$π^2$: Structure-Originated Reasoning Data Improves Long-Context Reasoning Ability of Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Do, Quyet V., Pham, Thinh, Nguyen, Nguyen, Li, Sha, Zunjare, Pratibha, Vu, Tu
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914450674548736
author Do, Quyet V.
Pham, Thinh
Nguyen, Nguyen
Li, Sha
Zunjare, Pratibha
Vu, Tu
author_facet Do, Quyet V.
Pham, Thinh
Nguyen, Nguyen
Li, Sha
Zunjare, Pratibha
Vu, Tu
contents We study a pipeline that curates reasoning data from initial structured data for improving long-context reasoning in large language models (LLMs). Our approach, $π^2$, constructs high-quality reasoning data through rigorous QA curation: 1) extracting and expanding tables from Wikipedia, 2) from the collected tables and relevant context, generating realistic and multi-hop analytical reasoning questions whose answers are automatically determined and verified through dual-path code execution, and 3) back-translating step-by-step structured reasoning traces as solutions of QA pairs given realistic web-search context. Supervised fine-tuning with \textsc{\small{gpt-oss-20b}} and \textsc{\small{Qwen3-4B-Instruct-2507}} on $π^2$ yields consistent improvements across four long-context reasoning benchmarks and our alike $π^2$-Bench, with average absolute accuracy gains of +4.3% and +2.7% respectively. Notably, our dataset facilitates self-distillation, where \textsc{\small{gpt-oss-20b}} even improves its average performance by +4.4% with its own reasoning traces, demonstrating $π^2$'s usefulness. Our code, data, and models are open-source at https://github.com/vt-pi-squared/pi-squared.
format Preprint
id arxiv_https___arxiv_org_abs_2604_05114
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle $π^2$: Structure-Originated Reasoning Data Improves Long-Context Reasoning Ability of Large Language Models
Do, Quyet V.
Pham, Thinh
Nguyen, Nguyen
Li, Sha
Zunjare, Pratibha
Vu, Tu
Computation and Language
Artificial Intelligence
Machine Learning
We study a pipeline that curates reasoning data from initial structured data for improving long-context reasoning in large language models (LLMs). Our approach, $π^2$, constructs high-quality reasoning data through rigorous QA curation: 1) extracting and expanding tables from Wikipedia, 2) from the collected tables and relevant context, generating realistic and multi-hop analytical reasoning questions whose answers are automatically determined and verified through dual-path code execution, and 3) back-translating step-by-step structured reasoning traces as solutions of QA pairs given realistic web-search context. Supervised fine-tuning with \textsc{\small{gpt-oss-20b}} and \textsc{\small{Qwen3-4B-Instruct-2507}} on $π^2$ yields consistent improvements across four long-context reasoning benchmarks and our alike $π^2$-Bench, with average absolute accuracy gains of +4.3% and +2.7% respectively. Notably, our dataset facilitates self-distillation, where \textsc{\small{gpt-oss-20b}} even improves its average performance by +4.4% with its own reasoning traces, demonstrating $π^2$'s usefulness. Our code, data, and models are open-source at https://github.com/vt-pi-squared/pi-squared.
title $π^2$: Structure-Originated Reasoning Data Improves Long-Context Reasoning Ability of Large Language Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2604.05114