OpenHermesPreferences: Dataset of ~1M AI preferences from teknium/OpenHermes-2.5 (opens in new tab)

(huggingface.co)

7 pointskashifr2y ago1 comments

1 comments

The dataset can be used for training preference models or aligning language models through techniques like Direct Preference Optimization (DPO).

j / k navigate · click thread line to collapse

kashifrOP2y ago

The dataset can be used for training preference models or aligning language models through techniques like Direct Preference Optimization (DPO).

j / k navigate · click thread line to collapse