Raw Counter-Strike 2 gameplay screenshots for vision-language model training.
Labels coming soon — will include ground truth JSON with game_state, analysis, and advice.
Think Before You See: VLMs as Game Agents Without Reinforcement Learning from Scratch
