Skill^2-Bench Measures Long-Horizon Reasoning via Skill Entropy
August 4, 2026
Skill^2-Bench introduces Skill Entropy to quantify the difficulty of switching between different reasoning skills in LLMs. The benchmark evaluates cross-skill long-horizon tasks across 558 skills and 9 distinct domains.
HOW THIS AFFECTS YOU
●
researcherYou can use Skill Entropy to more precisely evaluate how well models transition between distinct reasoning steps in multi-step tasks.