You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The base model needs to be fine-tuned to be able to perform with at least average quality on one of the four competition’s tasks (Find Cave, Build Waterfall, etc.) [4]. Average Quality means, here, that the agent should at least do something which somehow resembles what the actual task looks like instead of doing completely random actions. This is important for collecting feedback from a human later, since giving feedback on completely random trajectories would not lead to much information gain.
To fine-tune the base model, we use some of the provided demonstrations
Task:
Train four models, each, on the officially provided demonstrations:
Find out how many demonstrations are needed to reach an average quality. Do we need more (e.g. MineDojo https://minedojo.org/knowledge_base)?
Can we further improve the quality by defining Subtasks --> See own Issue for that
The text was updated successfully, but these errors were encountered:
The base model needs to be fine-tuned to be able to perform with at least average quality on one of the four competition’s tasks (Find Cave, Build Waterfall, etc.) [4]. Average Quality means, here, that the agent should at least do something which somehow resembles what the actual task looks like instead of doing completely random actions. This is important for collecting feedback from a human later, since giving feedback on completely random trajectories would not lead to much information gain.
To fine-tune the base model, we use some of the provided demonstrations
Task:
Train four models, each, on the officially provided demonstrations:
Find out how many demonstrations are needed to reach an average quality. Do we need more (e.g. MineDojo https://minedojo.org/knowledge_base)?
Can we further improve the quality by defining Subtasks --> See own Issue for that
The text was updated successfully, but these errors were encountered: