GRPO checkpoints from the chess easy-to-hard experiment (github.com/nandansarkar16/chess_rl): four arms per pre-training budget, group size 16.
David Demitri Africa PRO
davidafrica
AI & ML interests
None yet
Recent Activity
updated a dataset 2 days ago
davidafrica/e2h-polaris-gemma-4-e4b-it-rollouts published a dataset 2 days ago
davidafrica/e2h-polaris-gemma-4-e4b-it-rollouts updated a collection 2 days ago
pico-maml