Overthinking in LLMs: Optimize Test-Time Compute Scaling
Discover how overthinking impacts large language models and learn strategies to optimize test-time compute for better efficiency and accuracy.
Training Language Models to Code Like Students Using Dialogue
Learn how conversational serialization trains language models to mimic student coding and debugging, enhancing programming education with authentic data.
SciPredict: Can LLMs Accurately Predict Science Experiments?
Explore SciPredict benchmark evaluating LLMs' ability to predict scientific experiment outcomes in physics, biology, and chemistry with limited accuracy.
Camyla: Autonomous Medical Image Segmentation Research
Discover Camyla, an autonomous system revolutionizing medical image segmentation with over 2,700 models and 40 manuscripts in 28 days.
FACT-E: Reliable Causality-Based Chain-of-Thought Evaluation
Discover FACT-E, a causality-inspired method enhancing trustworthiness in Chain-of-Thought reasoning for more accurate AI model evaluations.