10亿参数模型仅用合规数据训练，英丹语表现达前沿水平
DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data
- 基于HRM架构从零训练，仅使用可接受的后训练数据
- 在20项评测中超越原版HRM-Text 1B，媲美更大模型
- 适合关注开源与伦理数据的研究者使用
当前大语言模型开发依赖海量且常不可接受的数据集，为坚持开源与道德数据的研究者带来高门槛。我们提出Mimir v1，一个基于层级推理模型（HRM）架构的10亿参数语言模型，从头训练，仅使用合规的后训练数据，在英语上实现极具竞争力的表现，并在丹麦语上创下新基准。该模型在161个数据集混合训练下，于20项英文、数学与代码以及丹麦语评测中，性能超越原始HRM-Text 1B，媲美更大规模的前沿模型如Qwen 3.5 4B和Gemma 4 E2B。模型已公开发布于Hugging Face Hub：https://huggingface.co/danish-foundation-models/DFM-Mimir
原文摘要 · Abstract (English)
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
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。