Clarification on AI coding assistants and Tournament Dataset access

Competition ADIA Lab
1

Section 5 of the Tournament Rules states that the Tournament Dataset must not be made available to anyone who is not an official participant.

Could you please clarify whether a participant may use a remote AI coding assistant, such as Codex or ChatGPT, if the service processes the participant-accessible training or reduced test data solely to develop, debug, and validate the participant’s submission?

If this is permitted, are any specific privacy, retention, or account settings required? If it is not permitted, should AI coding assistants be limited to public documentation, public code, and synthetic data?

Thank you.

2

Hi Kenny,

In all reasonable uses of LLMs, there is no sharing of the dataset: unless one puts the dataset in the context of the LLM, the LLM does not see the dataset, but just numeric or graphical summaries of it.