MedMT-Bench: Testing LLMs on Long Medical Conversations
MedMT-Bench evaluates LLMs' ability to handle long multi-turn medical conversations, revealing current AI limitations in clinical dialogue understanding.
Cluster-R1: Instruction-Following Large Reasoning Models
Discover Cluster-R1, a novel approach where large reasoning models follow instructions to improve autonomous data clustering and interpretation.
Symbolic-Mechanistic Evaluation for AI Beyond Accuracy
Discover a symbolic-mechanistic approach to AI evaluation that reveals true model generalization beyond misleading accuracy metrics.
MSA: Efficient Memory Sparse Attention for 100M Token AI Models
Discover MSA, a scalable memory sparse attention model enabling efficient AI processing of up to 100 million tokens with minimal performance loss.
Medical Coding with LLMs Using Privacy-Preserving Synthetic Data
Enhance medical coding accuracy using large language models trained on privacy-preserving synthetic clinical data for safer, efficient healthcare automatio...