Upload README.md

README.md

@@ -174,6 +174,8 @@ Half of the samples was generated by this model where prompts contained the adve
 
 [KTO](https://arxiv.org/abs/2402.01306) trainer from [Hugging Face TRL library](https://huggingface.co/docs/trl/en/kto_trainer) was employed for performing preference alignment. The LoRA adapter from the previous training stages was merged into the model, and a new LoRA adapter was created for the KTO training. The quantized base model serves as a reference.
 
+During the alignment, the model was encouraged to respect player's actions and agency, construct a coherent narrative, and use evocative language to describe the world and the outcome of the player's actions.
+
 #### QLoRa adapter configuration
 
 - Rank: 16
@@ -210,7 +212,7 @@ The model's performance in Adventure Mode has improved substantially. The writin
 ![Gradient Norm](img/kto_grad_norm.png)
 ![Learning rate](img/kto_learning_rate.png)
 ![Rewards](img/kto_train_rewards.png)
-![Log probabilities](img/train_logps.png)
+![Log probabilities](img/kto_train_logps.png)
 ![KL divergence](img/kto_train_kl_divergence.png)