9 papers
When to Compile a Computer-Use Agent? Measuring Payback and Making Compilation Decisions for Token Efficiency
Yulong Ming, Jie Xu, Zihan Wu +1
Compiling GUI procedures that agents execute repeatedly into programs can reduce their token costs. However, measuring payback and deciding when to compile have two challenges. Fir…
A Unified Issue Resolution Benchmark for Requirement Clarification, Planning, and Code Generation for Coding Agents
Xin Zhou, Chun Yong Chong, Kisub Kim +11
Large language model-powered coding agents are increasingly used to modify existing code repositories, for example, by adding features or fixing bugs. Yet existing repository-level…
What Resolve Rate Hides: Trajectory Structure Diagnostics for Coding Agents
Rui Shu, Chun Yong Chong, Xin Zhou +6
Coding agents are ranked almost entirely by resolve rate: whether their final patch passes the target tests. Yet two agents can reach the same outcome through very different proces…
PRAG: Efficient Privacy-Preserving RAG Service Supporting Arbitrary Top- Retrieval
Yulong Ming, Mingyue Wang, Jijia Yang +4
Retrieval-Augmented Generation (RAG) enables large language models to use external knowledge, but outsourcing the RAG service raises privacy concerns for both data owners and users…
MulVul: Retrieval-augmented Multi-Agent Code Vulnerability Detection via Cross-Model Prompt Evolution
Zihan Wu, Jie Xu, Yun Peng +2
Large Language Models (LLMs) struggle to automate real-world vulnerability detection due to two key limitations: the heterogeneity of vulnerability patterns undermines the effectiv…
Can Local Learning Match Self-Supervised Backpropagation?
Wu S. Zihan, Ariane Delrocq, Wulfram Gerstner +1
While end-to-end self-supervised learning with backpropagation (global BP-SSL) has become central for training modern AI systems, theories of local self-supervised learning (local-…
Dynamic Influence Tracker: Measuring Time-Varying Sample Influence During Training
Jie Xu, Zihan Wu
Existing methods for measuring training sample influence on models only provide static, overall measurements, overlooking how sample influence changes during training. We propose D…
LiveVal: Time-aware Data Valuation via Adaptive Reference Points
Jie Xu, Zihan Wu, Cong Wang +1
Time-aware data valuation enhances training efficiency and model robustness, as early detection of harmful samples could prevent months of wasted computation. However, existing met…
LMEraser: Large Model Unlearning through Adaptive Prompt Tuning
Jie Xu, Zihan Wu, Cong Wang +1
To address the growing demand for privacy protection in machine learning, we propose a novel and efficient machine unlearning approach for \textbf{L}arge \textbf{M}odels, called \t…