Brown, T.B.; Mann, B.; Ryder, N.; Subbiah, M.; Kaplan, J.; Dhariwal, P.; Neelakantan, A.; Shyam, P.; Sastry, G.; Askell, A.; Agarwal, S.; Herbert-Voss, A.; Krueger, G.; Henighan, T.; Child, R.; Ramesh, A.; Ziegler, D.M.; Wu, J.; Winter, C.; Hesse, C.; Chen, M.; Sigler, E.; Litwin, M.; Gray, S.; Chess, B.; Clark, J.; Berner, C.; McCandlish, S.; Radford, A.; Sutskever, I.; Amodei, D.: Language models are few-shot learners (2020)
0.02
0.023644082 = product of:
0.055169523 = sum of:
0.014966144 = product of:
0.029932288 = sum of:
0.029932288 = weight(_text_:p in 872) [ClassicSimilarity], result of:
0.029932288 = score(doc=872,freq=4.0), product of:
0.13319843 = queryWeight, product of:
3.5955126 = idf(docFreq=3298, maxDocs=44218)
0.03704574 = queryNorm
0.22471954 = fieldWeight in 872, product of:
2.0 = tf(freq=4.0), with freq of:
4.0 = termFreq=4.0
3.5955126 = idf(docFreq=3298, maxDocs=44218)
0.03125 = fieldNorm(doc=872)
0.5 = coord(1/2)
0.03266309 = weight(_text_:g in 872) [ClassicSimilarity], result of:
0.03266309 = score(doc=872,freq=4.0), product of:
0.13914184 = queryWeight, product of:
3.7559474 = idf(docFreq=2809, maxDocs=44218)
0.03704574 = queryNorm
0.23474671 = fieldWeight in 872, product of:
2.0 = tf(freq=4.0), with freq of:
4.0 = termFreq=4.0
3.7559474 = idf(docFreq=2809, maxDocs=44218)
0.03125 = fieldNorm(doc=872)
0.007540288 = weight(_text_:a in 872) [ClassicSimilarity], result of:
0.007540288 = score(doc=872,freq=24.0), product of:
0.04271548 = queryWeight, product of:
1.153047 = idf(docFreq=37942, maxDocs=44218)
0.03704574 = queryNorm
0.17652355 = fieldWeight in 872, product of:
4.8989797 = tf(freq=24.0), with freq of:
24.0 = termFreq=24.0
1.153047 = idf(docFreq=37942, maxDocs=44218)
0.03125 = fieldNorm(doc=872)
0.42857143 = coord(3/7)
- Abstract
- Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.
- Type
- a