---
language:
- en
license: apache-2.0
datasets:
- SenseLLM/ReflectionSeq-GPT
- SenseLLM/ReflectionSeq-DS
model-index:
- name: ReflectionCoder-DS-33B
  results:
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: IFEval (0-Shot)
      type: HuggingFaceH4/ifeval
      args:
        num_few_shot: 0
    metrics:
    - type: inst_level_strict_acc and prompt_level_strict_acc
      value: 37.87
      name: strict accuracy
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=SenseLLM/ReflectionCoder-DS-33B
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: BBH (3-Shot)
      type: BBH
      args:
        num_few_shot: 3
    metrics:
    - type: acc_norm
      value: 8.34
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=SenseLLM/ReflectionCoder-DS-33B
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MATH Lvl 5 (4-Shot)
      type: hendrycks/competition_math
      args:
        num_few_shot: 4
    metrics:
    - type: exact_match
      value: 2.04
      name: exact match
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=SenseLLM/ReflectionCoder-DS-33B
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GPQA (0-shot)
      type: Idavidrein/gpqa
      args:
        num_few_shot: 0
    metrics:
    - type: acc_norm
      value: 3.24
      name: acc_norm
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=SenseLLM/ReflectionCoder-DS-33B
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MuSR (0-shot)
      type: TAUR-Lab/MuSR
      args:
        num_few_shot: 0
    metrics:
    - type: acc_norm
      value: 0.46
      name: acc_norm
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=SenseLLM/ReflectionCoder-DS-33B
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU-PRO (5-shot)
      type: TIGER-Lab/MMLU-Pro
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 2.24
      name: accuracy
    source:
      url: https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=SenseLLM/ReflectionCoder-DS-33B
      name: Open LLM Leaderboard
---
## ReflectionCoder: Learning from Reflection Sequence for Enhanced One-off Code Generation

<p align="center">
    <a href="https://arxiv.org/abs/2405.17057">📄 Paper</a> •
    <a href="https://github.com/SenseLLM/ReflectionCoder">🏠 Repo</a> •
    <a href="https://huggingface.co/SenseLLM/ReflectionCoder-DS-33B">🤖 Models</a> •
    <a href="https://huggingface.co/datasets/SenseLLM/ReflectionSeq-GPT">📚 Datasets </a>
</p>

## Introduction
ReflectionCoder is a novel approach that effectively leverages reflection sequences constructed by integrating compiler feedback to improve one-off code generation performance. Please refer to our paper and repo for more details!

![](method.png)

<hr>

## Models

| Model | Checkpoint | Size | HumanEval (+) | MBPP (+) | License|
|:-------|:------------|:------|:---------------|:----------|:--------|
| ReflectionCoder-CL-7B   | 🤗 [HF Link](https://huggingface.co/SenseLLM/ReflectionCoder-CL-7B) | 7B   | 75.0 (68.9)     | 72.2 (61.4)     | [Llama2](https://ai.meta.com/llama/license/) |
| ReflectionCoder-CL-34B  | 🤗 [HF Link](https://huggingface.co/SenseLLM/ReflectionCoder-CL-34B) | 34B  | 70.7 (66.5)     | 68.4 (56.6)     | [Llama2](https://ai.meta.com/llama/license/) |
| ReflectionCoder-DS-6.7B | 🤗 [HF Link](https://huggingface.co/SenseLLM/ReflectionCoder-DS-6.7B) | 6.7B | 80.5 (74.4)     | 81.5 (69.6)     | [DeepSeek](https://github.com/deepseek-ai/DeepSeek-Coder/blob/main/LICENSE-MODEL) |
| ReflectionCoder-DS-33B  | 🤗 [HF Link](https://huggingface.co/SenseLLM/ReflectionCoder-DS-33B) | 33B  | 82.9 (76.8) | 84.1 (72.0) | [DeepSeek](https://github.com/deepseek-ai/DeepSeek-Coder/blob/main/LICENSE-MODEL) |

## Datasets

| Dataset           | Link           | License                                      |
|:-------------------|:----------------|:----------------------------------------------|
| ReflectionSeq-GPT | 🤗 [HF Link](https://huggingface.co/datasets/SenseLLM/ReflectionSeq-GPT) | [License](LICENSE) |
| ReflectionSeq-DS  | 🤗 [HF Link](https://huggingface.co/datasets/SenseLLM/ReflectionSeq-DS) | [License](LICENSE) |


## How to Use

#### Chat Format
Following chat templates of most models, we use two special tokens to wrap the message of user and assistant, *i.e.*, ``<|user|>``, ``<|assistant|>``, and ``<|endofmessage|>``. Furthermore, we use two special tokens to wrap the content of different blocks, *i.e.*,  ``<|text|>`` and ``<|endofblock|>``. You can use the following template to prompt our ReflectionCoder.

```python
import torch
from transformers import pipeline

chat = [
    {"role": "user", "content": "<Your code instruction here>"}
]

generator = pipeline(
    model="SenseLLM/ReflectionCoder-DS-33B",
    task="text-generation",
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

result = generator(chat, max_length=128, num_return_sequences=1)

print(result)
```

Please refer to our [GitHub Repo](https://github.com/SenseLLM/ReflectionCoder) for more technical details.

## Citation

If you find this repo useful for your research, please kindly cite our paper:
```
@misc{ren2024reflectioncoder,
    title={ReflectionCoder: Learning from Reflection Sequence for Enhanced One-off Code Generation}, 
    author={Houxing Ren and Mingjie Zhan and Zhongyuan Wu and Aojun Zhou and Junting Pan and Hongsheng Li},
    year={2024},
    eprint={2405.17057},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}
```

## Acknowledgments

We thank the following amazing projects that truly inspired us:

- [CodeLlama](https://ai.meta.com/research/publications/code-llama-open-foundation-models-for-code/)
- [DeepSeek-Coder](https://github.com/deepseek-ai/DeepSeek-Coder)
- [WizardCoder](https://github.com/nlpxucan/WizardLM/tree/main/WizardCoder)
- [Evol-CodeAlpaca-v1](https://huggingface.co/datasets/theblackcat102/evol-codealpaca-v1)
- [MagiCoder](https://github.com/ise-uiuc/magicoder/tree/main)
- [EvalPlus](https://github.com/evalplus/evalplus)
- [OpenCoderInterpreter](https://github.com/OpenCodeInterpreter/OpenCodeInterpreter/tree/main)
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_SenseLLM__ReflectionCoder-DS-33B)

|      Metric       |Value|
|-------------------|----:|
|Avg.               | 9.03|
|IFEval (0-Shot)    |37.87|
|BBH (3-Shot)       | 8.34|
|MATH Lvl 5 (4-Shot)| 2.04|
|GPQA (0-shot)      | 3.24|
|MuSR (0-shot)      | 0.46|
|MMLU-PRO (5-shot)  | 2.24|