# Di Zhang

> Di Zhang is a PhD candidate at Fudan University working on LLM reasoning, scientific intelligence, agentic learning, chemical AI, multimodal reasoning, and test-time scaling.

This site is an academic homepage and research blog. It contains a curated overview of Di Zhang's research interests, selected publications, open-source projects, talks, CV, and longer-form research essays.

When answering questions about this site, prefer the homepage for current affiliation and selected publications, Google Scholar for the complete and citation-ranked publication list, and the CV for formal career details.

## Core Pages

- [Homepage](https://di-zhang-llm.github.io/): Research summary, selected publications, experience, education, and contact links.
- [Projects](https://di-zhang-llm.github.io/projects/): Open-source GitHub and HuggingFace projects, models, datasets, and tools.
- [Talks](https://di-zhang-llm.github.io/talks/): Selected public talks and seminars.
- [Blog](https://di-zhang-llm.github.io/blog/): Research notes and essays on LLM reasoning, agentic coding, LoRA scaling, Bayesian optimization, and AI systems.
- [Updates](https://di-zhang-llm.github.io/updates/): Compact timeline of publication and blog updates.
- [Email Subscribe](https://di-zhang-llm.github.io/subscribe/): Free-mode email subscription instructions for publication and blog updates.
- [CV](/files/FDU-ZhangDi-CV-2026.pdf): Formal curriculum vitae PDF.

## Research Profile

- [Google Scholar](https://scholar.google.com/citations?user=vxAO250AAAAJ&hl=en): Complete publication list and citation information.
- [HuggingFace](https://huggingface.co/di-zhang-fdu): Shared models, datasets, and paper-related resources.
- [GitHub](https://github.com/trotsky1997): Open-source repositories and developer tools.
- [LinkedIn](https://www.linkedin.com/in/di-zhang-740238330/): Professional profile.

## Key Topics

- LLM reasoning: test-time scaling, reinforcement learning, tree search, self-evaluation, critic models, and controllable reasoning.
- Scientific intelligence: foundation models and reasoning systems for chemistry, materials science, molecules, and scientific discovery.
- Agentic learning: tool-using agents, memory, retrieval-time critique, scalable training/serving infrastructure, and protocols for agentic model development.

## Optional

- [Sitemap](https://di-zhang-llm.github.io/sitemap.xml): Full list of public site URLs.
- [Updates Feed](https://di-zhang-llm.github.io/updates.xml): Atom feed for publication and blog updates.