BadImplant: Multi-Targeted Injection Backdoor Attack on GNNs
Discover BadImplant, a novel injection-based multi-targeted backdoor attack on graph neural networks that preserves graph integrity and evades defenses.
Q-Probe: Advanced High-Resolution Image Quality Assessment
Discover Q-Probe, a context-aware framework enhancing high-resolution image quality assessment with superior accuracy and human-aligned evaluation.
Enhanced Evidence Extraction for Document Inconsistency with LLMs
Discover improved metrics and a redact-and-retry framework using LLMs to boost document inconsistency detection accuracy and efficiency.
LAsset: Automated Security Asset ID for SoC Verification
LAsset uses LLMs to automate security asset identification in SoC designs, boosting accuracy and scalability in hardware security verification.
PhyAVBench: Benchmark for Physically Grounded Text-to-Audio-Video
Discover PhyAVBench, a benchmark to evaluate audio physics grounding in text-to-audio-video models for realistic AI-generated audiovisual content.