[Jingxuan]
Jingxuan Fan
MLRE, Scale AI
Email: jingxuanf0214@gmail.com
Office: 1 World Trade Center, New York, NY
Biography
My name is Jingxuan. I am a machine learning research engineer at Scale AI, where I work on agentic post-training. I received my Ph.D. in neuroscience from HMS advised by Prof. Rachel Wilson, along with a concurrent M.S. in applied math at Harvard SEAS advised by Prof. Michael Brenner. My research interests span GenAI/LLM and NeuroAI, and you can reach me via email jingxuanf0214[AT]gmail[DOT]com.
Research
(* indicates equal first contribution, + indicates equal second contribution, highlight indicates representative papers)
@misc{hoy2026asynchronousnearlyfreeevolution,
  title={Asynchronous Is Nearly Free for Evolution Strategies on Long-Horizon Agentic Tasks}, 
  author={William Hoy and Jingxuan Fan and Nurcin Celik and Xu Pan},
  year={2026},
  eprint={2610.04196},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2610.04196}, 
}
William Hoy, Jingxuan Fan, Nurcin Celik, Xu Pan,
@misc{hoy2026continuouscontextmanagement,
  title={Continuous Context Management}, 
  author={William Hoy and Jingxuan Fan and Nurcin Celik and Xu Pan},
  year={2026},
  eprint={2609.35540},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2609.35540}, 
}
Veronica Chatrath*, Bryan Zhu*, Jingxuan Fan*, George Pu, Soham Dinesh Tiwari, Soham Dan, et al.,
@misc{chatrath2026readynotreliableenterprise,
  title={READY or Not: Reliable Enterprise Agent Deployment}, 
  author={Veronica Chatrath and Bryan Zhu and Jingxuan Fan and George Pu and Soham Dinesh Tiwari and Soham Dan and Ryan Young and Yuan Li and Yuang Yao and Apaar Shanker and Minglai Yang and Daniel Yue Zhang and Yunzhong He and Ying Liu and Chenguang Wang and Zhijun Yin and Yuan Xue},
  year={2026},
  eprint={2609.02095},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2609.02095}, 
}
Veronica Chatrath*, Bryan Zhu*, George Pu, Jingxuan Fan, Apaar Shanker, Varun Ursekar, et al.,
@misc{chatrath2026clinicarebenchclinicalcalibratedaudit,
  title={CliniCARE-Bench: Clinical Calibrated Audit of Medical Reasoning in EHR}, 
  author={Veronica Chatrath and Bryan Zhu and George Pu and Jingxuan Fan and Apaar Shanker and Varun Ursekar and Anahita Sharma and Jason Qin and Keqi Han and Soham Dinesh Tiwari and Soham Dan and Vijay Kalmath and Yuan Li and Daniel Yue Zhang and Chenguang Wang and Zainab Doctor and Zhijun Yin and Nigam H. Shah and Yuan Xue},
  year={2026},
  eprint={2608.07796},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2608.07796}, 
}
Alexander S. Bates, Jasper S. Phelps, Minsu Kim, et al. (incl. Jingxuan Fan), Mala Murthy, Jan Drugowitsch, Rachel I. Wilson, Wei-Chung Allen Lee,
@article{bates2026distributed,
  title={Distributed control circuits across a brain-and-cord connectome},
  author={Bates, Alexander S. and Phelps, Jasper S. and Kim, Minsu and Yang, Helen H. and Matsliah, Arie and Ajabi, Zaki and others and Murthy, Mala and Drugowitsch, Jan and Wilson, Rachel I. and Lee, Wei-Chung Allen},
  journal={Nature},
  volume={656},
  number={8129},
  pages={957--970},
  year={2026},
  publisher={Nature Publishing Group},
  doi={10.1038/s41586-026-10735-w},
}

@misc{fan2026scalingrewardmodelinghuman,
  title={Scaling Reward Modeling without Human Supervision}, 
  author={Jingxuan Fan and Yueying Li and Zhenting Qi and Dinghuai Zhang and Kianté Brantley and Sham M. Kakade and Hanlin Zhang},
  year={2026},
  eprint={2603.02225},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2603.02225}, 
}

@misc{wang2026circuitmechanismsspatialrelation,
  title={Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers}, 
  author={Binxu Wang and Jingxuan Fan and Xu Pan},
  year={2026},
  eprint={2601.06338},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2601.06338}, 
}

@misc{pan2025closingdataefficiencygapautoregressive,
  title={Closing the Data-Efficiency Gap Between Autoregressive and Masked Diffusion LLMs}, 
  author={Xu Pan and Ely Hahami and Jingxuan Fan and Ziqian Xie and Haim Sompolinsky},
  year={2025},
  eprint={2510.09885},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2510.09885}, 
}

@misc{pan2025userassistantbiasllms,
  title={User-Assistant Bias in LLMs}, 
  author={Xu Pan and Jingxuan Fan and Zidi Xiong and Ely Hahami and Jorin Overwiening and Ziqian Xie},
  year={2025},
  eprint={2508.15815},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2508.15815}, 
}

@misc{li2025multiheadrewardaggregationguided,
  title={Multi-head Reward Aggregation Guided by Entropy}, 
  author={Xiaomin Li and Xupeng Chen and Jingxuan Fan and Eric Hanchen Jiang and Mingye Gao},
  year={2025},
  eprint={2503.20995},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2503.20995}, 
}

@misc{li2025dataadaptivesafetyrulestraining,
  title={Data-adaptive Safety Rules for Training Reward Models}, 
  author={Xiaomin Li and Mingye Gao and Zhiwei Zhang and Jingxuan Fan and Weiyu Li},
  year={2025},
  eprint={2501.15453},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2501.15453}, 
}

@misc{fan2024hardmathbenchmarkdatasetchallenging,
    title={HARDMath: A Benchmark Dataset for Challenging Problems in Applied Mathematics}, 
    author={Jingxuan Fan and Sarah Martinson and Erik Y. Wang and Kaylie Hausknecht and Jonah Brenner and Danxian Liu and Nianli Peng and Corey Wang and Michael P. Brenner},
    year={2024},
    eprint={2410.09988},
    archivePrefix={arXiv},
    primaryClass={cs.LG},
    url={https://arxiv.org/abs/2410.09988}, 
}

@article{kwon2021amygdala,
    title={An amygdala circuit that suppresses social engagement},
    author={Kwon, Jeong-tae and Ryu, Changhyeon and Lee, Hyeseung and Sheffield, Alec and Fan, Jingxuan and Bigler, Shivani and Sullivan, Heather A and Choe, Han Kyung and Wickersham, Ian and Heiman, Myriam and Choi, Gloria B},
    journal={Nature},
    volume={593},
    number={7858},
    pages={114--118},
    year={2021},
    publisher={Nature Publishing Group},
    doi={10.1038/s41586-021-03413-6},
}

@article{wei2021single,
    title={Single-nanometer iron oxide nanoparticles as tissue-permeable {MRI} contrast agents},
    author={Wei, He and Wi{\'s}niowska, Agata and Fan, Jingxuan and Harvey, Peter and Li, Yuanyuan and Wu, Victoria and Hansen, Eric C and Zhang, Juanye and Kaul, Michael G and Frey, Abigail M and Adam, Gerhard and Frenkel, Anatoly I and Bawendi, Moungi G and Jasanoff, Alan},
    journal={Proceedings of the National Academy of Sciences},
    volume={118},
    number={42},
    pages={e2102340118},
    year={2021},
    publisher={National Academy of Sciences},
    doi={10.1073/pnas.2102340118},
}
News
[09/30/26] Our work on "Continuous Context Management" was accepted at the Transitioning from Pre-Training to Post-Training Workshop at NeurIPS 2026.
[09/24/26] Our work on "Scaling Reward Modeling without Human Supervision" was accepted at NeurIPS 2026 (main track).
[05/28/26] I graduated from Harvard with a PhD in Neuroscience and started at Scale AI working on agentic post-training.
[04/07/26] Our work on "User-Assistant Bias in LLMs" was accepted at ACL 2026.
[03/02/26] Our work on "Scaling Reward Modeling without Human Supervision" was accepted at ICLR SPOT 2026.
[02/20/26] Our work on "Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers" was accepted at CVPR 2026.
[10/05/25] Our work on "User-Assistant Bias in LLMs" was accepted at NeurIPS 2025 Multi-Turn Interaction Workshop.
[09/02/25] Our work on "Measuring and optimizing evidence preference tradeoff in LLM personalized recommendation" was accepted at NYRL.
[08/12/25] Our work on "The attention mechanism underlying relational object generation in text-to-image diffusion transformers" was accepted at NEMI.
[07/29/25] I joined AISST membership.
[06/16/25] I started my research internship at Amazon (New York) working on Ads/Recommendation GenAI.
[05/01/25] Our work on Data-adaptive Safety Rules for Training Reward Models was accepted at ICML 2025.
[04/21/25] I will be doing research internship at Amazon (New York) this summer.
[04/15/25] Our work on "Mapping a dynamic sensory panorama onto allocentric direction representations in goal-directed navigation" was selected as a flash talk at Janelia Grounding Cognition in Mechanistic Insight Conference.
[04/05/25] I gave a internal departmental talk on my work "Mechanisms for balancing course stabilization and exploration."
[03/05/25] Our work on Data-adaptive Safety Rules for Training Reward Models was accepted at ICLR BiAlign 2025 workshop.
[01/27/25] Our work on "Mapping a dynamic sensory panorama onto allocentric direction representations in goal-directed navigation" was accepted at Janelia Grounding Cognition in Mechanistic Insight Conference.
[01/23/25] Our submission to Humanity's Last Exam was selected as the top 550 (ranked 51) and included in the released dataset.
[01/22/25] Our work on HARDMath: A Benchmark Dataset for Challenging Problems in Applied Mathematics was accepted at ICLR 2025.
[10/09/24] Our work on HARDMath: A Benchmark Dataset for Challenging Problems in Applied Mathematics was accepted at NeurIPS MATH-AI 2024 workshop.
[05/12/24] I started my internship at Meta Reality Labs Research.
Education
Ph.D. in Neuroscience
Sept. 2020 - May 2026
Cambridge, MA
M.S. in Applied Math
Jan. 2023 - March 2024
Cambridge, MA
B.S. in Brain and Cognitive Sciences
Sept. 2017 - May 2020
Cambridge, MA
Experience
MLRE | Agentic Post-Training
May 2026 - present
New York, NY
Research Intern | Ads/Recommendation GenAI
June 2025 - Dec. 2025
New York, NY
Research Intern | Reality Labs Research
May 2024 - Aug. 2024
Redmond, WA
Software Engineer Intern
Sept. 2023 - Dec. 2023
Cambridge, MA
Research Intern | Merck Research Laboratories
June 2018 - Aug. 2018
Westpoint, PA
Services
Reviewer for BI 2024, ICLR 2025, ICLR 2026, CVPR 2026, NeurIPS 2026, ICLR 2027. Session Chair and moderator for Janelia Grounding Cognition in Mechanistic Insight Conference