Human Reliability Analysis; LLM + Fuzz Testing for UAVs; High Fidelity Scene Reconstruction and Understanding for AVs; LMMs for Video OCR

Tangible AI Digest - Research (31/12/2024 - 06/01/2025)

?? Large Language Models (LLMs)

1. Human Reliability Analysis: Xingyu Xiao and colleagues developed a novel framework (KRAIL) that combines expert knowledge with large language models to estimate human error probability (HEP) in complex systems. This approach enables semi-automated computation and shows superior performance in experiments using authoritative human reliability datasets. READ
2. Autonomous Systems: Taohong Zhu and colleagues developed SAFLITE, a predictive component that uses large language models to improve fuzz testing efficiency for Autonomous Systems (AS) (Zhitao: targeting UAVs). Experiments with various LLMs and fuzz testing tools showed significant increases (up to 234.5%) in detecting software vulnerabilities in autonomous drone control systems. READ

?? Autonomous Vehicles (AVs)

1. High-Fidelity Scene Reconstruction: Rami Wilson introduces GSAVS, a simulator leveraging Gaussian Splatting for high-fidelity scene reconstruction. This approach enhances the realism and customization of virtual environments, crucial for developing autonomous vehicle models. READ
2. Semantic Scene Understanding: Jiaqi Fan and team utilize multimodal large language models for semantic scene understanding and risk localization in traffic scenarios, achieving superior performance on existing datasets. READ

?? Large Multimodal Models (LMMs)

1. LMMs for Video OCR?: Led by Yulin Fei and Wei Chen, the team introduces a benchmark for evaluating video OCR capabilities of multimodal models, highlighting the need for improved OCR abilities in LMMs. READ

Stay tuned for more updates in next week’s digest!

Want to learn more about how we curate the latest AI digest or get your hands dirty on the analysis? Reach out to discover how Tangible AI Digest is created and stay connected with exclusive content, discussions, and updates. Subscribe now to stay ahead in the AI revolution!