Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

About the dormant threshold setting in the ReDO paper #226

Open
ZifanWu opened this issue Nov 15, 2024 · 2 comments
Open

About the dormant threshold setting in the ReDO paper #226

ZifanWu opened this issue Nov 15, 2024 · 2 comments

Comments

@ZifanWu
Copy link

ZifanWu commented Nov 15, 2024

Hello there,

There seems to be an inconsistency on the default dormant threshold setting ($\tau$ in the paper) between the main paper and the appendix:

In the main paper (Page 5, Section 5), it says:

For agents trained with ReDo, we use a threshold of τ = 0.1, unless otherwise noted, as we found this gave a better performance than using a threshold of 0 or 0.025.

In the appendix (Table 1), it says:

0.025 for default setting, 0.1 otherwise

Could you please specify the best setting for this hyperparameter? Thanks!

@GhadaSokar
Copy link

Hello,

Thank you for your question.

We used τ = 0.1 for all settings except in the default where we use τ = 0.025.

@ZifanWu
Copy link
Author

ZifanWu commented Nov 19, 2024

Understood. Thank you for the reply. If I understand correctly, the default setting (0.025) is for demonstrating the dormant phenomenon and 0.1, under which the perf is better, is for evaluation?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants