Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2507.07495 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916836810948608 |
|---|---|
| author | Parmar, Mihir Goyal, Palash Liu, Xin Song, Yiwen Ling, Mingyang Baral, Chitta Palangi, Hamid Pfister, Tomas |
| author_facet | Parmar, Mihir Goyal, Palash Liu, Xin Song, Yiwen Ling, Mingyang Baral, Chitta Palangi, Hamid Pfister, Tomas |
| contents | Recently, decomposing complex problems into simple subtasks--a crucial part of human-like natural planning--to solve the given problem has significantly boosted the performance of large language models (LLMs). However, leveraging such planning structures during post-training to boost the performance of smaller open-source LLMs remains underexplored. Motivated by this, we introduce PLAN-TUNING, a unified post-training framework that (i) distills synthetic task decompositions (termed "planning trajectories") from large-scale LLMs and (ii) fine-tunes smaller models via supervised and reinforcement-learning objectives designed to mimic these planning processes to improve complex reasoning. On GSM8k and the MATH benchmarks, plan-tuned models outperform strong baselines by an average $\sim7\%$. Furthermore, plan-tuned models show better generalization capabilities on out-of-domain datasets, with average $\sim10\%$ and $\sim12\%$ performance improvements on OlympiadBench and AIME 2024, respectively. Our detailed analysis demonstrates how planning trajectories improves complex reasoning capabilities, showing that PLAN-TUNING is an effective strategy for improving task-specific performance of smaller LLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_07495 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | PLAN-TUNING: Post-Training Language Models to Learn Step-by-Step Planning for Complex Problem Solving Parmar, Mihir Goyal, Palash Liu, Xin Song, Yiwen Ling, Mingyang Baral, Chitta Palangi, Hamid Pfister, Tomas Computation and Language Artificial Intelligence Recently, decomposing complex problems into simple subtasks--a crucial part of human-like natural planning--to solve the given problem has significantly boosted the performance of large language models (LLMs). However, leveraging such planning structures during post-training to boost the performance of smaller open-source LLMs remains underexplored. Motivated by this, we introduce PLAN-TUNING, a unified post-training framework that (i) distills synthetic task decompositions (termed "planning trajectories") from large-scale LLMs and (ii) fine-tunes smaller models via supervised and reinforcement-learning objectives designed to mimic these planning processes to improve complex reasoning. On GSM8k and the MATH benchmarks, plan-tuned models outperform strong baselines by an average $\sim7\%$. Furthermore, plan-tuned models show better generalization capabilities on out-of-domain datasets, with average $\sim10\%$ and $\sim12\%$ performance improvements on OlympiadBench and AIME 2024, respectively. Our detailed analysis demonstrates how planning trajectories improves complex reasoning capabilities, showing that PLAN-TUNING is an effective strategy for improving task-specific performance of smaller LLMs. |
| title | PLAN-TUNING: Post-Training Language Models to Learn Step-by-Step Planning for Complex Problem Solving |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2507.07495 |