Skip to content

What command and llm_prompt_type setting should be used for ScenicNL? #47

@hanquanjushi

Description

@hanquanjushi

In the paper https://arxiv.org/abs/2405.03709, the table shows that ScenicNL achieves the best performance. However, when using the gen_scenic tool, I’m not sure which command corresponds to ScenicNL. Specifically, I’m unclear about which option to select for the llm_prompt_type parameter.

Based on the gen_scenic --help output, the available options for llm_prompt_type are as follows:

--llm_prompt_type [predict_zero_shot|predict_few_shot|predict_scenic_tutorial|predict_python_api|predict_python_api_oneline|predict_lmql|predict_few_shot_with_rag|predict_few_shot_hyde|predict_few_shot_hyde_tot|predict_tot_then_hyde|predict_tot_then_split|predict_tot_into_nl|expert_discussion|expert_synthesis|ast_feedback|predict_lmql_to_hyde|predict_lmql_retry|predict_lmql_tot_retry]

Unfortunately, I couldn’t find a clear explanation in the documentation about which option corresponds to ScenicNL. Could you please clarify the following:

What is the specific command to use for ScenicNL?
Which value should be set for llm_prompt_type to correspond to ScenicNL?
If there are any additional parameters or configurations required, could you please provide details?
Thank you for your assistance!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions