CV

Research Interests

Code intelligence, repository-level code evolution, code retrieval, static program analysis, graph algorithms, and LLM agents.

Research internship interests: code intelligence / LLM agents.

Contact: a496263365a@163.com.

Education

Peking University — School of Computer Science

Direct-entry Ph.D. in Computer Science and Technology · September 2025–June 2030 (expected)

• Research area: software engineering.

Beijing Institute of Technology — School of Computer Science

Bachelor’s degree in Computer Science and Technology · September 2021–June 2025

• Average credit-weighted score over the first three undergraduate years: 92.7/100.
• Rank over the first three undergraduate years: 1/162 in the major.

Honors and Awards

• National Scholarship, twice.
• Beijing Outstanding Student (北京市三好学生).
• Beijing Outstanding Graduate (北京市优秀毕业生).

Research Experience and Publications

FeatX: Editing Software by Editing Features for Repository-Level Code Evolution

Xutian Li, Yifeng Zhu, Xianlin Zhao, Yanzhen Zou, Lu Zhang, Bing Xie

Published · ASE 2026, Tool and Dataset Track

Research period: July 2025–May 2026

FeatX addresses the need to manually manage context and locate feature implementations in code-centric workflows. It uses an explicit feature hierarchy and feature–code mappings to drive repository-level code editing.

• Combines JavaParser-based static analysis with LLM-guided exploratory retrieval to assemble cross-file context.
• Locates target code from semantic feature changes, plans modification dependencies and execution order, and produces reviewable code diffs.
• Across 38 real feature-change commits, achieved 0.385 function-level change-localization F1, a 42.6% relative improvement over the strongest baseline.
• In a 10-participant user study, reduced average NASA-TLX workload scores by 41% compared with ChatGPT-based conversation.

Publication: ASE 2026, Tool and Dataset Track.

Paper: doi.org/10.1145/3832783.3834646.

FeatLens: Feature-Guided Dependency Retrieval for Repository-Level Code Generation

Xutian Li, Bo Xiong, Yifeng Zhu, Runbang Yan, Yanzhen Zou, Lu Zhang, Bing Xie

Preprint · arXiv preprint, 2026

Research period: December 2025–October 2026

FeatLens connects functional requirements with dependencies dispersed across a repository, reducing the need for multi-step online LLM exploration.

• Builds an offline feature–function index and expands entities along call, use, and ownership relations.
• Combines semantic relevance and structural heuristics to construct a personalized PageRank restart distribution and select context under a fixed budget.
• Makes zero LLM calls during online dependency retrieval. Offline indexing and downstream code generation are separate stages; this is not a claim that the entire pipeline is LLM-free.
• At the same Top-15 budget, improved dependency recall over UniXcoder RAG by 11.2 percentage points on DevEval and 17.4 percentage points on EvoCodeBench.
• On DevEval, achieved 55.03% generation Pass@1 with GPT-5-mini.

Status: Preprint. The results above follow the manuscript experiments reported in my CV.

Preprint: arxiv.org/abs/2609.26480.

FeatOutline: Feature-Oriented Repository Summarization for Enhanced Code Comprehension and Maintenance

Yifeng Zhu, Xutian Li, Xianlin Zhao, Zexiong Ma, Kunze Li, Haizhuo Yuan, Yanzhen Zou, Bing Xie

Published online on November 7, 2026 · ACM Transactions on Software Engineering and Methodology (TOSEM), 2026

Research period: November 2024–November 2025

FeatOutline addresses the challenge of organizing repository summaries by features while preserving fine-grained links to their implementations.

• Formalizes feature hierarchies and feature–code mappings, directly connecting features to implementing methods to support comprehension and localization.
• Evaluates traceability at both file and method levels, distinguishing file relevance from precise method localization.
• Across 32 tasks derived from real commits, achieved 57.4% method-level Link@3 F1, compared with 13.6% for the strongest baseline.

Publication: ACM Transactions on Software Engineering and Methodology (TOSEM), published online on November 7, 2026.

Paper: doi.org/10.1145/3862475.

Course Project

Repetitive Generation Analysis and Neuron Intervention in Language Models

Computational linguistics course project · September 2025–January 2026

Code: github.com/uniqueFranky/LLMRep

• Reproduced a method for identifying repetition-related neurons. Used PyTorch forward hooks to collect MLP activations, selected candidate neurons by differences in mean activation before and after repetition, and zeroed their activations during inference.
• In the course report’s GPT-2 / IWSLT experiment, neuron suppression alone reduced Rep-3 from 70.66% to 24.54%, while perplexity increased from 1.589 to 3.732. These results show a repetition–perplexity trade-off, not a lossless improvement.

Teaching Experience

• 2026 Spring, Teaching Assistant, Data Structures and Algorithms B (数据结构与算法 B).

Technical Skills

• Programming: Java, Python, JavaScript, C++.
• Research and development: static program analysis, code retrieval, graph algorithms, and LLM agent tooling.