English

When Thoughts Meet Facts: Reusable Reasoning for Long-Context LMs

Computation and Language 2026-04-29 v2 Artificial Intelligence Machine Learning

Abstract

Recent Long-Context Language Models (LCLMs) can process hundreds of thousands of tokens in a single prompt, enabling new opportunities for knowledge-intensive multi-hop reasoning by integrating large sets of retrieved documents or, in some cases, directly all necessary information. However, simply feeding more documents into the context window fails to capture how evidence should be connected. We address this gap with thought templates, which recast reasoning as reusable thought caches, derived from prior problem solving traces, structuring how evidence is combined and guiding multi-hop inference with factual documents. To keep these templates effective, we propose an update strategy that iteratively refines templates derived from training data through natural-language feedback. Across diverse benchmarks and LCLM families, our approach delivers consistent gains over strong baselines in both retrieval-based and retrieval-free settings. Furthermore, we show that optimized templates can be distilled into smaller open-source models, demonstrating its broad applicability and transparent reasoning reuse. We refer to our framework as Thought Template Augmented LCLMs (ToTAL).

Keywords

Cite

@article{arxiv.2510.07499,
  title  = {When Thoughts Meet Facts: Reusable Reasoning for Long-Context LMs},
  author = {Soyeong Jeong and Taehee Jung and Sung Ju Hwang and Joo-Kyung Kim and Dongyeop Kang},
  journal= {arXiv preprint arXiv:2510.07499},
  year   = {2026}
}

Comments

ACL Findings 2026

R2 v1 2026-07-01T06:25:09.158Z