search

wkwan / ScrimBrain

Reinforcement Learning in Fortnite With Real-Time Screen Capture and Windows Input Simulation
30 stars 1 forks source link

Use LLM’s/Multimodal Models to Generate a Paper for ScrimBrain Using the Repo, Tensorboard Log Files, and YT Video #1

Open wkwan opened 1 month ago

wkwan commented 1 month ago

Goal: To collaborate with more RL researchers on ScrimBrain

Problem: @wkwan writes like a zoomer

Solution: Generate a scientific paper for ScrimBrain using LLM's and multimodal models, by somehow feeding them the repo, Tensorboard log files (not uploaded publicly yet), and YouTube video. The generated paper will probably need to be edited.

Example structure (might need a multi-prompt approach):

  1. Intro
  2. Methodology
  3. ⁠Experiment
  4. ⁠Result (figures could include the TB graphs and the visualizations of the convolutional filters generated by visualize_model.py)
  5. ⁠Discussion
  6. ⁠Conclusion
mei-chen commented 1 month ago

This could be a good resource to use for getting a draft of the intro together – https://storm.genie.stanford.edu/

mei-chen commented 1 month ago

reinforcement learning for using raw image intake to play game.pdf I just generated a rough draft using Storm ^

wkwan commented 1 month ago

@mei-chen ty! Might use that to beef up the lit review in the ChatGPT generated result

Need to edit and add figures/numbers.