Update README.md

README.md

@@ -15,6 +15,8 @@ We have developed and released the family [Ichigo-llama3s](https://huggingface.c
 We SFT [homebrewltd/Ichigo-llama3.1-s-base-v0.3](https://huggingface.co/homebrewltd/Ichigo-llama3.1-s-base-v0.3) with nearly 1B tokens from [Instruction Speech WhisperVQ v3](homebrewltd/mixed-instruction-speech-whispervq-v3-full) dataset.
 This is the model checkpoint from step 7000. Due to some noise in the training data, it has an artificially higher score on the Speech Instruction benchmark.
 
+This model is a supervised fine-tuned (SFT) version of homebrewltd/Ichigo-llama3.1-s-base-v0.3, trained on over 1 billion tokens from the [Instruction Speech WhisperVQ v4](jan-hq/mixed-instruction-speech-whispervq-v3-full-phase2-3) dataset which built upon [Instruction Speech WhisperVQ v3](homebrewltd/mixed-instruction-speech-whispervq-v3-full), adding multi-turn speech conversations and noise rejection capabilities for enhanced performance. This version, we introduce of noise-augmented multi-turn conversations, where we synthetically inject noise into both speech and text-only dialogue data. As a result, the model demonstrates improved robustness against noisy environmental inputs and enhanced multi-turn conversation capabilities, making it more reliable in real-world applications.
+
 **Model developers** Homebrew Research.
 
 **Input** Text and sound.