Grounding

As explained in Building LLM Applications with DSPy

A method of preparing a language model with substantial information about a task so its responses and instructions reflect that task.

Grounding prepares a prompt language model to generate task-specific instructions by giving it substantial information about the task. In MIPROv2, that information includes the module’s source code and a dataset description with training examples, helping candidate instructions reflect what the task model must do. section 6.2.1

MIPROv2 uses grounding as part of its instruction-generation step:

• Summarize the dataset and source code; these summaries require language-model calls. section 6.2.1
• Use the summaries to prompt for alternative instructions. Tips can guide summary and candidate-generation calls, and can be toggled with tip_aware_proposer. section 6.2.1
• Evaluate candidate instructions efficiently: grounding aims to improve their quality when testing each candidate is costly. section 6.2.1

The dataset reveals a strong focus on flight bookings, with users consistently seeking specific routes, departure times, and additional travel logistics such as fare details and airport information. This focus highlights significant demand for travel assistance applications that provide efficient booking processes along with comprehensive travel-planning insights. Overall, the observations suggest a user preference for quick, clear communications and tailored solutions that encompass the entire travel experience.

A generated dataset summary describes flight routes, departure times, fares, and other travel-planning needs, illustrating the task context available to the instruction proposer. section 6.2.1

How Building LLM Applications with DSPy covers it

How deep it goes

Describes task inputs and summary-then-generation steps, but does not specify how the summaries are constructed.

Examples

Uses flight-booking intent classification and a travel dataset summary mentioning routes, times, fares, and airport information.

What you can do after

Apply grounding through MIPROv2 to generate task-specific candidate instructions from source code and dataset summaries.

How it connects

Links MIPROv2 grounding with instruction optimization, few-shot demonstrations, BootstrapFewShot, COPRO, SIMBA, and GEPA.

liveBookmanning.com

Read Building LLM Applications with DSPy on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Natural Language Processing in Action, Second EditionGrounding is the process of anchoring a language model’s generated response in relevant external facts, sources, or user-specific information.
• 

  liveBookmanning.com
  Generative AI in ActionGrounding is the technique of supplying a generative model with relevant external information as context for its response.
• 

  liveBookmanning.com
  Build AI-Enhanced Web AppsGrounding is the technique of checking generated information against a reliable external corpus to improve factual accuracy.
• 

  liveBookmanning.com
  LLMs in ProductionGrounding is the practice of supplying a language model with relevant external context so its generation is anchored to available information.
• 

  liveBookmanning.com
  AI Agents in ActionGrounding is the assessment of how well a prompt or agent profile satisfies criteria for being constrained by relevant information or facts.
• 

  liveBookmanning.com
  AI Agents in Action, Second EditionGrounding is the process of anchoring a model's responses in relevant external information, data, or context.
• 

  liveBookmanning.com
  Building Reliable AI SystemsA technique for anchoring an AI system's outputs in external information or evidence.

Compare how 8 books explain grounding →

Related concepts

• SIMBAuses · SIMBA is a DSPy prompt optimizer that iteratively evaluates candidate prompts, reflects on execution trajectories, and adds demonstrations or rules to improve performance.
• MIPROv2uses · MIPROv2 is a DSPy optimizer that searches for effective instructions and demonstrations for language-model programs using performance feedback and a Bayesian proposal model.
• GEPAuses · GEPA (Genetic-Pareto) is a reflective evolutionary optimization method that generates and evaluates candidate prompts using task information and feedback.
• hallucinationaddresses · A hallucination is model-generated content that is false, fabricated, or unsupported by relevant evidence but presented as plausible or factual.
• retrieval-augmented generationproduced by · Retrieval-augmented generation is a technique that retrieves relevant information from external sources and supplies it as context to a generative model when producing an output.

This explainer was generated by Manning's knowledge graph from the text of Building LLM Applications with DSPy. The section links point to the parts of the book it draws on.