Dear Zigeng,
Thank you for sharing this interesting idea and open-source code. I noticed during your ICLR rebuttal, you extended the experiments to general-purpose language tasks (such as MMLU, HellaSwag, ARC-C, PIQA, GPQA-CoT, and QASPER). Would it be possible for you to also release the evaluation code for these tasks? Thank you!
Dear Zigeng,
Thank you for sharing this interesting idea and open-source code. I noticed during your ICLR rebuttal, you extended the experiments to general-purpose language tasks (such as MMLU, HellaSwag, ARC-C, PIQA, GPQA-CoT, and QASPER). Would it be possible for you to also release the evaluation code for these tasks? Thank you!