Language Models are Few-Shot Learners
scale, and a task specified in the prompt rather than the weights
Read it
- 19 sections
- 0 figures
- 0 equations
- 144 references
- cited by 0 here
- difficulty 3 of 5
- the publisher's page
Contents
- 1Introduction
- 2Approach
- 3Results
- 4Measuring and Preventing Memorization Of Benchmarks
- 5Limitations
- 6Broader Impacts
- 7Related Work
- 8Conclusion
- Acknowledgments
- ADetails of Common Crawl Filtering
- BDetails of Model Training
- CDetails of Test Set Contamination Studies
- DTotal Compute Used to Train Language Models
- EHuman Quality Assessment of Synthetic News Articles
- FAdditional Samples from GPT-3
- GDetails of Task Phrasing and Specifications
- HResults on All Tasks for All Model Sizes