Instruction Tuning
الضبط التعليمي
أسلوب ضبط دقيق يُدرَّب فيه النموذج على أمثلة من التعليمات مقرونة بالاستجابات المطلوبة، مما يحسّن قدرته على اتّباع نوايا المستخدم في مهام متنوعة.
A fine-tuning technique where models are trained on examples of instructions paired with desired responses, improving the model's ability to follow human intent across diverse tasks.
Also translated asالتدريب الإرشادي للنموذج، التوليف بالتعليمات، الضبط الدقيق بالتعليمات، مواءمة اتباع التعليمات
First appears in this corpus in: Finetuned Language Models Are Zero-Shot Learners (2022)
Appears in these papers
- Alpaca: A Strong, Replicable Instruction-Following Model2023in the sky ✦
- BLIP-2: Bootstrapping Language-Image Pre-Training with Frozen Image Encoders and Large Language Models2023in the sky ✦
- Finetuned Language Models Are Zero-Shot Learners2022in the sky ✦
- Finetuned Language Models Are Zero-Shot Learners2022in the sky ✦
- Gemini: A Family of Highly Capable Multimodal Models2023in the sky ✦
- Gemma: Open Models Based on Gemini Research and Technology2024in the sky ✦
- Generative Agents: Interactive Simulacra of Human Behavior2023in the sky ✦
- HyDE: Precise Zero-Shot Dense Retrieval without Relevance Labels2022in the sky ✦
- LIMA: Less Is More for Alignment2023in the sky ✦
- LIMA: Less Is More for Alignment2023in the sky ✦
- LLaMA: Open and Efficient Foundation Language Models2023in the sky ✦
- Visual Instruction Tuning2023in the sky ✦
- Mistral 7B2023in the sky ✦
- Mistral 7B2023in the sky ✦
- PaLM: Scaling Language Modeling with Pathways2022in the sky ✦
- Self-Instruct: Aligning Language Models with Self-Generated Instructions2022in the sky ✦
- Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection2023in the sky ✦