1 paper · 1 filter
Matthew Ho, Chen Si, Zhaoxiang Feng +5
While inference-time scaling enables LLMs to carry out increasingly long and capable reasoning traces, the patterns and insights uncovered during these traces are immediately disca…