Cuiyun Gao
6 papers · Latest:
Schedule-and-Calibrate: Utility-Guided Multi-Task Reinforcement Learning for Code LLMs
ASTOR is a multi-task RL framework for code LLMs that uses utility-driven data scheduling and policy optimization, outperforming specialists.
When Model Editing Meets Service Evolution: A Knowledge-Update Perspective for Service Recommendation
EVOREC is a framework for service recommendation that uses model editing and constrained decoding to adapt to evolving services and overcome outdated facts.
Cascaded Code Editing: Large-Small Model Collaboration for Effective and Efficient Code Editing
This paper proposes Cascaded Code Editing, combining large models for edit sketch generation and small models for efficient application.
On the Effectiveness of Context Compression for Repository-Level Tasks: An Empirical Investigation
This paper empirically investigates context compression for repository-level code tasks, finding it effective for performance and efficiency.
Evaluating LLM-Based 0-to-1 Software Generation in End-to-End CLI Tool Scenarios
This paper introduces CLI-Tool-Bench, a new benchmark for evaluating LLM-based 0-to-1 software generation, revealing current models struggle with end-to-end CLI tool creation.
Dependency-Guided Repository-Level C-to-Rust Translation with Reinforcement Alignment
DepTrans is a new framework that automates C-to-Rust code migration using reinforcement learning and dependency-guided refinement, achieving high accuracy.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.