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

Prioritize Bash if LLM responses both Bash and Conclusion #716

Merged
merged 1 commit into from
Nov 11, 2024

Conversation

DonggeLiu
Copy link
Collaborator

Previously, when the LLM response included both bash commands and a conclusion, we ignored the bash command and compiled the fuzz target in the conclusion alone. This approach is suboptimal, as conclusions may be premature, and the bash command often provides useful context to refine them.

To improve accuracy, we should prioritize the bash command when both are present in a response.

A more robust solution would involve structuring tasks to prevent mixed responses (e.g., using task-focused agents). But until then, prioritizing bash commands in mixed responses provides a fallback/default solution.

@DonggeLiu
Copy link
Collaborator Author

/gcbrun exp -n dg -ag

@DonggeLiu
Copy link
Collaborator Author

@DonggeLiu
Copy link
Collaborator Author

Thanks @mihaimaruseac : )
I plan to break these steps into individual agents in the next step, do you have any advices/suggestions?

@mihaimaruseac
Copy link
Member

Oh, that's great, the prompt is quite long right now and I think that splitting it over multiple agents might bring more benefits.

I think one best rule of thumb is to have each agent use only one tool. So we should not have an agent use both bash and have access to the compiler, for example

@DonggeLiu DonggeLiu merged commit 6ae19df into main Nov 11, 2024
5 checks passed
@DonggeLiu DonggeLiu deleted the agents-prioritize-bash branch November 11, 2024 21:34
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

Successfully merging this pull request may close these issues.

3 participants