questions about data generation

[lollipopwsy]

Hello, Authors! 
Do I just need to run ./gene_data.sh in the llm-mcts/scripts directory to generate data?
Does the commented-out code need to be run? @1989Ryan I am waiting for your response. Thank you!

[1989Ryan]

Please refer to https://github.com/1989Ryan/llm-mcts?tab=readme-ov-file#generate-data to uncomment the command that generate the expert data you need.

[lollipopwsy]

But when I check the script, there are only simple tasks, full tasks, unseen-apartment simple and full tasks, unseen-item simple and full tasks, unseen_comp unseen-item full tasks, and unseen_comp full tasks.
How do these correspond to the seen and unseen simple, comp, novelsimple, novelcomp2, novelcomp3 mentioned in the paper?

@1989Ryan

[1989Ryan]

Basically, you can map unseen_comp -> novelcomp and full -> comp and combine them. Those flags are for different purposes and might overlap in definition. I used the term in the Watch and Help code, which differs from our paper.