Xilin Liu

Papers on Lattice

Total citations

Topics

h-index

Publication activitypapers/week, last 8 weeks

Research focus

Code Generation & Program Synthesis (5)Tool Use & Agents (3)Eval Frameworks & Benchmarks (2)Recommendation & Information Retrieval (1)Data Curation & Synthetic Data (1)Architecture Design (Transformers, SSMs, MoE) (1)

Frequent co-authors

Dewu Zheng (4)Ensheng Shi (4)Zibin Zheng (4)Kefeng Duan (3)Jiachi Chen (3)Kefeng Duan (2)Dewu Zheng (2)Ming-Wei Liu (2)

Papers (6)

Sep 1, 2026

Kefeng Duan +18Sep 1, 2026·also Affiliation:

Adaptive Critical Token-Aware Retrieval for Repository-Level Code Generation

Critical tokens can make or break code generation, and ACToR ensures they receive the precise context needed to succeed.

Kefeng Duan, Kefeng Duan, Dewu Zheng +16

Code Generation & Program Synthesis Recommendation & Information Retrieval

Kefeng Duan +16Sep 1, 2026·also Affiliation:

Efficient SWE Agent Benchmarking via Trajectory-Aware Evaluation

PTA-IRT reveals that incorporating detailed problem-solving trajectories can dramatically enhance the evaluation accuracy of software engineering agents.

Kefeng Duan, Kefeng Duan, Dewu Zheng +14

Code Generation & Program Synthesis Eval Frameworks & Benchmarks Tool Use & Agents

Aug 27, 2026

Dewu Zheng +5Aug 27, 2026·also Affiliation:

From Static to Dynamic: Benchmarking Real-World Code Review with MCR-Bench

Mainstream LLMs miss 40% of defects in multi-round code reviews, revealing their limitations in real-world software development contexts.

Dewu Zheng, Xi-Wen Wang, Kefeng Duan +3

Code Generation & Program Synthesis Eval Frameworks & Benchmarks

Dewu Zheng +6Aug 27, 2026·also Affiliation:

SWE-Prime: Fewer Trajectories, Better Performance

Training on just 10% of carefully curated trajectories can lead to performance improvements of over 24% in software issue resolution tasks.

Dewu Zheng, Rui Ye, Yang Ye +4

Code Generation & Program Synthesis Data Curation & Synthetic Data

Jul 21, 2026

Jul 21, 2026·also Huawei, HUST, ZJU

PhoenixRepair: Rethinking Repair Strategy Exploration in Software Agents

PhoenixRepair redefines how software agents explore repair strategies, achieving a 76% resolution rate by leveraging multi-location sampling and iterative refinement.

Tianyue Jiang, Xinabang He, Jiachi Chen +3

Code Generation & Program Synthesis Tool Use & Agents

Apr 2, 2026

Apr 2, 2026·also Affiliation:, HITSZ, Huawei

Memory in the LLM Era: Modular Architectures and Strategies in a Unified Framework

LLM agent memory architectures matter, and a systematic comparison reveals a new SOTA method by combining the best of existing approaches.

Yanchen Wu, Teng Lin, Fangyuan Zhang +5

Architecture Design (Transformers, SSMs, MoE)Reasoning & Chain-of-Thought Tool Use & Agents