I feel that the paper is really pretty good. May I know which conference I am currently submitting to?
LeiyanGithub opened this issue · 10 comments
Tanks for your attention to our paper, this paper is now under review of ARR.
Is it because the EMNLP application was not accepted?
Yes 😭
So sad, but we think this is a pretty good job. Give me some insights! What aspects do the reviewers consider to be inadequate? Is it because the initial version experiments were not complete enough, not a more complete version now?
Do you plan to submit to ACL？
We will decide based on the subsequent reviews' comments.
I am a little curious why you chose Lora instead of fully fine-tuning？
We primarily consider the computational resources available and the stability of training.
Hello, may I ask how to extract answers from the model's response in zero-shot situations, like vicuna?
For example:
{"id": "SVAMP11", "instruction": "Given a math problem, please solve it", "input": "Winter is almost here and most animals are migrating to warmer countries. 27 bird families flew away for the winter from near a mountain. If there were 41 bird families living near the mountain initially How many bird families were left near the mountain?", "answer": "14.0", "response": "There were 41 bird families living near the mountain initially."}
We directly extract the last number in the last sentence of the response.