Papers › Improving Language Understanding by Generative Pre-Training

Improving Language Understanding by Generative Pre-Training

11 Jun 2018Preprint 2018 6archive 2025-07-28

Alec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever

Natural language understanding comprises a wide range of diverse tasks such as textual entailment, question answering, semantic similarity assessment, and document classification. Although large unlabeled text corpora are abundant, labeled data for learning these specific tasks is scarce, making it challenging for discriminatively trained models to perform adequately. We demonstrate that large gains on these tasks can be realized by generative pre-training of a language model on a diverse corpus of unlabeled text, followed by discriminative fine-tuning on each specific task. In contrast to previous approaches, we make use of task-aware input transformations during fine-tuning to achieve effective transfer while requiring minimal changes to the model architecture. We demonstrate the effectiveness of our approach on a wide range of benchmarks for natural language understanding. Our general task-agnostic model outperforms discriminatively trained models that use architectures specifically crafted for each task, significantly improving upon the state of the art in 9 out of the 12 tasks studied. For instance, we achieve absolute improvements of 8.9% on commonsense reasoning (Stories Cloze Test), 5.7% on question answering (RACE), and 1.5% on textual entailment (MultiNLI).

PaperPDFCode

Code

13 repositories listed; official and paper-mentioned ones first.

lvyufeng/bert4ms mindsporeApache-2.0 report

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

Cloze TestDocument ClassificationLanguage ModelingLanguage ModellingNatural Language InferenceNatural Language UnderstandingQuestion AnsweringSemantic SimilaritySemantic Textual Similarity

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Natural Language Inference MultiNLI Finetuned Transformer LM Matched 82.1 #41 of 67 Archive leaderboard report
Natural Language Inference MultiNLI Finetuned Transformer LM Mismatched 81.4 #41 of 67 Archive leaderboard report
Natural Language Inference SNLI Fine-Tuned LM-Pretrained Transformer % Test Accuracy 89.9 #15 of 98 Archive leaderboard report
Natural Language Inference SNLI Fine-Tuned LM-Pretrained Transformer % Train Accuracy 96.6 #15 of 98 Archive leaderboard report
Natural Language Inference SNLI Fine-Tuned LM-Pretrained Transformer Parameters 85m #15 of 98 Archive leaderboard report
Natural Language Inference SciTail Finetuned Transformer LM Accuracy 88.3 #4 of 13 Archive leaderboard report
Question Answering RACE Finetuned Transformer LM RACE 59.0 #4 of 7 Archive leaderboard report
Question Answering RACE Finetuned Transformer LM RACE-h 57.4 #4 of 7 Archive leaderboard report
Question Answering RACE Finetuned Transformer LM RACE-m 62.9 #4 of 7 Archive leaderboard report
Question Answering StoryCloze Finetuned Transformer LM Accuracy 86.5 #8 of 23 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

Introduced by this paper: GPT

AdamAttentionAttention DropoutBPECosine AnnealingDense ConnectionsDiscriminative Fine-TuningDropoutGPTLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionResidual ConnectionSoftmaxWeight Decay

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