Papers › Partially Shuffling the Training Data to Improve Language Models
Partially Shuffling the Training Data to Improve Language Models
Ofir Press
Although SGD requires shuffling the training data between epochs, currently none of the word-level language modeling systems do this. Naively shuffling all sentences in the training data would not permit the model to learn inter-sentence dependencies. Here we present a method that partially shuffles the training data between epochs. This method makes each batch random, while keeping most sentence ordering intact. It achieves new state of the art results on word-level language modeling on both the Penn Treebank and WikiText-2 datasets.
In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Tasks
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Language Modelling | Penn Treebank (Word Level) | AWD-LSTM-DOC + Partial Shuffle | Params | 23M | #15 of 43 | Archive leaderboard | report |
| Language Modelling | Penn Treebank (Word Level) | AWD-LSTM-DOC + Partial Shuffle | Test perplexity | 52.0 | #15 of 43 | Archive leaderboard | report |
| Language Modelling | Penn Treebank (Word Level) | AWD-LSTM-DOC + Partial Shuffle | Validation perplexity | 53.79 | #15 of 43 | Archive leaderboard | report |
| Language Modelling | Penn Treebank (Word Level) | AWD-LSTM-MoS + Partial Shuffle | Params | 22M | #18 of 43 | Archive leaderboard | report |
| Language Modelling | Penn Treebank (Word Level) | AWD-LSTM-MoS + Partial Shuffle | Test perplexity | 53.92 | #18 of 43 | Archive leaderboard | report |
| Language Modelling | Penn Treebank (Word Level) | AWD-LSTM-MoS + Partial Shuffle | Validation perplexity | 55.89 | #18 of 43 | Archive leaderboard | report |
| Language Modelling | WikiText-2 | AWD-LSTM-DOC + Partial Shuffle | Number of params | 37M | #23 of 38 | Archive leaderboard | report |
| Language Modelling | WikiText-2 | AWD-LSTM-DOC + Partial Shuffle | Test perplexity | 57.85 | #23 of 38 | Archive leaderboard | report |
| Language Modelling | WikiText-2 | AWD-LSTM-DOC + Partial Shuffle | Validation perplexity | 60.16 | #23 of 38 | Archive leaderboard | report |
| Language Modelling | WikiText-2 | AWD-LSTM-MoS + Partial Shuffle | Number of params | 35M | #25 of 38 | Archive leaderboard | report |
| Language Modelling | WikiText-2 | AWD-LSTM-MoS + Partial Shuffle | Test perplexity | 59.98 | #25 of 38 | Archive leaderboard | report |
| Language Modelling | WikiText-2 | AWD-LSTM-MoS + Partial Shuffle | Validation perplexity | 62.38 | #25 of 38 | Archive leaderboard | report |
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
Methods
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections