GenerTeam/sequence-recovery

Next K-mer Prediction Abouts The Next K-mer Prediction task is a zero-shot evaluation method introduced in the GENERator paper to assess the quality of pretrained models. It involves inputting a sequence segment into the model and having it predict the next K base pairs. The predicted sequence is then compared to the actual sequence to assess accuracy. Sequence: The input sequence has a maximum length of 96k base pairs (bp). You can control the number of input… See the full description on the dataset page: https://huggingface.co/datasets/GenerTeam/sequence-recovery.