@inproceedings{feng-etal-2026-pace,
    title = "{PACE}: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning",
    author = "Feng, Ruixiang  and
      Wen, Yuntao  and
      Zhou, Silin  and
      Shi, Ke  and
      Wang, Yifan  and
      Le, Ran  and
      An, Zhenwei  and
      Chen, Zongchao  and
      Yang, Chen  and
      Peng, Guangyue  and
      Jia, Yiming  and
      Wang, Dongsheng  and
      Zhang, Tao  and
      Chen, Lisi  and
      Song, Yang  and
      Gao, Shen  and
      Shang, Shuo",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.1545/",
    pages = "30884--30903",
    ISBN = "979-8-89176-395-1",
    abstract = "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 PACE, 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 PACE achieves a substantial reduction in token usage (up to 55.7{\%}) while simultaneously improving accuracy (up to 4.1{\%}) on math benchmarks, with generalization ability to code, science, and general domains."
}