Get Started
Home
Topics
Search
Library
Research questionWhen do natural-language rules outperform examples for LLM in-context learning?LLMs can acquire a task from either a natural-language rule description or examples of desired inputs and outputs. The more reliable prompting format may depend on whether the task requires algebraic abstraction, computation, distributional sensitivity, or parametric knowledge.
AI
LLM Pretraining & Post-training
Machine Learning
Natural Language Processing
Reasoning
Latest papersRecent research connected to this question, newest first.LLMs Learn Better In-Context from Rules than from ExamplesThe evidence compares rule-based and example-based in-context learning on five tasks spanning games, arithmetic, and linguistic inference. It includes base and instruction-tuned models and examines whether adding examples or increasing their number changes learning efficacy.research paper · Sep 2, 2026
Related questions
How should low-resource LLM fine-tuning use task-level language priors with ambiguous or incomplete data?How can prompt representations distinguish specificity to expose fine-grained LLM weaknesses?How can language models reliably follow instructions containing many simultaneous constraints?How can instruction-tuned LLMs learn corpus-specific knowledge without exhaustive synthetic QA or instruction fine-tuning?