Jiajun Fan

CS Ph.D. Student · University of Illinois Urbana-Champaign

jiajunf3@illinois.edu jiajunfan.com Google Scholar Full CV (PDF)

RL Post-Training · Generative Models Multimodal Reasoning LLMs Superhuman Deep RL

Experience

Amazon — AGI Foundations (Audio / Multimodal LLM)

Applied Scientist Intern

Seattle, WA, USAMay 2025 – Dec 2025

Full-stack distributed RL post-training: sole owner of the end-to-end pipeline for multimodal reasoning LLMs — training infrastructure, algorithm and reward design, evaluation and model delivery — and sole first author of the resulting publication (CESAR, ICLR 2026).
Built and scaled a multi-node, multi-GPU GRPO training stack for omni-modal LLMs, and designed a process-reward framework that moves RLVR beyond outcome verification and removes any dependence on human-annotated reasoning traces.

Education

University of Illinois Urbana-Champaign

Ph.D. in Computer Science GPA 4.0/4.0

Urbana, IL, USAAug 2024 – May 2029 (expected)

Research: Autonomous RL post-training for flow/diffusion models & multimodal reasoning LLMs
Service: Reviewer @ ICML 2025–26, ICLR 2025–26, NeurIPS 2024–26, COLM 2026, CVPR 2026, WACV 2027, AAAI 2027, AISTATS 2025

Tsinghua University

M.Eng. in Computer Technology GPA 3.97/4.0 · Top 1.3%

Beijing, ChinaSept 2021 – Jun 2024

Service: Reviewer @ NeurIPS 2022–23, ICML 2023–24, ICLR 2024

Nankai University

B.Eng. in Intelligent Science and Technology Rank 1/83

Tianjin, ChinaSept 2017 – Jun 2021

GPA 93.28/100 (3.9/4.0) · National Scholarship ×2 (Top 1%)

Selected Publications * = first / co-first author

ICML 2026

Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation

J. Sun, P. Jiang, S. Wang, J. Fan, H. Wang, S. Ouyang, M. Zhong, Y. Jiao, C. Huang, X. Xu, P. Han, P. Li, J. Huang, G. Liu, H. Ji, J. Han

+10.3% over strong RAG and reranking baselines

ICML 2026

Batched Contextual Reinforcement

B. Yang, H. Ma, J. Fan, G. Liu

15.8–62.6% fewer tokens at equal or better accuracy

COLM 2026

Procedure-Aware Reinforcement Learning for Tool-Augmented Large Language Models

Q. Zheng*, J. Fan*, C. Cheng, G. Liu

3B model competitive with GPT-5-mini and Claude-Haiku-4.5 on BFCL-V3

TMLR 2026

ProteinZero: Self-Improving Protein Generation via Online Reinforcement Learning

Z. Wang*, J. Fan*, R. Guo, T. Nguyen, H. Ji, G. Liu

Online RL for self-improving protein design, without human-curated data

ICLR 2026

Incentivizing Consistent, Effective and Scalable Reasoning Capability in Audio LLMs via Reasoning Process Rewards

J. Fan*, R. Ren, J. Li, R. Pandey, P.G. Shivakumar, I. Bulyko, A. Gandhe, G. Liu, Y. Gu

SOTA MMAU · Outperforms Gemini 2.5 Pro & GPT-4o Audio

ICLR 2026

SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration

Y. Li, Y. Meng, Z. Sun, K. Ji, C. Tang, J. Fan, X. Ma, S.-T. Xia, Z. Wang, W. Zhu

1.5× lossless speedup (LIBERO) · 2.4× (SimplerEnv)

NeurIPS 2025

Adaptive Divergence Regularized Policy Optimization for Fine-tuning Generative Models

J. Fan*, T. Wei, C. Cheng, Y. Chen, G. Liu

2B SD3 surpasses 4.8B & 12B models

NeurIPS 2025

Variational Supervised Contrastive Learning

Z. Wang, J. Fan, T. Nguyen, H. Ji, G. Liu

SOTA 79.36% Top-1 on ImageNet-1K (ResNet-50)

ICLR 2025

Online Reward-Weighted Fine-Tuning of Flow Matching with Wasserstein Regularization

J. Fan*, S. Shen, C. Cheng, Y. Chen, C. Liang, G. Liu

First online RLHF for flow matching models

2025

Fine-tuning Flow Matching Generative Models with Intermediate Feedback

J. Fan*, C. Cheng, S. Shen, X. Zhou, G. Liu · Under Review

2026

ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models

Y. Li, H. Liu, K. Ji, Y. Meng, J. Fan, Y. Wang, S. Qin, C. Wu, S.-T. Xia, Z. Wang

2.55× / 3.77× speedup · 13.8 → 26.3 Hz on real-world GR00T tasks

TPAMI 2026

PRANCE: Joint Token-Optimization and Structural Channel-Pruning for Adaptive ViT Inference

Y. Li, C. Tang, Y. Meng, J. Fan, Z. Chai, X. Ma, Z. Wang, W. Zhu

ICLR 2023 Oral

Learnable Behavior Control: Breaking Atari Human World Records via Sample-Efficient Behavior Selection

J. Fan*, Y. Zhuang, Y. Liu, J. Hao, B. Wang, J. Zhu, H. Wang, S.-T. Xia

Ranked 5/4176 · 24 Atari world records · 78× data efficiency

ICML 2022

Generalized Data Distribution Iteration

J. Fan*, C. Xiao

Beats Agent57 with 500× less data · 22 world records

Awards & Honors

2024GPA 4.0/4.0, UIUC Ph.D. Program
2024GPA 3.97/4.0, Top 1.3% — Tsinghua University M.Eng.
2021Outstanding Graduates (Top 1%) — Nankai UniversityTop 1%
2021Excellent Graduation Thesis — Nankai University
2021Tang Lixin ScholarshipTop 1%
2020National Scholarship (Rank 1/83)Top 1%
2020Nomination for Zhou Enlai Scholarship
2019National Scholarship (Rank 1/83)Top 1%
20193rd Prize, RoboCup@HOME Education World Final — Sydney, Australia
2019Bronze Medal, ACM/ICPC Asia Regional — Xuzhou
2018National 2nd Prize, Mathematical Contest in ModelingTop 5%