Our new X account is live! Follow @wizwand_team for updates

Language Modeling on Penn Treebank (PTB) (val)

46.63Perplexity

AWD-LSTM + MoS + Partial Shuffled + Adversarial Training

Updated 2d ago

Evaluation Results

Method
AWD-LSTM + MoS + Partial Shuffled + Adversarial Training 2019.06	46.63	-	-	-	-
AWD-LSTM + MoS + Adversarial Training 2019.06	47.15	-	-	-	-
AWD-LSTM-MoS + dynamic evaluation with FRAGE 2018.09	47.38	-	-	-	-
AWD-LSTM + MoS + Partial Shuffled 2019.06	47.93	-	-	-	-
AWD-LSTM + MoS 2019.06	48.33	-	-	-	-
AWD-LSTM-MoS + dynamic evaluation 2018.09	48.33	-	-	-	-
AWD-LSTM + Adversarial Training 2019.06	49.31	-	-	-	-
ADV-AWD-LSTM + MaxUp 2020.02	50.83	-	-	-	-
AWD-LSTM 2019.06	51.6	-	-	-	-
ADV-AWD-LSTM 2020.02	51.6	-	-	-	-
AWD-LSTM + continuous cache pointer with FRAGE 2018.09	52.3	-	-	-	-
AWD-LSTM + continuous cache pointer 2018.09	53.9	-	-	-	-
AWD-LSTM + MoS + Partial Shuffled + Adversarial Training 2019.06	54.1	-	-	-	-
AWD-LSTM-DOC 2022.03	54.1	-	-	-	-
AWD-LSTM + MoS + Adversarial Training 2019.06	54.98	-	-	-	-
AWD-LSTM-MoS with FRAGE 2018.09	55.52	-	-	-	-
AWD-LSTM + MoS + Partial Shuffled 2019.06	55.89	-	-	-	-
ADV-AWD-LSTM + MaxUp 2020.02	56.25	-	-	-	-
LSTM + MoS 2019.02	56.5	22	-	-	-
AWD-LSTM-MoS 2022.03	56.5	-	-	-	-
AWD-LSTM + MoS 2019.06	56.54	-	-	-	-
AWD-LSTM + MoS 2020.02	56.54	-	-	-	-
AWD-LSTM-MoS 2018.09	56.54	-	-	-	-
AWD-FWM (ours) 2020.11	56.76	-	-	-	-
AWD-LSTM + Adversarial Training 2019.06	57.15	-	-	-	-
ADV-AWD-LSTM 2020.02	57.15	-	-	-	-
AWD-LSTM-MoS w/o finetune with FRAGE 2018.09	57.55	-	-	-	-
Random search WS 2019.02	57.8	23	0.25	1	1.25
AWD-LSTM-MoS w/o finetune 2018.09	58.08	-	-	-	-
ASHA + LSTM + DropConnect 2019.02	58.1	24	-	-	13
DARTS (second order) 2019.02	58.1	23	1	1	2
LSTM + SE 2019.10	58.1	-	-	-	-
DARTS (2nd) 2019.10	58.1	-	-	-	-
AWD-LSTM + FRAGE 2020.02	58.1	-	-	-	-
AWD-LSTM with FRAGE 2018.09	58.1	-	-	-	-
DARTS (second order) 2019.02	58.2	23	1	1	2
DARTS 2019.06	58.3	-	-	-	-
ON-LSTM 2022.03	58.3	-	-	-	-
AWD-LSTM 2020.02	58.5	-	-	-	-
ASHA baseline 2019.02	58.6	23	-	-	2
DM-LSTM 2022.03	58.6	-	-	-	-
Fraternal dropout 2022.03	58.9	-	-	-	-
AWD-TXL (ours) 2020.11	59.39	-	-	-	-
GDAS 2019.10	59.8	-	-	-	-
LSTM + DropConnect 2019.02	60	24	-	-	-
AWD-LSTM 2019.06	60	-	-	-	-
AWD-LSTM 2018.09	60	-	-	-	-
AWD-LSTM 2022.03	60	-	-	-	-
AWD-LSTM 2022.05	60	-	-	-	-
AWD-LSTM 2020.11	60	-	-	-	-
DARTS (first order) 2019.02	60.2	23	0.5	1	1.5
DARTS (1st) 2019.10	60.2	-	-	-	-
AWD-LSTM w/o finetune with FRAGE 2018.09	60.2	-	-	-	-
LSTM 2019.10	60.7	-	-	-	-
AWD-LSTM w/o finetune 2018.09	60.7	-	-	-	-
ENAS 2019.02	60.8	24	0.5	-	-
ENAS 2019.10	60.8	-	-	-	-
LSTM + SC 2019.10	60.9	-	-	-	-
Random search baseline 2019.02	61.8	23	-	-	2
V-RHN 2019.10	67.9	-	-	-	-
Variational RHN 2022.03	67.9	-	-	-	-
Variational LSTM + weight tying 2019.06	71.1	-	-	-	-
Pointer Sentinel-LSTM 2022.03	72.4	-	-	-	-
VL-HMM 2022.05	125	-	-	-	-
Rank-Space Models 2022.05	135.6	-	-	-	-
VL-HMM 2022.05	136	-	-	-	-
Rank-Space Models 2022.05	137	-	-	-	-
LHMM 2022.05	141.4	-	-	-	-
HMM+RNN 2022.05	142.3	-	-	-	-
VL-HMM 2022.05	146	-	-	-	-