Yilan Chen

Publications

LLM Post-training & Agents

• Reward-Aligned Distillation. [project page]
  Yilan Chen, Zian Su, Manish Prajapat, Dingmin Wang, Rein Houthooft, Rui Song, Vijay Lingam.
  Under review; also accepted at NeurIPS 2026 PretoPost Workshop.

• PINT: Privileged Information Internalization for Agentic RL.
  Zian Su, Yilan Chen, Manish Prajapat, Rein Houthooft, Rui Song, Vijay Lingam.
  Under review; also accepted at NeurIPS 2026 PretoPost Workshop.

• Token Utility for On-Policy Distillation via Local Trust-Region Optimization.
  Narayanan Sadagopan, Yilan Chen, Vijay Lingam.
  Under review; also accepted at NeurIPS 2026 PretoPost Workshop.

• BridgeSim: Unveiling the OL-CL Gap in End-to-End Autonomous Driving.
  Seth Z. Zhao¹, Luobin Wang¹, Hongwei Ruan, Yuxin Bao, Yilan Chen, et al.
  Conference on Robot Learning (CoRL 2026).

• Cross-Task Linearity Emerges in the Pretraining-Finetuning Paradigm. [code]
  Zhanpeng Zhou¹, Zijun Chen¹, Yilan Chen, Bo Zhang, Junchi Yan.
  Forty-first International Conference on Machine Learning (ICML 2024).

Machine Learning Theory

• Generalization Bound of Gradient Flow through Training Trajectory and Data-dependent Kernel.
  Yilan Chen, Zhichao Wang, Wei Huang, Andi Han, Taiji Suzuki, Arya Mazumdar.
  Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025).

• How Does Label Noise Gradient Descent Improve Generalization in the Low SNR Regime？
  Wei Huang¹, Andi Han¹, Yujin Song, Yilan Chen, Denny Wu, Difan Zou, Taiji Suzuki.
  Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025).

• Provable and Efficient Dataset Distillation for Kernel Ridge Regression. [code] [project page]
  Yilan Chen, Wei Huang, Tsui-Wei Weng.
  Thirty-eighth Conference on Neural Information Processing Systems (NeurIPS 2024).

• Analyzing Generalization of Neural Networks through Loss Path Kernels. [code] [slides] [poster] [video]
  Yilan Chen, Wei Huang, Hao Wang, Charlotte Loh, Akash Srivastava, Lam M. Nguyen, Tsui-Wei Weng.
  Thirty-seventh Conference on Neural Information Processing Systems (NeurIPS 2023).

• Analyzing Deep PAC-Bayesian Learning with Neural Tangent Kernel: Convergence, Analytic Generalization Bound, and Efficient Hyperparameter Selection.
  Wei Huang¹, Chunrui Liu¹, Yilan Chen, Richard Yi Da Xu, Miao Zhang, Tsui-Wei Weng.
  Transactions on Machine Learning Research (TMLR 2023).

• On the Equivalence between Neural Network and Support Vector Machine. [code][slides][poster][video]
  Yilan Chen, Wei Huang, Lam M. Nguyen, Tsui-Wei Weng.
  Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS 2021).

Interpretable Machine Learning

• The Importance of Prompt Tuning for Automated Neuron Explanations. [code]
  Justin Lee¹, Tuomas Oikarinen¹, Arjun Chatha, Keng-Chi Chang, Yilan Chen, Tsui-Wei Weng.
  NeurIPS 2023 Workshop on Attributing Model Behavior at Scale.

• Quantifying the Knowledge in a DNN to Explain Knowledge Distillation for Classification.
  Quanshi Zhang¹, Xu Cheng¹, Yilan Chen, Zhefan Rao.
  IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI 2023).

• Explaining Knowledge Distillation by Quantifying the Knowledge.
  Xu Cheng, Zhefan Rao², Yilan Chen², Quanshi Zhang.
  IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2020).