Diachronia year prediction
Guess the time when a Polish excerpt was published [ver. 1.0.0]
Git repo URL: git://gonito.net/diachronia-year-prediction / Branch: master
Run git clone --single-branch git://gonito.net/diachronia-year-prediction -b master to get the challenge data
Browse at https://gonito.net/gitlist/diachronia-year-prediction.git/master
Leaderboard
# | submitter | when | ver. | description | test-A RMSE | × | |
---|---|---|---|---|---|---|---|
1 | Kaszuba | 2023-06-13 09:24 | 1.0.0 | test solution | 14.95 | 2 | |
2 | zrostek | 2023-04-25 10:06 | 1.0.0 | plT5 summarization with spaces epochs=3 t5 | 14.98 | 37 | |
3 | zrostek | 2023-05-08 07:09 | 1.0.0 | plT5-large summarization with spaces epochs=3 large t5 | 15.00 | 37 | |
4 | zrostek | 2023-05-25 09:22 | 1.0.0 | pretrained nanoT5 (adafactor, legacy) summarization (50 GB), step 70000 out of 364210 (plus 0.5 year) epochs=3 no-pretrained t5 | 15.22 | 37 | |
5 | zrostek | 2022-12-19 08:20 | 1.0.0 | finetuned HerBERT epochs=10 learning-rate=1.0e-6 transformer bert | 16.09 | 37 | |
6 | zrostek | 2022-12-12 08:08 | 1.0.0 | transformer linear regression linear-regression transformer | 16.41 | 37 | |
7 | p/tlen | 2022-12-03 20:31 | 1.0.0 | Vowpal Wabbit regression with a small neural network bits=29 nnsize=6 passes=10 vowpal-wabbit | 17.76 | 3 | |
8 | zrostek | 2023-03-09 08:58 | 1.0.0 | Roberta trained from scratch with 6,2 + 19 GB extra filtered Polish text roberta no-pretrained | 18.68 | 37 | |
9 | Antkowiak | 2023-05-28 12:32 | 1.0.0 | sieć Soplica (korpus: Pan Tadeusz) tf tokenization | 40.09 | 4 | |
10 | Szyszko | 2023-06-12 16:10 | 1.0.0 | Solution neural-network bert | 52.71 | 1 | |
11 | zrostek | 2022-12-05 14:44 | 1.0.0 | linear regression baseline linear-regression tf-idf baseline | 53.32 | 37 | |
12 | p/tlen | 2022-11-19 15:24 | 1.0.0 | null-model baseline null-model baseline | 55.70 | 3 |