Conceptinference-time scaling / 测试时计算扩展 / test-time compute
Inference-Time Scaling
A technical approach that improves AI model performance by increasing computation during the inference phase rather than during training. It has emerged as an important alternative direction as the returns from pre-training scaling laws diminish.
Timeline (last 90 days)
Oct 9
One of the fundamental reasons the industry has actively explored alternative paths such as post-training, reinforcement learning, and inference-time scaling in recent years is the diminishing marginal returns of pre-training scaling laws.
Unverified50%