Efficient Token Pruning for Large Vision Language Models
Discover a novel token pruning method that improves efficiency and performance in large vision language models without extra training.
BioCOMPASS: Biomarker-Driven Immunotherapy Response Prediction
Discover BioCOMPASS, a transformer-based model integrating biomarkers to improve immunotherapy response predictions and enhance treatment personalization.
Spectral Compact Training for Efficient Large Language Models
Discover Spectral Compact Training, a memory-efficient method enabling large language model training on consumer hardware with truncated SVD and Stiefel QR...
Automated CEFR-Based Assessment of Scratch Programming Skills
Discover a CEFR-aligned framework using Fuzzy C-Means to automate and enhance assessment of programming skills in Scratch for personalized learning.
GRASP: Optimizing Multi-Agent Collaboration with Gradient Realignment
Discover GRASP, a novel framework enhancing multi-agent collaboration by realigning gradients and active shared perception for faster, stable optimization.