Back
Rahul Agarwal
Founder | Agentic AI... • 1d
Prompting isn’t simply typing a question into an AI — it’s a structured communication method and one of the key skills needed to work effectively with LLMs. Here’s a clearer breakdown of the different approaches: 1. Foundational Prompt Styles Zero-shot: Just give the task with no examples. One-shot: Provide one example before asking for the output. Few-shot: Share multiple examples so the model learns the pattern. 2. Techniques for Better Reasoning Chain-of-Thought: Ask the model to think step-by-step. Self-Consistency: Generate multiple reasoning paths and pick the best output. Tree-of-Thought: Explore multiple reasoning branches (more advanced). ReAct: Blend reasoning with actions like tool or API calls. 3. Instruction & Role-Based Approaches Instruction-based: Direct and specific guidance (e.g., “Explain this in two sentences”). Role prompting: Assign the model a role (e.g., “Act as a cybersecurity expert”). Hybrid prompts: Combine clear instructions with example-based learning. 4. Structuring and Linking Prompts Prompt chaining: Use one response as input for the next step. Dynamic prompts: Insert real-time context or live data. Meta prompts: Ask the model to review, refine, or verify its own output. 5. Multimodal Prompting Text + images Text + audio/video (Used in models designed to handle multiple input types.) 6. Specialized Prompting for Specific Fields Coding prompts: Precise syntax and constraints. Healthcare/Legal prompts: High accuracy, structured formats, and strict language. 7. Testing & Refining Prompts (Not techniques themselves, but essential.) Ablation: Remove parts of a prompt to see what matters. Injection testing: Check how well the prompt holds up in real applications. What Doesn’t Count as Prompting RAG systems: Retrieval + generation workflows where prompts are just one piece. Autonomous agents or tool-based systems: Prompting is part of the mechanism, not the method. Modern prompting is shifting from simple instructions to system-level design. If you work with LLMs, mastering these approaches isn’t optional — it’s foundational.
More like this
Recommendations from Medial
Account Deleted
Hey I am on Medial • 6m
10 Pro Techniques of prompt engineering for Lovable 1. Structured Prompts: Break your prompt into clear sections: Context, Task, Guidelines, and Constraints. Label each section explicitly for clarity. 2. Specificity and Detail: Clearly descr
See More
Kimiko
Startups | AI | info... • 8m
【 CHATGPT IMAGE PROMPT GUIDE 】 Structure your prompts like a creative director: Instruction → Scene → Subject → Setting → Style → Treat it like a visual brief. → Keywords (bokeh, 35mm) do the heavy lifting. → Save + reuse this structure across GPT-
See More
Daksh Sikarwar
Make your idea to a ... • 2m
AI Prompt That Never Fails (RTCCO) Role • Task • Context • Constraints • Output. “Act as a performance marketer. Task: write 3 ad angles. Context: yoga studio in Agra. Constraints: 60–80 chars, no clichés. Output: table with hook + CTA.” Try once, yo
See More
SPACEROCK TRADING
Top 7 Global indices... • 10m
Hi friends....Is there any best AI based 30 Seconds or 1min Trailer making or create application for free against text prompt ? Professionals Please suggest other than Google 1st page results, I. need application which experienced own with instant o
See More
SPACEROCK TRADING
Top 7 Global indices... • 10m
Hi friends....Is there any best AI based 30 Seconds or 1min Trailer making or create application for free against text prompt ? Professionals Please suggest other than Google 1st page results, I. need application which experienced own with instant o
See More
Download the medial app to read full posts, comements and news.