Update README.md

README.md

@@ -1,8 +1,26 @@
 ---
 license: mit
+datasets:
+- llm-blender/mix-instruct
+metrics:
+- BERTScore
+- BLEURT
+- BARTScore
+- Pairwise Rank
+tags:
+- pair_ranker
+- reward_model
+- RLHF
 ---
 
-PairRanker used in llm-blender, trained on deberta-v3-large. 
+PairRanker used in llm-blender, trained on deberta-v3-large. This is the ranker model used in experiments in LLM-Blender paper, 
+which is trained on [mixinstruct](https://huggingface.co/datasets/llm-blender/mix-instruct) dataset for 5 epochs.
+
+|  PairRanker type  | Source max length | Candidate max length | Total max length |
+|:-----------------:|:-----------------:|----------------------|------------------|
+| [pair-ranker](https://huggingface.co/llm-blender/pair-ranker) (This model)              | 128               | 128                  | 384              |
+| [pair-reward-model](https://huggingface.co/llm-blender/pair-reward-model/) | 1224              | 412                  | 2048             |
+
 
 - Github: [https://github.com/yuchenlin/LLM-Blender](https://github.com/yuchenlin/LLM-Blender)
 - Paper: [https://arxiv.org/abs/2306.02561](https://arxiv.org/abs/2306.02561)
@@ -37,6 +55,4 @@ Then you are good to use pairrankers with
 - `blender.compare()` to compare 2 candiates.
 See LLM-Blender Github [README.md](https://github.com/yuchenlin/LLM-Blender#rank-and-fusion)
 and jupyter file [blender_usage.ipynb](https://github.com/yuchenlin/LLM-Blender/blob/main/blender_usage.ipynb)
-for detailed usage examples.
-
-
+for detailed usage examples.