Update README.md

README.md

@@ -12,13 +12,19 @@ base_model: moreh/MoMo-72B-lora-1.8.7-DPO
 
 Smaug arrives!
 
-We recently released Smaug-72B-v0.1 which has taken first place on the Open LLM Leaderboard by HuggingFace. It is the first open-source model to have an average score more than 80.
+We recently released Smaug-72B-v0.1 which has taken first place on the Open LLM Leaderboard by HuggingFace. It is the first open-source model to surpass an average score of 80%.
 
 Smaug-72B is finetuned directly from [moreh/MoMo-72B-lora-1.8.7-DPO](https://huggingface.co/moreh/MoMo-72B-lora-1.8.7-DPO) and is ultimately based on [Qwen-72B](https://huggingface.co/Qwen/Qwen-72B).
 
-To do so we built on techniques and datasets used in our previous model efforts, with the addition of some new datasets and a new approach. We believe this new approach is generally useful in training across a wide range of model types and downstream use cases, and it powers both our Smaug-34B and 72B models.
+We created Smaug-72B-v0.1 using a new fine-tuning technique, DPO-Positive (DPOP), and new pairwise preference versions of ARC, HellaSwag, and MetaMath (as well as other existing datasets). We introduce the technique and the full training details in our new paper: https://arxiv.org/abs/2402.13228.
 
-We are currently working on writing up this new technique in the form of a technical report which we aim to release on arXiv soon (we may also be releasing a new member of the Smaug lineup at that time!). We are excited to release the details to the open source community for them to build on and improve Smaug and spawn more dragons to dominate the LLM space. Keep watching this space for our announcements!
+We show that on datasets in which the edit distance between pairs of completions is low (such as in math-based datasets), standard DPO loss can lead to a reduction of the model's 
+likelihood of the preferred examples, as long as the relative probability between the preferred and dispreferred classes increases. 
+Using these insights, we design DPOP, a new loss function and training procedure which avoids this failure mode. 
+Surprisingly, we also find that DPOP outperforms DPO across a wide variety of datasets and downstream tasks, including datasets with high edit distances between completions.
+
+We believe this new approach is generally useful in training across a wide range of model types and downstream use cases, and it powers all of our Smaug models.
+With the release of our paper and datasets, we are excited for the open source community to continue to build on and improve Smaug and spawn more dragons to dominate the LLM space!
 
 
 ### Evaluation Results