File size: 2,486 Bytes
f17846d c90f821 f17846d 5be2bff f17846d c90f821 f17846d fb497b5 5be2bff f17846d c90f821 f17846d 5be2bff f17846d 5be2bff | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 | ---
license: apple-amlr
base_model:
- Qwen/Qwen3-4B-Instruct-2507
tags:
- self-distillation
- code-generation
- ssd
library_name: transformers
---
# SimpleSD-4B-instruct
This model was produced using **Simple Self-Distillation (SSD)**, a method that improves code generation by fine-tuning a language model on its own sampled outputs—without rewards, verifiers, teacher models, or reinforcement learning.
- **Self-distillation sampling:** temperature=1.6, top_p=0.8, top_k=20
- **Evaluation sampling:** temperature=1.1, top_p=0.8, top_k=20
## Notes
- These are research checkpoints for reproducibility.
- They are not optimized Qwen releases.
- They don't represent a broader open-source model strategy.
## Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("apple/SimpleSD-4B-instruct")
tokenizer = AutoTokenizer.from_pretrained("apple/SimpleSD-4B-instruct")
```
## Method
SSD samples solutions from the base model using non-unit temperature and top-k/top-p truncation, then fine-tunes on those samples via standard supervised learning. Despite its simplicity, SSD yields large gains on competitive programming benchmarks, with improvements concentrating on harder problems. The mechanism traces to resolving a *precision–exploration conflict*: SSD reshapes token distributions in a context-dependent way so that a single global decoding configuration becomes far more effective at evaluation time.
## Results
LiveCodeBench (%)
| Model | LCBv6 pass@1 | LCBv6 pass@5 | LCBv5 pass@1 | LCBv5 pass@5 |
|---|---|---|---|---|
| Qwen3-4B-Instruct-2507 (base) | 34.0 | 41.0 | 34.3 | 45.4 |
| **+ SSD (this model)** | **41.5** (+7.5) | **56.8** (+15.8) | **45.7** (+11.4) | **61.9** (+16.5) |
## Paper
[**Embarrassingly Simple Self-Distillation Improves Code Generation**](https://arxiv.org/abs/2604.01193)
```bibtex
@misc{zhang2026embarrassinglysimpleselfdistillationimproves,
title={Embarrassingly Simple Self-Distillation Improves Code Generation},
author={Ruixiang Zhang and Richard He Bai and Huangjie Zheng and Navdeep Jaitly and Ronan Collobert and Yizhe Zhang},
year={2026},
eprint={2604.01193},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2604.01193},
}
```
## License
This model is released under the [Apple Machine Learning Research Model License](https://huggingface.co/apple/SimpleSD-4B-instruct/blob/main/LICENSE).
|