Jianqiao Lu   陆剑桥

About

I am a Research Engineer at ByteDance Seed Foundation Model, contributing to large-scale foundation model pretraining. My current focus is scaling and efficiency—especially training stability, throughput/cost, and long-context training & inference efficiency.

Interests

• Scaling & efficiency: training stability, throughput, and cost.
• Long context: efficient training and inference/prefill for long sequences.
• Systems & architecture: practical optimizations that improve stability and efficiency at scale.

Previously

I was a PhD student supervised by Prof. Zhiyi Huang at The University of Hong Kong (HKU).

Contact

• Email: tjlujianqiao@gmail.com

Collaboration

During my PhD, I worked closely with Dr. Zhijiang Guo, Dr. Zhengying Liu, Dr. Yinya Huang, and Yingjia Wan on formal and informal mathematical reasoning for large language models — covering autoformalization, theorem proving, and process-supervised verification.

News

• Sep. 2024:  (First Author) AutoPSV: Automated Process-Supervised Verifier, accepted to NeurIPS 2024.
• Sep. 2024:  Proving Theorems Recursively, accepted to NeurIPS 2024.
• Sep. 2024:  Mr.Bean: A Comprehensive Meta-Reasoning Benchmark for Analyzing Large Language Models, accepted to NeurIPS 2024.
• Sep. 2024:  FVEL: Interactive Formal Verification Environment with Large Language Models via Theorem Proving, accepted to NeurIPS 2024 (Datasets and Benchmarks Track). [Project]
• Sep. 2024:  Online Matching Meets Sampling Without Replacement, accepted to WINE 2024.
• May 2024:  Planning, Creation, Usage: Benchmarking LLMs for Comprehensive Tool Utilization in Real-World Complex Scenarios, accepted to ACL 2024. [Code]
• Oct. 2023:  (First Author) Improving End-to-End Speech Processing by Efficient Text Data Utilization with Latent Synthesis, accepted to EMNLP 2023.

Preprint

Large Language Models

• Jun. 2024:  (First Author) FormalAlign: Automated Alignment Evaluation in Autoformalization, arXiv.
• May 2024:  (First Author) Process-Driven Autoformalization in Lean 4, arXiv.
• Jan. 2024:  (Equal Contribution) MHPP: Exploring the Capabilities and Limitations of Language Models Beyond Basic Code Generation, arXiv.
• Jan. 2024:  (First Author) YODA: Teacher-Student Progressive Learning for Language Models, arXiv.
• Oct. 2023:  (First Author) SELF: Self-Evolution with Language Feedback, arXiv.