Small AI Models for Legal Document Reasoning: Study
Explore how small AI models under 10B parameters perform legal reasoning tasks, matching larger models at lower cost and complexity.
Summary-Driven RL for Enhanced Video Understanding
Discover how Summary-Driven Reinforcement Learning improves video understanding by boosting reasoning and temporal comprehension in multi-modal LLMs.
DenseSwinV2: Advanced Cassava Leaf Disease Classification Model
Discover DenseSwinV2, a hybrid CNN-Transformer model achieving 98% accuracy in cassava leaf disease classification with robust attention mechanisms.
DiReCT: Physics-Based Video Generation with Contrastive Learning
DiReCT enhances video generation by disentangling contrastive trajectories, improving physics consistency and visual quality without extra training time.
Multimodal Coherence Score: Improving AI Data Quality
Discover the Multimodal Coherence Score (MCS), a novel metric to evaluate and improve data fusion quality in AI beyond task accuracy.