Research map / Large language models / Language, safety and interpretability
Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learners
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use differentiable prompts instead of conventional fine-tuning | BERT-large · SemEval-2010 Task 8, K=8 | micro F1: 26.3 -> 51.8 (+25.5) | Helped |
| Use differentiable prompts instead of conventional fine-tuning | BERT-large · TACRED-Revisit, K=8 | micro F1: 7.4 -> 25.8 (+18.4) | Helped |
| Use differentiable prompts instead of conventional fine-tuning | BERT-large · WiKi80, K=8 | accuracy: 46.3 -> 68.5 (+22.2) | Helped |
| Use differentiable prompts instead of conventional fine-tuning | SCIBERT · ChemProt, K=8 | micro F1: 30.2 -> 57.2 (+27.0) | Helped |
| Add the fluency constraint objective (ablation) | DART · SemEval-2010 Task 8, K=8 | micro F1: 50.3 -> 51.8 (+1.5) | Helped |
| Add differentiable template optimization (ablation) | DART · SemEval-2010 Task 8, K=8 | micro F1: 49.8 -> 51.8 (+2.0) | Helped |
| Add differentiable label optimization (ablation) | DART · SemEval-2010 Task 8, K=8 | micro F1: 47.5 -> 51.8 (+4.3) | Helped |
| Vary the learned template length from 3 to 2, 5, or 10 tokens (ablation) | DART · SST-2 | accuracy: 93.5 (0.5); 92.6 (0.6); 91.2 (1.1); 90.6 (0.5) | Mixed |
Closest papers by meaning
- Bidirectional Language Models Are Also Few-shot Learners (ICLR 2023 · similarity 0.88)
- Tuning Language Models as Training Data Generators for Augmentation-Enhanced Few-Shot Learning (ICML 2023 · similarity 0.87)
- FLEX: Unifying Evaluation for Few-Shot NLP (NeurIPS 2021 · similarity 0.87)
- Dialogue for Prompting: A Policy-Gradient-Based Discrete Prompt Generation for Few-Shot Learning (AAAI 2024 · similarity 0.85)
- Large Language Models are Good Prompt Learners for Low-Shot Image Classification (CVPR 2024 · similarity 0.85)
- BayesPrompt: Prompting Large-Scale Pre-Trained Language Models on Few-shot Inference via Debiased Domain Abstraction (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.