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

Language Modeling on Penn Treebank (test)

46.34Perplexity

GL-LWGC-AWD-MOS-LSTM + dynamic evaluation

Updated 3d ago

Evaluation Results

Method
GL-LWGC-AWD-MOS-LSTM + dynamic evaluation 2017.08	46.34	-	-	-	-
AWD-LSTM-DOC (fin) x 5 2018.08	47.17	-	-	-	-
AWD-LSTM-MoS+PDR + dynamic evaluation 2018.08	47.3	-	-	-	-
Ensemble of All 2023.11	47.31	-	-	-	-
AWD-LSTM-MoS + dynamic evaluation 2017.08	47.69	-	-	-	-
AWD-LSTM-MoS 2017.11	47.69	-	-	-	-
AWD-LSTM-MoS + dynamic evaluation 2019.05	47.69	-	-	-	-
AWD-LSTM-MoS + dynamic evaluation 2018.08	47.7	-	-	-	-
AWD-LSTM-DOC x 5 2018.08	48.44	-	-	-	-
AWD-LSTM+PDR + dynamic evaluation 2018.08	49.3	-	-	-	-
AWD-LSTM-DRILL + dynamic evaluation 2019.05	49.4	-	-	-	-
AWD-LSTM + dynamic evaluation 2017.08	51.1	-	-	-	-
AWD-LSTM + dynamic eval 2017.09	51.1	-	-	-	-
AWD-LSTM 2017.11	51.1	-	-	-	-
AWD-LSTM + dynamic evaluation 2018.08	51.1	-	-	-	-
AWD-LSTM + dynamic evaluation 2019.05	51.1	-	-	-	-
AWD-LSTM+PDR + continuous cache pointer 2018.08	51.6	-	-	-	-
DOC + Finetune* + Partial Shuffle 2019.03	52	-	-	-	-
AWD-LSTM-DOC (fin) 2018.08	52.38	-	-	-	-
DOC + Finetune* 2019.03	52.38	-	-	-	-
AWD-LSTM-DOC 2023.11	52.38	-	-	-	-
GL-LWGC-AWD-MOS-LSTM + finetune 2017.08	52.57	-	-	-	-
DOC + Finetune + Partial Shuffle 2019.03	52.58	-	-	-	-
AWD-LSTM - 3-layer LSTM (tied) + continuous cache pointer 2017.08	52.8	-	-	-	-
AWD-LSTM + neural cache 2017.09	52.8	-	-	-	-
AWD-LSTM 2017.11	52.8	-	-	-	-
AWD-LSTM + continuous cache pointer 2018.08	52.8	-	-	-	-
AWD-LSTM + continuous cache pointer 2019.05	52.8	-	-	-	-
AWD-LSTM-DOC 2018.08	52.87	-	-	-	-
DOC + Finetune 2019.03	52.87	-	-	-	-
GL-LWGC-AWD-MOS-LSTM 2017.08	52.95	-	-	-	-
DOC + Partial Shuffle 2019.03	53.28	-	-	-	-
DOC 2019.03	53.44	-	-	-	-
2-layers GL-LWGC-AWD-MOS-LSTM + finetune 2017.08	53.54	-	-	-	-
AWD-LSTM-MoS+PDR 2018.08	53.8	-	-	-	-
MoS + Finetune + Partial Shuffle 2019.03	53.92	-	-	-	-
AWD-LSTM-MOS 2023.11	54	-	-	-	-
TrellisNet 2021.03	54.19	-	-	-	-
TrellisNet-MoS (1.4x larger) 2018.10	54.19	-	-	-	-
AWD-LSTM-MoS 2018.08	54.4	-	-	-	-
AWD-LSTM-MoS + finetuning 2019.06	54.4	-	-	-	-
AWD-LSTM-MoS + finetune 2017.08	54.44	-	-	-	-
AWD-LSTM-MoS 2017.11	54.44	-	-	-	-
AWD-LSTM-MoS 2018.08	54.44	-	-	-	-
AWD-LSTM-MoS 2019.05	54.44	-	-	-	-
Transformer-XL 2019.06	54.5	-	-	-	-
MoS + Finetune 2019.03	54.64	-	-	-	-
TrellisNet-MoS 2018.10	54.67	-	-	-	-
Li et al. (2018) 2022.06	55.2	-	-	-	-
MoS + Partial Shuffle 2019.03	55.35	-	-	-	-
Transformer-XL 2023.11	55.41	-	-	-	-
Reference Model 2017.08	55.58	-	-	-	-
AWD-LSTM+PDR 2018.08	55.6	-	-	-	-
AWD-LSTM-DRILL 2019.05	55.7	-	-	-	-
AWD-LSTM + Phrase Induction + finetuning 2019.06	55.7	-	-	-	-
ENAS 2018.02	55.8	-	-	-	-
ENAS 2018.10	55.8	-	-	-	-
AWD-LSTM-MoS 2017.08	55.97	-	-	-	-
AWD-LSTM-MoS 2017.11	55.97	-	-	-	-
AWD-LSTM-MoS 2018.10	55.97	-	-	-	-
LSTM 2018.02	56	-	-	-	-
DARTS 2018.10	56.1	-	-	-	-
Differentiable NAS 2019.06	56.1	-	-	-	-
MoS 2019.03	56.18	-	-	-	-
ON-LSTM 2019.06	56.2	-	-	-	-
Fast Weights 2023.11	56.45	-	-	-	-
Fraternal dropout + AWD-LSTM 3-layer 2017.10	56.8	-	-	-	-
AWD-LSTM + Fraternal Dropout 2018.08	56.8	-	-	-	-
TrellisNet (1.4x larger) 2018.10	56.8	-	-	-	-
TrellisNet 2018.10	56.97	-	-	-	-
AWD-LSTM-DRILL 2019.05	57	-	-	-	-
EGRU 2022.06	57	-	-	84.8	42.9
EGRU 2023.11	57.18	-	-	-	-
EGRU 2022.06	57.2	-	-	79.7	46.3
AWD-LSTM 2017.08	57.3	-	-	-	-
AWD-LSTM - 3-layer LSTM (tied) 2017.08	57.3	-	-	-	-
AWD-LSTM 2017.09	57.3	-	-	-	-
LSTM 2018.02	57.3	-	-	-	-
AWD-LSTM 2017.11	57.3	-	-	-	-
Merity et al. (2017a) - AWD-LSTM 3-layer (baseline) 2017.10	57.3	-	-	-	-
AWD-LSTM 2018.08	57.3	-	-	-	-
AWD-LSTM 2018.08	57.3	-	-	-	-
AWD-LSTM 2019.05	57.3	-	-	-	-
AWD-LSTM + finetuning 2019.06	57.3	-	-	-	-
Merity et al. (2018) 2022.06	57.3	-	-	-	-
AWD-LSTM + Phrase Induction 2019.06	57.5	-	-	-	-
AWD-LSTM MoC 2018.10	57.55	-	-	-	-
LSTM 2018.02	57.6	-	-	-	-
AWD-LSTM 2017.09	57.7	-	-	-	-
Lookahead (ADAM) 2019.07	57.72	-	-	-	-
AWD-LSTM + Phrase Induction (No Attention) 2019.06	58	-	-	-	-
2-layer skip connection LSTM 2017.08	58.3	-	-	-	-
4-layer skip connection LSTM (tied) 2017.08	58.3	-	-	-	-
LSTM + BB tuning 2017.09	58.3	-	-	-	-
4-layer skip connection LSTM (tied) 2017.11	58.3	-	-	-	-
2-layer skip connection LSTM 2017.11	58.3	-	-	-	-
Melis et al. (2017) - 4-layer skip connection LSTM 2017.10	58.3	-	-	-	-
LSTM with skip connections 2018.08	58.3	-	-	-	-
LSTM + skip conn. (Black-box tuned) 2018.10	58.3	-	-	-	-
2-layer skip connection LSTM 2019.05	58.3	-	-	-	-

Showing 100 of 419 rows