Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore how to enhance AI model performance through Inference-Time Scaling (ITS) in this 28-minute conference talk from DevConf.US 2025. Learn about an innovative approach that optimizes computational resource allocation during inference rather than relying on traditional methods of scaling model size or training data. Discover how ITS restructures search and evaluation strategies at test-time to significantly improve model output quality without requiring retraining or expanding model parameters. Examine the top ITS methods and gain practical knowledge on implementing these techniques using existing models with off-the-shelf toolkits including reward_hub and inference_time_scaling library. Understand how this orthogonal solution addresses the growing constraints of cost, latency, and diminishing returns in AI model improvement.
