Paper Accepted by ACL 2025

Our paper has been accepted by ACL 2025 Findings.

Our paper titled “The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding?” by Yutao Sun, Mingshuai Chen, Tiancheng Zhao, Ruochen Xu, Zilun Zhang, and Jianwei Yin has been accepted by ACL 2025 (Findings) – The 63rd Annual Meeting of the Association for Computational Linguistics at Vienna, Austria. This paper presents Crescent – a simple yet effective framework for generating high-quality synthetic data for improving the mathematical reasoning capabilities of an LLM with zero supervision signals while preserving its general performance.