fix genqa eval normlization bug, add more task#3
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: aa968d725a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| def pubmedqa_cf_prompt(line, task_name: str = None): | ||
| """CF variant: score full answer text (yes/no/maybe) via logprobs.""" | ||
| ctx = " ".join(line["CONTEXTS"]) if isinstance(line["CONTEXTS"], list) else line["CONTEXTS"] |
There was a problem hiding this comment.
After switching the configs below to qiaojin/PubMedQA, the prompt functions still use the old pubmed_qa column names. The new dataset rows expose lowercase question and a context dict with contexts, not QUESTION/CONTEXTS, so selecting pubmedqa:cf, pubmedqa:mcf, or pubmedqa:mcf_em will raise KeyError before any examples are scored. Either keep the old dataset or update both prompt functions to read the new schema.
Useful? React with 👍 / 👎.
1 participant
No description provided.