Decider model author defends gains as private split matches public leaderboard

antoinechaffin says the decider model's gains come from removing the causal mask and more training data, and the tiny public-private leaderboard gap shows it isn't overfitting.

2026-10-07 ~ 2026-10-07 · 2 related posts

• Small public-private leaderboard gap cited as proof of no benchmaxxing — antoine_chaffin · 2026-10-07
• Decider model gains from unmasking and more data, authors deny benchmaxxing — antoine_chaffin · 2026-10-07