Nanbeige4-3B Technical Report: Exploring the Frontier of Small Language Models

• URL: http://arxiv.org/abs/2512.06266v1
• Date: Sat, 06 Dec 2025 03:36:27 GMT
• Title: Nanbeige4-3B Technical Report: Exploring the Frontier of Small Language Models
• Authors: Chen Yang, Guangyue Peng, Jiaying Zhu, Ran Le, Ruixiang Feng, Tao Zhang, Wei Ruan, Xiaoqi Liu, Xiaoxue Cheng, Xiyun Xu, Yang Song, Yanzipeng Gao, Yiming Jia, Yun Xing, Yuntao Wen, Zekai Wang, Zhenwei An, Zhicong Sun, Zongchao Chen,
• Abstract summary: Nanbeige4-3B is a family of small-scale but high-performing language models.<n>Pretrained on 23T high-quality tokens and finetuned on over 30 million diverse instructions, we extend the boundary of the scaling law for small language models.
• Score: 23.832817775138675
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We present Nanbeige4-3B, a family of small-scale but high-performing language models. Pretrained on 23T high-quality tokens and finetuned on over 30 million diverse instructions, we extend the boundary of the scaling law for small language models. In pre-training, we design a Fine-Grained Warmup-Stable-Decay (FG-WSD) training scheduler, which progressively refines data mixtures across stages to boost model performance. In post-training, to improve the quality of the SFT data, we design a joint mechanism that integrates deliberative generation refinement and chain-of-thought reconstruction, yielding substantial gains on complex tasks. Following SFT, we employ our flagship reasoning model to distill Nanbeige4-3B through our proposed Dual Preference Distillation (DPD) method, which leads to further performance gains. Finally, a multi-stage reinforcement learning phase was applied, leveraging verifiable rewards and preference modeling to strengthen abilities on both reasoning and human alignment. Extensive evaluations show that Nanbeige4-3B not only significantly outperforms models of comparable parameter scale but also rivals much larger models across a wide range of benchmarks. The model checkpoints are available at https://huggingface.co/Nanbeige.

Related papers

• Scaling Transformers for Discriminative Recommendation via Generative Pretraining [15.796591192359044]
  We propose a framework named GPSD (textbfGenerative textbfPretraining for textbfScalable textbfDiscriminative Recommendation) to address the overfitting issue.<n>Extensive experiments conducted on both industrial-scale and publicly available datasets demonstrate the superior performance of GPSD.
  arXiv  Detail & Related papers  (2025-06-04T08:31:33Z)
• HAD: Hybrid Architecture Distillation Outperforms Teacher in Genomic Sequence Modeling [52.58723853697152]
  We propose a Hybrid Architecture Distillation (HAD) approach for DNA sequence modeling.<n>We employ the NTv2-500M as the teacher model and devise a grouping masking strategy.<n>Compared to models with similar parameters, our model achieved excellent performance.
  arXiv  Detail & Related papers  (2025-05-27T07:57:35Z)
• Long-Short Chain-of-Thought Mixture Supervised Fine-Tuning Eliciting Efficient Reasoning in Large Language Models [23.34070841541423]
  We propose Long-Short Chain-of-Thought Mixture Supervised Fine-Tuning (LS-Mixture SFT)<n>Our experiments demonstrate that models trained using LS-Mixture SFT, compared to those trained with direct SFT, achieved an average accuracy improvement of 2.3%.<n>This work offers an approach to endow non-reasoning models with reasoning capabilities through supervised fine-tuning.
  arXiv  Detail & Related papers  (2025-05-06T12:18:11Z)
• Overtrained Language Models Are Harder to Fine-Tune [64.44743256512237]
  Large language models are pre-trained on ever-growing token budgets.<n>We show that extended pre-training can make models harder to fine-tune, leading to degraded final performance.
  arXiv  Detail & Related papers  (2025-03-24T23:11:56Z)
• SaRA: High-Efficient Diffusion Model Fine-tuning with Progressive Sparse Low-Rank Adaptation [52.6922833948127]
  In this work, we investigate the importance of parameters in pre-trained diffusion models.<n>We propose a novel model fine-tuning method to make full use of these ineffective parameters.<n>Our method enhances the generative capabilities of pre-trained models in downstream applications.
  arXiv  Detail & Related papers  (2024-09-10T16:44:47Z)
• FD-Align: Feature Discrimination Alignment for Fine-tuning Pre-Trained Models in Few-Shot Learning [21.693779973263172]
  In this paper, we introduce a fine-tuning approach termed Feature Discrimination Alignment (FD-Align) Our method aims to bolster the model's generalizability by preserving the consistency of spurious features. Once fine-tuned, the model can seamlessly integrate with existing methods, leading to performance improvements.
  arXiv  Detail & Related papers  (2023-10-23T17:12:01Z)
• Improving Non-autoregressive Translation Quality with Pretrained Language Model, Embedding Distillation and Upsampling Strategy for CTC [51.34222224728979]
  This paper introduces a series of innovative techniques to enhance the translation quality of Non-Autoregressive Translation (NAT) models. We propose fine-tuning Pretrained Multilingual Language Models (PMLMs) with the CTC loss to train NAT models effectively. Our model exhibits a remarkable speed improvement of 16.35 times compared to the autoregressive model.
  arXiv  Detail & Related papers  (2023-06-10T05:24:29Z)
• Mixture-of-Experts Meets Instruction Tuning:A Winning Combination for Large Language Models [125.91897197446379]
  We find that MoE models benefit more from instruction tuning than dense models. Our most powerful model, FLAN-MOE-32B, surpasses the performance of FLAN-PALM-62B on four benchmark tasks.
  arXiv  Detail & Related papers  (2023-05-24T04:22:26Z)
• METRO: Efficient Denoising Pretraining of Large Scale Autoencoding Language Models with Model Generated Signals [151.3601429216877]
  We present an efficient method of pretraining large-scale autoencoding language models using training signals generated by an auxiliary model. We propose a recipe, namely "Model generated dEnoising TRaining Objective" (METRO) The resultant models, METRO-LM, consisting of up to 5.4 billion parameters, achieve new state-of-the-art on the GLUE, SuperGLUE, and SQuAD benchmarks.
  arXiv  Detail & Related papers  (2022-04-13T21:39:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.