Research map / Large language models / Language, safety and interpretability
True Few-Shot Learning with Language Models
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 |
|---|---|---|---|
| select prompts with CV/MDL on training examples instead of random prompt | GPT-3 family (0.1B-175B) · LAMA-UHN | accuracy | No clear effect |
| true few-shot (CV/MDL) prompt selection instead of held-out validation selection | GPT-3 family · LAMA-UHN | accuracy: -5-7% absolute | Hurt |
| conservative CV_alpha selection penalizing variance (ablation) | GPT-3 175B prompt selection · LAMA | chance of beating average prompt: 50% vs 51% | No clear effect |
| increase labeled examples from 5 to 40 (ablation) | CV/MDL prompt selection · LAMA | accuracy | No clear effect |
| more forward passes to estimate CV/MDL (ablation) | prompt selection, GPT-3 6.7B · LAMA | accuracy | No clear effect |
| choose ADAPET hyperparameters with CV instead of validation | ADAPET · SuperGLUE (32-shot) | average score: 77.2 -> 73.0 | Hurt |
| CV/MDL hyperparameter selection vs random hyperparameters | ADAPET · SuperGLUE (32-shot) | average score: 73.9 vs 73.4 (MDL) | No clear effect |
| CV prompt selection | GPT-3 175B · CB | accuracy | Hurt |
Closest papers by meaning
- The Surprising Effectiveness of Test-Time Training for Few-Shot Learning (ICML 2025 · similarity 0.91)
- Optimization Inspired Few-Shot Adaptation for Large Language Models (NeurIPS 2025 · similarity 0.89)
- Many-Shot In-Context Learning (NeurIPS 2024 · similarity 0.89)
- Task Contamination: Language Models May Not Be Few-Shot Anymore (AAAI 2024 · similarity 0.89)
- Large Language Models are Demonstration Pre-Selectors for Themselves (ICML 2025 · similarity 0.89)
- From Few to Many: Self-Improving Many-Shot Reasoners Through Iterative Optimization and Generation (ICLR 2025 · similarity 0.88)
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.