Failure-Transparent Agents
Benchmarking Post-Failure Reporting in Tool-Using Language Models
Zhu et al. · arXiv:2609.35732
Benchmarks whether tool-using language models truthfully report failed actions and tests a structured evidence contract.
Papers
Reliable AI agents, cybersecurity, graph learning, and Android malware detection.
Open a title to read the paper. Select Cite to copy BibTeX. Download all
LeaseGuard
Incumbent-Preserving Admission Control for Privileged LLM Agents
Zhu et al. · arXiv:2609.24077
Prevents privileged agents from displacing healthy resource owners through deterministic admission control.
SFCMOEA
A Hybrid Clique-Based Method with Structural Feature Node Extraction for Community Detection in Overlapping Networks
Ma et al. · CMC 87(1)
Combines maximal cliques, structural feature nodes, and MOEA/D for overlapping community detection.
Partition-Matched Evaluation
Community Features under Distribution Shift in Android Malware Function-Call Graphs
Zhu et al. · arXiv:2609.25256
Tests whether graph-community features remain useful under Android malware shift using partition-matched controls.
Structure-Aware Calibration
Stress-Testing Malware Graph Neural Networks under Type Shift
Zhu et al. · arXiv:2609.28517
Studies calibration of malware graph neural networks under unseen-type shift and evaluates a support-aware fallback.
Residual Community Prototypes
Under-Reject Held-Out Malware Families in FCG-MFD
Zhu et al. · arXiv:2609.24980
Evaluates residual community prototypes for rejecting malware families not observed during training.
Android Malware Detection
Android Malware Detection Based on Deep Learning
Zhang, Zou & Zhu · IEEE ICCC
Applies deep learning to Android malware detection and classification.