Curriculum Vitae
Education
- M.S. in Computer Science, National University of Singapore (NUS), 2026.8 – 2028.7 (expected)
- B.S. in Business Analytics, Shanghai University of Finance and Economics (SUFE), 2022.9 – 2026.6
- TOEFL: 101 · GRE: 327 · JLPT: N2
Research & Industry Experience
Tencent Youtu AI Lab — Research Intern (2025.10 – 2026.5)
- LLM-driven adversarial data synthesis for content-safety evaluation.
- Contributed to a unified multi-label content-moderation model on an XLM-RoBERTa backbone.
- Co-developed a multimodal embedding model for cross-modal retrieval.
SUFE FinAI Center — Research Intern, advised by Prof. Liwen Zhang (张立文) (2025.6 – present)
- Financial reasoning LLM: CoT data curation, SFT + GRPO training, evaluation on FinEval / FinQA / ConvFinQA.
- Co-design of financial evaluation suites and agent-safety benchmarks.
Shanghai Jiao Tong University (SJTU) — Research Intern, advised by Prof. Xiaodong Gu (顾小东) (2026.3 – present)
- Research direction: LLM for Code — code generation, repair, and program understanding.
ICBC Head Office, Private Banking Department — Research Intern (2026.1 – 2026.4)
- Agent + RAG system for scientist discovery, relationship construction, and one-customer-one-strategy marketing.
- Private-banking sparring assistant producing sales scripts and recommendations.
LanMa Technology (澜码科技) — Agent Development Intern (2024.1 – 2024.4)
- Agent development: workflow orchestration, tool invocation, and prompt engineering for enterprise-grade LLM agents.
Publications
Zhi Yang, Runguo Li, Qiqi Qiang, Jiashun Wang, Fangqi Lou, et al. (2026). "FinVault: Benchmarking Financial Agent Safety in Execution-Grounded Environments." arXiv preprint arXiv:2601.07853. Submitted to EMNLP 2026.
Selected Projects
ARH — AI Research Helper (Open Source · github.com/LiRunGuo/Arhelper)
- Background: A CLI-first research assistant agent for AI researchers — reading papers, searching code, running experiments, writing summaries.
- Approach: Plan/Execute safety model with approval-gated writes; three-layer memory (regex + LLM preference extraction, BM25 cross-session recall, end-of-session reflection summary); skill self-learning; multi-LLM fallback (OpenAI / Anthropic / Ollama); automatic context compression; checkpoint & resume.
- Stack: Python 3.9+, Pydantic, FastAPI, SQLAlchemy; Telegram & CLI gateways; arXiv / HuggingFace / GitHub / OpenReview crawlers.
- Status: 0.2.0-alpha, MIT licensed, all data stored locally in
~/.arh/.
Skills
- LLM Development — SFT / GRPO / PPO, distillation, deployment (vLLM, HuggingFace)
- Multimodal Models — VLM fine-tuning, model merging (SLERP / TIES / DARE), adversarial data synthesis
- Agents & RAG — agent workflow orchestration, retrieval, knowledge bases
- Frameworks & Tools — Python, PyTorch, HuggingFace, FAISS, Linux, Docker
Talks
Teaching
Honors & Awards
- National Encouragement Scholarship
- People’s Scholarship, 2nd Class
- Silver Medal, FLTRP English Competition (Municipal)
Languages
- Chinese (native), English (TOEFL 101 · GRE 327), Japanese (JLPT N2)
