Papers by Guillermo Sapiro
CLARO: Controlled Attribute-Driven Reasoning Optimization for Efficient Chain-of-Thought (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent approaches guide reasoning length through token penalties or truncation, risking the omission of necessary steps. |
| Approach: | They propose a framework to internalize conciseness and attribute-driven reasoning optimization . they propose guiding models to embed high-quality structural attributes within a token budget . |
| Outcome: | The proposed method outperforms state-of-the-art models across benchmarks yielding accuracy gains of up to 63.6%. |