Challenging America word-gap prediction

Guess a word in a gap. [ver. 3.0.0]

Git repo URL: git://gonito.net/challenging-america-word-gap-prediction / Branch: master
Run git clone --single-branch git://gonito.net/challenging-america-word-gap-prediction -b master to get the challenge data
Browse at https://gonito.net/gitlist/challenging-america-word-gap-prediction.git/master

Leaderboard

# submitter when ver. description test-A PerplexityHashed ×
1 kubapok 2021-12-11 17:25 3.0.0 roberta_large_no_ft 52.58 26
2 Jakub 2023-06-27 20:07 3.0.0 Updated input truncation 85.17 31
3 s478846 2023-05-30 15:49 3.0.0 neural solution neural-network n-grams bow 105.79 42
4 s444501 2023-05-30 11:50 3.0.0 s444501 neural with bagging bagging_left_ctx=25 bagging_right_ctx=25 batch_size=4000 embed_size=300 epochs=1 hidden_size=150 learning-rate=1.0e-4 ngram_left_ctx=7 [...] neural-network 126.69 17
5 [anonymized] 2023-06-11 17:15 3.0.0 zad9 132.60 10
6 s444391 2023-06-14 10:05 3.0.0 Embeddings and bow solution embed_size=300 epochs=1 hidden_size=300 learning-rate=1.0e-4 vocab_size=5000 neural-network 132.60 13
7 s444356 2023-05-30 14:53 3.0.0 neural network, bag of words batch-size=10000 context-size=25 dropout=0.3 embed-size=200 epochs=1 hidden-size=1000 hidden-size_2=500 learning-rate=1.0e-3 [...] neural-network bow 134.70 73
8 s444354 2023-06-13 21:05 3.0.0 trigram model embed_size=300 epochs=1 hidden_size=750 learning-rate=1.0e-4 vocab_size=5000 138.03 24
9 s444018 2023-05-30 22:47 3.0.0 s444018 batch-size=10000 dropout=0.3 embed-size=300 hidden-size=1000 second-hidden-size=500 vocab-size=12500 neural-network bow 139.63 27
10 s409771 2023-06-07 19:14 3.0.0 gpt2-large gpt2-large 142.89 8
11 s478815 2023-06-02 19:15 3.0.0 neural network 149.96 20
12 ked 2023-05-25 23:12 3.0.0 zad9_v3 neural-network 177.87 15
13 s478873 2023-05-31 09:26 3.0.0 s478873 batch-size=5000 embed-size=100 epochs=1 hidden_size=100 top=100 vocab-size=10000 neural-network bow 180.53 27
14 s444517 2023-06-07 14:41 3.0.0 gpt-2, only left-context gpt2 transformer-decoder 183.70 12
15 [anonymized] 2023-06-26 21:55 3.0.0 gpt-2 left context out files left-to-right gpt2 transformer-decoder 185.64 22
16 Jakub Eichner 2023-05-31 08:17 3.0.0 Bag of words NN 188.60 19
17 [anonymized] 2023-06-08 18:03 3.0.0 s444421 top=10 gpt2 215.47 9
18 Mikołaj Pokrywka 2023-06-08 11:07 3.0.0 all done 239.28 15
19 [anonymized] 2023-04-22 22:15 3.0.0 s478831 239.65 15
20 444498 2023-05-26 17:13 3.0.0 trigram, tetragram 239.79 7
21 Cezary 2023-05-10 07:51 3.0.0 This is my trigram nn solution batch_size=15000 embed_size=200 epochs=1 hidden_size=190 learning-rate=1.0e-3 top=10 vocab_size=20000 neural-network trigram 243.43 5
22 s444386 2023-05-30 20:39 3.0.0 nn with bag of words neural-network bow 243.69 27
23 s444452 2023-06-08 16:14 3.0.0 lstm lstm 257.14 10
24 s478840 2023-05-10 20:25 3.0.0 s478840 batch-size=1500 embed-size=250 hidden-size=100 top=600 vocab-size=20000 neural-network trigram 263.30 21
25 Kamil Guttmann 2023-05-09 17:59 3.0.0 nn, trigram, predict from next two tokens batch-size=5000 embed-size=100 embed_size=256 epochs=1 hidden_size=2048 topk=150 vocab-size=20000 neural-network trigram 275.80 12
26 s478855 2023-06-07 22:39 3.0.0 s478855 decoder 283.52 12
27 s444415 2023-05-30 19:13 3.0.0 Nn ngram model top=600 neural-network n-grams bow 300.73 24
28 Adam Wojdyła 2023-06-28 15:51 3.0.0 s444507 top=30 neural-network 303.80 43
29 Łukasz Jędyk 2022-05-29 10:25 3.0.0 434708 lstm ensemble 343.21 7
30 Przemek 2022-04-10 18:49 3.0.0 434766 plusalpha plusaplha 343.70 4
31 s444476 2023-05-08 22:26 3.0.0 trigram nn done HIDDENLAYER=1000 355.85 7
32 [name not given] 2022-04-10 23:14 3.0.0 434742 n-grams 379.56 13
33 Jakub Pietrzak 2022-04-09 20:59 3.0.0 470628 n-grams plusaplha 412.64 8
34 s444383 2023-05-15 10:16 3.0.0 bigram 420.77 9
35 Jakub Pogodziński 2022-04-11 09:17 3.0.0 437622 alpha n-grams goodturing 427.45 3
36 [name not given] 2022-04-10 17:19 3.0.0 s470611 n-grams backoff 436.81 11
37 [anonymized] 2022-06-26 18:43 3.0.0 434695 smoothing plusalpha plusaplha 436.81 6
38 [anonymized] 2022-05-07 14:40 3.0.0 426206 neural-network bigram 447.01 5
39 [anonymized] 2022-05-08 17:33 3.0.0 434732 neural-network bigram 454.57 5
40 [anonymized] 2022-04-10 19:34 3.0.0 s434804 n-grams plusaplha 533.82 5
41 zrostek 2022-05-29 22:19 3.0.0 470619 lstm ensemble 597.73 6
42 [anonymized] 2022-04-10 19:19 3.0.0 s430705 plusalpha n-grams goodturing 628.51 7
43 Piotr Kopycki 2022-04-10 21:01 3.0.0 470629 plusaplha 714.72 2
44 Wojciech Jarmosz 2022-04-10 12:57 3.0.0 s434704 n-grams plusaplha goodturing 828.47 8
45 s434788 2022-04-10 22:44 3.0.0 434788 plusalpha plusaplha NaN 3
46 s444455 2023-06-12 23:24 3.0.0 embeddings, bov 130.68 14
47 Jakub Adamski 2023-06-14 15:53 3.0.0 nn trigram neural-network trigram 132.60 23
48 s444417 2023-05-30 12:04 3.0.0 nn, vocab_size: 1500, embed_size: 200, batch_size: 5000, epoch:1, hl, l_ctx, r_ctx = 10 batch_size=5000 embed_size=200 vocab_size=15000 neural-network 143.54 11
49 Marcin Kostrzewski 2023-09-25 10:16 3.0.0 GRU, both left and right context used, embedding_size=128, hidden_size=256, gru_layers=4, epochs=10 175.95 18
50 s478839 2023-06-24 22:05 3.0.0 gpt2-fine-tuned top=30 train_dataset=50000 gpt2 fine-tuned 226.54 28
51 s443930 2023-06-10 20:59 3.0.0 zadanie12-1 231.20 26
52 Martyna Druminska 2023-05-30 22:37 3.0.0 trigram model batch_size=800 embed_size=300 epochs=1 hidden_size=128 learning-rate=1.0e-4 vocab_size=38000 neural-network trigram 232.67 24
53 [anonymized] 2022-04-04 16:44 3.0.0 434749: n-gramowy model oparty na 3-gram (fill in the middle) + backoff to 2-gram + backoff to 2-gram reversed + alpha smoothing n-grams backoff 322.06 2
54 MaciejSobkowiak 2022-04-10 22:59 3.0.0 s434784 n-grams 379.52 4
55 [anonymized] 2022-04-03 21:45 3.0.0 434780 n-grams 379.52 8
56 Anna Nowak 2022-05-01 09:22 3.0.0 434760, bigram neural-network final neural-network bigram 453.65 21
57 Piotr 2022-05-09 19:39 3.0.0 440058 neural network neural-network bigram 454.57 5
58 [anonymized] 2022-04-11 18:28 3.0.0 test n-grams 519.82 2
59 [anonymized] 2022-04-05 17:11 3.0.0 440054 n-grams 568.10 5
60 s444465 2023-06-27 22:17 3.0.0 s444465 bigram neural model fixed v2 batch_size=2500 embed_size=200 epochs=5 learning-rate=1.0e-3 vocab_size=40000 neural-network bigram 934.09 13
61 Wiktor Bombola 2023-06-02 12:28 3.0.0 change the standarization method of output, wildcard 1024.24 20
62 [anonymized] 2023-03-29 10:05 3.0.0 test Infinity 1

Graphs by parameters

context-size

[direct link]

embed_size

[direct link]

hidden-size

[direct link]

top

[direct link]

top_k

[direct link]

topk

[direct link]

vocab-size

[direct link]

vocab_size

[direct link]