I am attempting to reproduce the transfer learning from HotpotQA to FEVER (Section 5.4 of the paper). In this process, I have encountered several difficulties:
(1) How exactly are the insights from HotpotQA transferred to FEVER? The paper states “We use gpt-4-0613 for adapting the HotpotQA insights into FEVER insights”, but I could not find the corresponding prompt or any released code.
(2) It seems that the implementation of the FEVER environment in the code has issues, and may not be consistent with HotpotQA. Specifically, it cannot return the search results:
I am attempting to reproduce the transfer learning from HotpotQA to FEVER (Section 5.4 of the paper). In this process, I have encountered several difficulties:
(1) How exactly are the insights from HotpotQA transferred to FEVER? The paper states “We use gpt-4-0613 for adapting the HotpotQA insights into FEVER insights”, but I could not find the corresponding prompt or any released code.
(2) It seems that the implementation of the FEVER environment in the code has issues, and may not be consistent with HotpotQA. Specifically, it cannot return the search results: