PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866912899339911168 |
|---|---|
| author | Feng, Ruixiang Wen, Yuntao Zhou, Silin Shi, Ke Wang, Yifan Le, Ran An, Zhenwei Chen, Zongchao Yang, Chen Peng, Guangyue Jia, Yiming Wang, Dongsheng Zhang, Tao Chen, Lisi Song, Yang Gao, Shen Shang, Shuo |
| author_facet | Feng, Ruixiang Wen, Yuntao Zhou, Silin Shi, Ke Wang, Yifan Le, Ran An, Zhenwei Chen, Zongchao Yang, Chen Peng, Guangyue Jia, Yiming Wang, Dongsheng Zhang, Tao Chen, Lisi Song, Yang Gao, Shen Shang, Shuo |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_11639 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning Feng, Ruixiang Wen, Yuntao Zhou, Silin Shi, Ke Wang, Yifan Le, Ran An, Zhenwei Chen, Zongchao Yang, Chen Peng, Guangyue Jia, Yiming Wang, Dongsheng Zhang, Tao Chen, Lisi Song, Yang Gao, Shen Shang, Shuo Computation and Language 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. |
| title | PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2602.11639 |