OneBench
Shorthand for Thought: Compressing LLM Reasoning via Entropy-Guided Supertokens | OneBench: AI Insights