Launch Week 3: Five days of launches
Confident AI
Docs
Pricing
Products
LLM Evaluation
LLM Evaluation

Benchmark LLM systems with research-backed metrics.

LLM Observability
LLM Observability

Trace, monitor, and alert on production LLM systems.

AI Red Teaming
AI Red Teaming

Stress-test LLM apps against adversarial attacks.

AI Governance
AI Governance

Enforce AI standards and controls across teams.

Book a DemoSign Up
Book a DemoSign Up

Image Editing

Image Editing is a single-turn image metric to determine if an image was edited as instructed

Overview

The image editing metric is a single-turn image metric that uses MLLM-as-a-judge to assess how well your LLM application edited an image to match the instruction it was given.

Required Parameters

These are the parameters you must supply in your test case to run evaluations for image editing metric:

inputlist of string | MLLMImageRequired

The editing instruction and the original image you supplied to your image editing application.

actual_outputlist of string | MLLMImageRequired

The edited image your image editing application generates for the given input.

How Is It Calculated?

The image editing metric uses an MLLM to score the edited image on semantic consistency against the editing instruction and on perceptual quality, then combines the lowest score of each group into a single score.


Image Editing=min(α1,…,αi)⋅min(β1,…,βi)\text{Image Editing} = \sqrt{\text{min}(\alpha_1, \ldots, \alpha_i) \cdot \text{min}(\beta_1, \ldots, \beta_i)}

The final score is the square root of the product of the minimum semantic consistency score and the minimum perceptual quality score.

Usage

To run the image editing metric on Confident AI, add it to a single-turn metric collection. This will allow you to use image editing metric for:

  • Single-turn E2E testing
  • Single-turn component-level testing
  • Online and offline evals for traces and spans

Last updated on

Built byConfident AI