DFM Mimir v1：仅使用许可的后训练数据，在10亿参数规模下实现前沿性能的开放HRM模型
DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data
AI总结：
研究人员因大模型开发依赖非许可数据受阻，本文提出仅用许可后训练数据训练的10亿参数HRM模型Mimir v1，其在多基准测试中性能优于同规模HRM-Text 1B，可与更大前沿模型媲美，且在丹麦语任务达新SOTA，已发布于Hugging Face Hub
AI中文摘要：
当前大语言模型的开发依赖于海量且往往不被许可的数据集，这为致力于开源和道德数据源的研究人员设置了很高的门槛。我们推出Mimir v1，这是一款基于分层推理模型（Hierarchical Reasoning Model，HRM）架构的10亿参数语言模型，从零开始训练，仅使用许可的后训练数据，在英语任务上展现出极具竞争力的性能，并在丹麦语任务上创下新的最优水平。该模型在161个数据集的混合数据集上训练，在20个涵盖英语、数学与代码及丹麦语的基准测试中，Mimir v1的性能优于原始HRM-Text 1B，且可与Qwen 3.5 4B、Gemma 4 E2B等更大规模的前沿模型相媲美。该模型已在Hugging Face Hub上发布：this https URL
英文摘要：
Current large language model development relies on massive, often non-permissible datasets, creating a high barrier for researchers committed to open-source and ethically sourced data. We introduce Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture, that is trained from scratch and delivers highly competitive performance for English and sets a new state of the art for Danish using only permissible post-training data. Trained on a mixture of 161 datasets, Mimir v1 outperforms the original HRM-Text 1B and competes with larger frontier models like Qwen 3.5 4B and Gemma 4 E2B, tested across 20 benchmarks for English, Math & Code and Danish. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir
