About
Research & Development
JPS: A New Frontier in Understanding Multimodal AI Vulnerabilities
TLDR: JPS is a novel method for jailbreaking Multimodal Large Language Models (MLLMs) that goes beyond simply bypassing safety filters. It focuses on generating...
Research & Development
Integrated Influence: A New Path to Understanding AI Decisions Through Data Attribution
TLDR: A new research paper introduces "Integrated Influence," a novel data attribution method that overcomes limitations of existing techniques by incorporating a baseline dataset...
Research & Development
Chemist Eye: AI-Driven Safety for Self-Driving Laboratories
TLDR: Chemist Eye is a new AI-powered system designed to enhance safety in self-driving laboratories (SDLs). It uses cameras (RGB, depth, infrared) and vision-language...
Research & Development
Enhancing AI Tool Use with Dependency-Aware Retrieval
TLDR: The "Tool Graph Retriever" (TGR) is a new method that helps AI agents find the right tools more effectively by understanding how tools...
Research & Development
HIA: A Budget-Friendly Approach to Aligning Large Language Models
TLDR: HIA (Heuristic-Guided Inference-time Alignment) is a new framework that aligns Large Language Models (LLMs) with user preferences efficiently and cost-effectively. It uses a...
Research & Development
Enhancing Code Generation with Reasoning-Aware Reinforcement Learning
TLDR: This research introduces Posterior-GRPO (P-GRPO), a novel reinforcement learning framework for large language models (LLMs) that rewards the quality of the intermediate reasoning...