Simple Few-Shot Prompts Beat Complex Techniques for Small Models on NLI

A paper from Hellenic Open University and colleagues systematically evaluates small open-source decoder-only models (Llama, Mistral, Gemma, Phi 4), showing that simple prompts with well-chosen few-shot examples outperform complex prompting techniques on NLI tasks.

2026-11-11 ~ 2026-11-11 · 2 related posts

• Simple prompts with few examples match complex ICL methods on NLI, small-LLM study finds — rohanpaul_ai · 2026-11-11
• Study finds simple few-shot prompts let small open LLMs match complex prompting on NLI — rohanpaul_ai · 2026-11-11