Chunks as Arms: Multi-Armed Bandit-Guided Sampling for Long-Context LLM Preference Optimization
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866911577713672192 |
|---|---|
| author | Duan, Shaohua Huang, Pengcheng Li, Xinze Liu, Zhenghao Yi, Xiaoyuan Yan, Yukun Wang, Shuo Gu, Yu Yu, Ge Sun, Maosong |
| author_facet | Duan, Shaohua Huang, Pengcheng Li, Xinze Liu, Zhenghao Yi, Xiaoyuan Yan, Yukun Wang, Shuo Gu, Yu Yu, Ge Sun, Maosong |
| contents | Long-context modeling is critical for a wide range of real-world tasks, including long-context question answering, summarization, and complex reasoning tasks. Recent studies have explored fine-tuning Large Language Models (LLMs) with synthetic data to enhance their long-context capabilities. However, the effectiveness of such approaches is often limited by the low diversity and factual inconsistencies in the generated data. To address these challenges, we propose LongMab, a novel framework that leverages a Multi-Armed Bandit (MAB) rollout strategy to identify the most informative chunks from the given long context for sampling high-quality and diverse responses and constructing preference data pairs for Direct Preference Optimization (DPO) training. Specifically, we treat context chunks as arms of MAB, select chunks based on their expected reward scores to input into LLMs to generate responses, and iteratively update these scores based on reward feedback. Both exploration and exploitation during the rollout process enable the LLM to focus on the most relevant context segments, thereby generating and collecting high-quality and diverse responses. Experimental results on both Llama and Qwen show the effectiveness of LongMab by achieving more than a 4% improvement on long-context reasoning benchmarks. All data and code will be released on https://github.com/NEUIR/LongMab-PO. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_13993 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Chunks as Arms: Multi-Armed Bandit-Guided Sampling for Long-Context LLM Preference Optimization Duan, Shaohua Huang, Pengcheng Li, Xinze Liu, Zhenghao Yi, Xiaoyuan Yan, Yukun Wang, Shuo Gu, Yu Yu, Ge Sun, Maosong Computation and Language Artificial Intelligence Long-context modeling is critical for a wide range of real-world tasks, including long-context question answering, summarization, and complex reasoning tasks. Recent studies have explored fine-tuning Large Language Models (LLMs) with synthetic data to enhance their long-context capabilities. However, the effectiveness of such approaches is often limited by the low diversity and factual inconsistencies in the generated data. To address these challenges, we propose LongMab, a novel framework that leverages a Multi-Armed Bandit (MAB) rollout strategy to identify the most informative chunks from the given long context for sampling high-quality and diverse responses and constructing preference data pairs for Direct Preference Optimization (DPO) training. Specifically, we treat context chunks as arms of MAB, select chunks based on their expected reward scores to input into LLMs to generate responses, and iteratively update these scores based on reward feedback. Both exploration and exploitation during the rollout process enable the LLM to focus on the most relevant context segments, thereby generating and collecting high-quality and diverse responses. Experimental results on both Llama and Qwen show the effectiveness of LongMab by achieving more than a 4% improvement on long-context reasoning benchmarks. All data and code will be released on https://github.com/NEUIR/LongMab-PO. |
| title | Chunks as Arms: Multi-Armed Bandit-Guided Sampling for Long-Context LLM Preference Optimization |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2508.13993 |