# HyperCLOVA X THINK Technical Report

Source: [HyperCLOVA X THINK Technical Report](<https://www.sota2.com/research/paper/68fa79cc1e209131cea7f6e2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-27T06:56:50.936Z

Authors: NAVER Cloud HyperCLOVA X Team

Published: 2025-06-27T17:23:12.000Z

## Abstract

We introduce HyperCLOVA X THINK, the first reasoning-focused large language model in the HyperCLOVA X family, pre-trained on roughly $6$ trillion high-quality Korean, and English tokens, augmented with targeted synthetic Korean data. It was implemented as a compute-memory-balanced Peri-LN Transformer scaled with $\\mu$P, pre-trained through a three-stage curriculum that expands the context window to $128$K tokens, and post-trained via supervised fine-tuning with Reinforcement Learning from Verifiable Rewards supports both detailed rationale and concise-answer modes. It delivers competitive performance against similarly sized models on Korea-focused benchmarks such as KMMLU, CSAT, KoBALT-700, HAERAE-1.0, and KoBigBench, while preserving robust bilingual consistency and translation quality. In addition, a vision-augmented variant matches or exceeds GPT-4.1 on the KCSAT STEM benchmark, all of which are achieved with substantially lower training compute than existing models of similar sizes. We also present a pruning and distillation technique that will soon be applied to HyperCLOVA X THINK for an open-source and business-friendly foundation model. Altogether, these capabilities position HyperCLOVA X THINK as a robust foundation for Korean AI innovation and a valuable resource for the global research community.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Multi-Hop Question Answering on Multi-Hop QA Zero-Shot](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-multi-hop-qa-zero-shot.md>): Multi-hop Question Answering; Multi-Hop QA Zero-Shot; Accuracy (Zero-Shot): 65.13.

- [Multi-Hop Question Answering on Multi-Hop QA Full Context](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-multi-hop-qa-full-context.md>): Multi-hop Question Answering; Multi-Hop QA Full Context; Accuracy: 84.47.

- [Multi-Hop Question Answering on Multi-Hop QA Partial Context](<https://www.sota2.com/research/sota/multi-hop-question-answering-on-multi-hop-qa-partial-context.md>): Multi-hop Question Answering; Multi-Hop QA Partial Context; Accuracy: 42.97.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2506.22403>)

[PDF](<https://arxiv.org/pdf/2506.22403.pdf>)