Determiner Usage Evaluation on Manchester Corpus (12 child dyads)
0.902Biasgpt-bert-babylm-small
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| gpt-bert-babylm-smallArch.=MLM2026.05 | 0.902 | 18.5 | 18.5 | 43.1 | 0.997 | 0.001 | |
| ChildrenArch.=Human2026.05 | 0.834 | 25.1 | 24.2 | 22.6 | 0.527 | 0.484 | |
| roberta-base-strict-2023Arch.=MLM2026.05 | 0.834 | 30.1 | 24.9 | 47.5 | 0.001 | 0.001 | |
| CaretakersArch.=Human2026.05 | 0.815 | 30.1 | 32 | 20 | 0.191 | 0.256 | |
| ltg-bert-bncArch.=MLM2026.05 | 0.783 | 31.4 | 28.7 | 22.5 | 0.138 | 0.505 | |
| babylm-baseline-100m-gpt-bert-causal-focusArch.=AR2026.05 | 0.782 | 31.4 | 28.8 | 24.7 | 0.122 | 0.041 | |
| roberta-baseArch.=MLM2026.05 | 0.776 | 32.3 | 29.2 | 23.9 | 0.096 | 0.142 | |
| opt-125mArch.=AR2026.05 | 0.756 | 34.6 | 30.3 | 25.1 | 0.027 | 0.018 | |
| t5-baseArch.=S2S2026.05 | 0.753 | 34.7 | 30.4 | 23.9 | 0.03 | 0.085 | |
| gpt2Arch.=AR2026.05 | 0.752 | 34.9 | 30.5 | 25.2 | 0.031 | 0.012 |