Language Reasoning Models (LRMs) achieve strong performance by scaling test-time computation but often suffer from ``overthinking'', producing excessively long reasoning traces that increase latency and memory usage. Existing LRMs typically enforce conciseness with uniform length penalties, which over-compress crucial early deduction steps at the sequence level and indiscriminately penalize all queries at the group level. To solve these limitations, we propose \textbf{\model}, a dual-level framework for prefix-protected and difficulty-aware compression under hierarchical supervision. At the sequence level, prefix-protected optimization employs decaying mixed rollouts to maintain valid reasoning paths while promoting conciseness. At the group level, difficulty-aware penalty dynamically scales length constraints based on query complexity, maintaining exploration for harder questions while curbing redundancy on easier ones. Extensive experiments on DeepSeek-R1-Distill-Qwen (1.5B/7B) demonstrate that \model achieves a substantial reduction in token usage (up to \textbf{55.7\%}) while simultaneously improving accuracy (up to \textbf{4.1\%}) on math benchmarks, with generalization ability to code, science, and general domains.
@article{arxiv.2602.11639,
title = {PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning},
author = {Ruixiang Feng and Yuntao Wen and Silin Zhou and Ke Shi and Yifan Wang and Ran Le and Zhenwei An and Zongchao Chen and Chen Yang and Guangyue Peng and Yiming Jia and Dongsheng Wang and Tao Zhang and Lisi Chen and Yang Song and Shen Gao and Shuo Shang},
journal= {arXiv preprint arXiv:2602.11639},
year = {2026}
}