descriptions.py

LLM_BLENDER_OVERALL_DESC = """
LLM-Blender is an innovative ensembling framework to attain consistently superior performance by leveraging the diverse strengths of multiple open-source large language models (LLMs). 
LLM-Blender cut the weaknesses through ranking and integrate the strengths through fusing generation to enhance the capability of LLMs.

Our framework consists of two complementary modules: **PairRanker** and **GenFuser**, 
addressing the observation that optimal LLMs for different examples can significantly vary. 
**PairRanker** employs a specialized pairwise comparison method to distinguish subtle differences between candidate outputs. 
**GenFuser** aims to merge the top-ranked candidates from the aggregation of PairRanker's pairwise 
comparisons into an improved output by capitalizing on their strengths and mitigating their weaknesses.

| [Paper](https://arxiv.org/abs/2306.02561) 
| [Code](https://github.com/yuchenlin/LLM-Blender) 
| [Dataset](https://huggingface.co/datasets/llm-blender/mix-instruct)
| [Models](https://huggingface.co/llm-blender)
| [Tweet](https://twitter.com/billyuchenlin/status/1668666357058277377)

Try LLM-Blender now! 👇
"""

PairRM_OVERALL_DESC = """## 🤗 [PairRM](https://huggingface.co/llm-blender/PairRM)

PairRM is a reward model based on PairRanker architecture that has been trained on various high-quality and 
large-scale dataset with human preference annotations and exhibits great correlation with human preferences.

**While PairRM is a extremely small model (0.4B), our tests on various human alignment benchmarks show approaching performance of GPT-4.** (See [PairRM](https://huggingface.co/llm-blender/PairRM) for more detail results)

PairRM could be easily applied to 3 scenarios:
1. [Directly compare two responses or two conversations](https://huggingface.co/llm-blender/PairRM#use-case-1-compare-responses-quality-evaluator)
2. [Do best-of-n sampling to enhancing the ability of LLM](https://huggingface.co/llm-blender/PairRM#use-case-2-best-of-n-sampling-decoding-enhancing)
3. [RLHF alignment](https://huggingface.co/llm-blender/PairRM#use-case-3-rlhf)

This demo allows user to interact with PairRM in the first scenario. Feel free to compare the quality of two responses or two conversations by inputting them in the text box below.

Try PairRM now! 👇
"""

CITATION = """

## Citation
```bibtex
@inproceedings{llm-blender-2023,
    title = "LLM-Blender: Ensembling Large Language Models with Pairwise Comparison and Generative Fusion",
    author = "Jiang, Dongfu and Ren, Xiang and Lin, Bill Yuchen",
    booktitle = "Proceedings of the 61th Annual Meeting of the Association for Computational Linguistics (ACL 2023)",
    year = "2023"
}

```
"""