Binfeng (Bill) Xu

I am a Senior Research Engineer at Samsung Research America (SRA) training large foundation models towards better reasoning/planning, multimodality and deterministic system control.

I earned MS. degree in Data Science from NYU in 2022, during which I briefly worked with Alfredo Canziani and Yann LeCun on policy learning for autonomous driving. Prior to this, I completed BS. degree from WFU in 2020, doubling Computer Science and Statistics, advised by Grey Ballard on efficient Machine Learning (Tucker Decomposition) for neuro fMRI and by Paúl Pauca on multiple Computer Vision applications. I was active on Kaggle ML competitions before 2022, ranking top 1% globally and titled 'Competition Master'.

Since 2018, I had experienced a broad range of AI/ML research and applications. I like to build products/systems out of ideas end-to-end.

Email  |  CV  |  Twitter  |  GitHub  |  LinkedIn  |  Kaggle

Interests & Projects (dm for discussion or collaboration)

World Modeling and Machine Intelligence. Generally bipartites into:

  • Conceptual learning: Modeling multimodal world patterns with (very) long-context autoregressive transformers. eg. LWM.
  • Generative reasoning: While Symbolic Learning models (eg. AlphaGo) shows the strength at (relatively) small action space, I'm most interested in GFlowNet in stochastic prediction of wild world space.

Tool-augmented LLM Agents: Taking autonomous actions with optimized reasoning and planning from large models.

  • Gentopia: A clean framework to build hierarchical agents with config, simplifying specialization, eval, sharing and inheritance. [demo].
  • ReWOO: Eliminate stacking redundancy in ALM systems by decoupling LLM reasoning from observations.

LLMs for RecSys. Using pretrained world knowledge in LLMs to generate catalog / knowledge graphs, so as to create new recalls and escape popularity bias; Using LLMs to mock users, initializing traffic to help cold-start.

Quantitative Models. I built a quant model in 2020 to scan bitcoin trade signals. Later TempoQuant helped me predict Nasdaq dip on 10/13/22 (few believed me on the new bull). I'm thinking of open-sourcing the framework.

Papers

Gentopia: A Collaborative Platform for Tool-Augmented LLMs
Binfeng Xu, Xukun Liu, Hua Shen, Zeyu H, Yuhan L, Murong Y, Zhiyuan P, Yuchen L, Ziyu Y, Dongkuan Xu

ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models
Binfeng Xu, Zhiyuan Peng, Bowen Lei, Subhabrata Mukherjee, Yuchen Liu, Dongkuan Xu

Efficient Computation of Tucker Decomposition of Correlation-Based Tensors
Binfeng Xu, Grey Ballard, Robert Lyday, Paul Laurienti

Iterative Constringency Optimization: Preclustering Approach to Agent Interactive Data
Binfeng Xu, Nicole Dalzell

Reviewer: ACL 23', KDD 23', AAAI 24',
Misc

Petting Kobu, a cute Norway Forest 🐱; Cyberpunk; Digital nomad (someday); Fan of all games by Hidetaka Miyazaki, who motivated me once into indie game devs; Good at Dota2 (once). Photography @500px; Minimalist;