About
Research & Development
Safeguarding Enterprise AI: Protect’s Multi-Modal Approach to LLM Safety
TLDR: Protect is a new multi-modal guardrailing system for enterprise LLMs, handling text, image, and audio inputs across four safety dimensions: toxicity, sexism, data...
Research & Development
Unpacking Prompt Design: How Simple Language Outperforms Detail in Zero-Shot Posture Classification
TLDR: A study on zero-shot classification of human postures (sitting, standing, walking/running) using Vision-Language Models (VLMs) found that for high-performing models like MetaCLIP 2...
Research & Development
Precise Parking: Correcting GPS Errors with a Low-Rank Approach
TLDR: A new unsupervised method called RAA (Rectify And Align) uses a low-rank constraint to accurately correct and align GPS points of parking spots,...
Research & Development
Unlocking Deeper Narrative Comprehension in AI with LiteraryQA
TLDR: LiteraryQA is a new, high-quality dataset derived from NarrativeQA, specifically focusing on literary works. It addresses issues like noisy documents and flawed question-answer...
Research & Development
Crafting Authentic Game NPCs: How “Deflanderization” Balances Personality and Purpose
TLDR: This research paper introduces "Deflanderization" techniques to prevent LLM-based game NPCs from becoming overly simplistic or losing task efficiency due to exaggerated role-playing....
Research & Development
CanvasMAR: A Novel Approach for Enhanced Video Generation with Global Frame Prediction
TLDR: CanvasMAR is a new masked autoregressive model for video generation that addresses the 'slow-start problem' and 'error accumulation' by introducing a 'canvas mechanism.'...