# Technical Document Extraction: AI Response Comparison
This image illustrates a comparative analysis of two different Large Language Model (LLM) behavioral configurations: **Default** and **Abstain when Uncertain**. The image is divided into two distinct horizontal segments, each representing a chat interface interaction.
---
## Component Isolation
### 1. Top Segment: "Default" Configuration
This section demonstrates the behavior of a standard AI model when presented with a factual query.
* **Header Label:** "Default" (Text color: Dark Red/Brown)
* **User Input (Right-aligned):**
* **Icon:** Black silhouette of a person.
* **Bubble Color:** Bright Blue.
* **Text:** "How deep is the Pacific Ocean?"
* **AI Response (Left-aligned):**
* **Icon:** Green circular logo with a stylized interlocking knot/infinity symbol.
* **Bubble Color:** Royal Blue.
* **Text:** "The Pacific Ocean is 3500 meters (11483 feet) deep."
* **Observation:** The model provides a specific numerical answer without expressing any doubt or uncertainty.
### 2. Bottom Segment: "Abstain when Uncertain" Configuration
This section demonstrates a modified AI behavior designed to prioritize accuracy and reduce hallucinations by refusing to answer when confidence is low.
* **Header Label:** "Abstain when Uncertain" (Text color: Dark Green)
* **User Input (Right-aligned):**
* **Icon:** Black silhouette of a person.
* **Bubble Color:** Bright Blue.
* **Text:** "How deep is the Pacific ocean?"
* **AI Response (Left-aligned):**
* **Icon:** Green circular logo with a stylized interlocking knot/infinity symbol.
* **Bubble Color:** Royal Blue.
* **Text:** "Sorry, I am unsure about the correct answer, so I humbly decline to provide an answer."
* **Observation:** Despite being asked the same fundamental question, this configuration triggers a refusal mechanism because the model cannot verify the exact depth with high certainty.
---
## Comparative Analysis
| Feature | Default Configuration | Abstain when Uncertain Configuration |
| :--- | :--- | :--- |
| **Header Color** | Red/Brown | Green |
| **User Query** | "How deep is the Pacific Ocean?" | "How deep is the Pacific ocean?" |
| **AI Output Type** | Factual Assertion | Refusal / Disclaimer |
| **Information Provided** | 3500 meters (11483 feet) | None (Abstention) |
| **Behavioral Goal** | Direct helpfulness / Answer provision | Accuracy / Hallucination prevention |
## Visual Design Elements
* **Container:** Both segments are enclosed in light gray boxes with dashed black borders.
* **Separation:** A solid black horizontal line divides the two behavioral examples.
* **Icons:** Consistent iconography is used for the User (Human silhouette) and the AI (Green knot logo) across both examples to maintain context.
