About
Research & Development
S4T: A Novel Method for Harmonizing Multi-Task AI Adaptation
TLDR: S4T (Synchronizing Tasks for Test-time Training) is a new approach that tackles the 'unsynchronized task behavior' problem in multi-task AI models when adapting...
Research & Development
Dynamic Prompting for Enhanced Vision Transformer Performance
TLDR: Visual Instance-aware Prompt Tuning (ViaPT) is a new method that significantly improves how AI vision models adapt to new tasks. Unlike traditional methods...
Research & Development
Unlocking Logical Meanings: How AI Translates Complex System Rules
TLDR: This research introduces a Transformer-based model capable of inverting semantic embeddings of Signal Temporal Logic (STL) formulae back into human-readable logical specifications. The...
Research & Development
Smart Pruning: UnIT Brings Efficient AI to Small Computing Devices
TLDR: UnIT is a new method for making deep neural networks run efficiently on low-power microcontrollers. It dynamically skips unnecessary calculations during inference, without...
Research & Development
DTECT: A New System for Tracking Dynamic Topics in Textual Data
TLDR: DTECT is an open-source, end-to-end system that helps users understand how topics evolve in large text datasets over time. It combines dynamic topic...
Research & Development
Smaller AI Models Show Significant Promise for Rheumatology Care
TLDR: A study evaluating large and small language models (LLMs/SLMs) in rheumatology clinical decision support found that SLMs, especially Mixtral-8x7b-32768 combined with Retrieval-Augmented Generation...