduce confounding effects, thereby improving urban mobility prediction.

Causality Enhanced Origin-Destination Flow Prediction in Data-Scarce Cities
Tao Feng, Yunke Zhang, Xiaochen Fan, Huandong Wang, Yong Li
arxiv, 2025

We propose CE-OFP, which is a unified framework that transfers urban knowledge from data-rich cities to data-scarce ones by leveraging reinforcement learning for causal discovery, variational auto-encoders for feature reconstruction, and graph attention networks for knowledge distillation, resulting in up to an 11% reduction in OD flow prediction RMSE.

Deep Reinforcement Learning for Modelling Protein Complexes
Tao Feng, Ziqi Gao, Jiaxuan You, Chenyi Zi, Yan Zhou, Chen Zhang, Jia Li
ICLR, 2024

We propose GAPN, a generative adversarial policy network that efficiently models multi-chain protein complexes by formulating the assembly as a graph search and leveraging policy gradient with adversarial rewards to achieve significant accuracy and efficiency improvements.

Molecular De Novo Design through Transformer-based Reinforcement Learning
Tao Feng, Pengcheng Xu, Tianfan Fu, Jimeng Sun
arxiv, 2024

We introduce a Transformer-based generative model for molecular de novo design that outperforms traditional RNN methods by effectively capturing long-term dependencies to generate active compounds with desired properties.

How Far Are We From AGI: Are LLMs All We Need?
Tao Feng*, Chuanyang Jin*, Jingyu Liu*, Kunlun Zhu*, Haoqin Tu, Zirui Cheng, Guanyu Lin, Jiaxuan You
TMLR, 2024

We provide a comprehensive exploration of AGI by defining its required capabilities, discussing alignment challenges, and outlining a strategic roadmap for responsible and impactful development across various domains.

Deep Reinforcement Learning for Demand Driven Services in Logistics and Transportation Systems: A Survey
Zefang Zong, Tao Feng, Jingwei Wang, Tong Xia, Yong Li
TKDD, 2024

We review the challenges and deep reinforcement learning solutions for optimizing dispatching and routing in urban demand-driven services.