About
Research & Development
Automating Normative Reasoning: A New Approach for Input/Output Logics
TLDR: This research paper introduces a novel method to automate reasoning in Input/Output (I/O) Logics, a type of deontic logic used for formalizing conditional...
Research & Development
REFINE: Enhancing Multimodal AI Performance Through Targeted Error Feedback
TLDR: REFINE is a teacher-student framework that improves Multimodal Large Language Models (MLLMs) by systematically structuring errors into an "Error-book." It uses three types...
Research & Development
Smart Safety for Hybrid Systems: Introducing Uppaal Coshy
TLDR: Uppaal Coshy is a new tool that automatically creates "shields" (safety strategies) for complex hybrid systems with continuous state spaces. It works by...
Research & Development
Beyond Jailbreaks: Unpacking the True Criminal Potential of Large Language Models
TLDR: A new study introduces VENOM, a framework to assess the genuine criminal capabilities of Large Language Models (LLMs) beyond superficial jailbreak attacks. It...
Research & Development
Unpacking AI Performance in Moroccan Legal Question Answering
TLDR: MizanQA is a new benchmark dataset with over 1,700 multiple-choice questions designed to evaluate Large Language Models (LLMs) on Moroccan legal question answering....
Research & Development
Domain-Aligned AI Model Sharpens Extreme Climate Event Projections
TLDR: The research introduces the Domain Aligned Climate Downscaling (DACD) model, an AI-based spatiotemporal downscaling model that significantly improves the simulation and projection of...