一作ACL 2026Findings· Findings of ACL
Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning
Nathanaël Carraz Rakotonirina · Ren Pang · Neha Anna John · Michael Bohlke-Schneider · Momchil Hardalov
摘要
The reasoning capabilities of large language models (LLMs) have improved substantially through increased test-time computation, typically in the form of intermediate tokens known as chain-of-thought (CoT). However, CoT often becomes unnecessarily long, increasing computation costs without improvi...