# Chain-based Distillation for Effective Initialization of Variable-Sized Small Language Models

arXiv: https://arxiv.org/abs/2605.07783 · HTML: https://mlsanity.com/paper/2605.07783/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 31 of its 40 references that are in the map.
- First posted on arXiv 8 May 2026.

## Builds on (most cited first, 30 of 31)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [Self-Instruct: Aligning Language Models with Self-Generated Instructions](https://mlsanity.com/paper/2212.10560/) (2022)
- [BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions](https://mlsanity.com/paper/1905.10044/) (2019)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling](https://mlsanity.com/paper/2304.01373/) (2023)
- [Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization](https://mlsanity.com/paper/1808.08745/) (2018)
- [On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes](https://mlsanity.com/paper/2306.13649/) (2023)
- [Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes](https://mlsanity.com/paper/2305.02301/) (2023)
- [Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor](https://mlsanity.com/paper/2212.09689/) (2022)
- [Compact Language Models via Pruning and Knowledge Distillation](https://mlsanity.com/paper/2407.14679/) (2024)
- [SCOTT: Self-Consistent Chain-of-Thought Distillation](https://mlsanity.com/paper/2305.01879/) (2023)
- [Less is More: Task-aware Layer-wise Distillation for Language Model Compression](https://mlsanity.com/paper/2210.01351/) (2022)
- [DISC-FinLLM: A Chinese Financial Large Language Model based on Multiple Experts Fine-tuning](https://mlsanity.com/paper/2310.15205/) (2023)
- [Towards the Law of Capacity Gap in Distilling Language Models](https://mlsanity.com/paper/2311.07052/) (2023)
- [Scaling Smart: Accelerating Large Language Model Pre-training with Small Model Initialization](https://mlsanity.com/paper/2409.12903/) (2024)
- [Learngene: From Open-World to Your Learning Task](https://mlsanity.com/paper/2106.06788/) (2021)
- [Learngene: Inheriting Condensed Knowledge from the Ancestry Model to Descendant Models](https://mlsanity.com/paper/2305.02279/) (2023)
- [Large Language Model Meets Graph Neural Network in Knowledge Distillation](https://mlsanity.com/paper/2402.05894/) (2024)
- [Transformer as Linear Expansion of Learngene](https://mlsanity.com/paper/2312.05614/) (2023)
- [H2O-Danube-1.8B Technical Report](https://mlsanity.com/paper/2401.16818/) (2024)
- [Building Variable-sized Models via Learngene Pool](https://mlsanity.com/paper/2312.05743/) (2023)
- [In-Context Learning Distillation for Efficient Few-Shot Fine-Tuning](https://mlsanity.com/paper/2412.13243/) (2024)
- [Scaling Up On-Device LLMs via Active-Weight Swapping Between DRAM and Flash](https://mlsanity.com/paper/2504.08378/) (2025)
- [MiniRBT: A Two-stage Distilled Small Chinese Pre-trained Model](https://mlsanity.com/paper/2304.00717/) (2023)
- [Chain-of-Model Learning for Language Model](https://mlsanity.com/paper/2505.11820/) (2025)

## Abstract (arXiv)

Large language models (LLMs) achieve strong performance but remain costly to deploy in resource-constrained settings. Training small language models (SLMs) from scratch is computationally expensive, while conventional knowledge distillation requires repeated access to large teachers for different target sizes, leading to poor scalability. To solve these problems, we propose Chain-based Distillation (CBD), a scalable paradigm for efficiently initializing variable-sized language models. A sparse and limited sequence of intermediate models (called anchors) is constructed via stepwise distillation, forming a distillation chain that progressively transfers knowledge from the source LLMs. To support heterogeneous settings, we introduce bridge distillation for cross-architecture and cross-vocabulary transfer. Models of variable sizes are initialized via parameter interpolation between adjacent anchors, eliminating repeated large teacher inference. Experiments show that the proposed method substantially improves efficiency and downstream performance. A 138M-parameter SLM without recovery pre-training, outperforms scratch-trained models on a 10B-token corpus on the specific task. CBD also demonstrates versatility in heterogeneous settings for initialize models with different architectures and vocabularies.