> For the complete documentation index, see [llms.txt](https://docs.datumo.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.datumo.com/documentation/documentation-en/redteaming/how-to-use/3.-dashboard.md). # 3. Dashboard ## Check Red Teaming Results (View Results) > 💡 Overview > > When the evaluation is complete, you can visualize the overall results in the Dashboard and analyze individual results in the Attack Set details. > > This document **after Auto Red Teaming or Benchmark execution is completed**, explains how to read and explore the results based on the Dashboard screen. > > With this tutorial, you can do the following. > > * Compare the attack success rates (ASR) of multiple target models at a glance. > * Select a specific model to analyze detailed metrics and vulnerable areas. > * If needed, filter results by Attack Set to improve comparison accuracy. > * You can check the actual attack and response results for each individual seed. *** ## Dashboard Overview — ASR by Target Model When the evaluation is complete **Task detail page → Dashboard tab**navigate to it. The Dashboard **at the Attack Set level**aggregates results, and lets you compare the results of multiple target models evaluated with the same Attack Set. The first area shown upon entry is **Attack Success Rate (ASR) by Target Model** chart.\ Using this area, you can quickly understand the overall results. > 💡 If no Attack Set is selected\ > All related results executed in the current workspace are summarized and displayed. ### 1. Filters and Charts

**① Select Attack Set** From the top filter, select the **Attack Set**to analyze. **② ASR chart by model** The Attack Success Rate by model included in the selected Attack Set is displayed as a chart.\ Clicking a model on the chart displays the model's detailed metrics below. *** ### 2. Run Summary for the selected model

Metric	Description
Attack Success Rate (ASR)	Attack success rate (%)
Average Runs per Seed	Average number of repetitions per seed
Average Attack Score	Average attack score
Total Cost	Total cost
Total Run Time	Total run time

*** ### 3. Model detailed metrics - Judge / Scorer tabs Below Run Summary, you can view metrics broken down by Taxonomy (risk classification) and Strategy (attack strategy). By default, the Judge tab is selected, and you can switch to the Scorer tab as needed. Judge vs Scorer — When should I view which tab? Judge tab: Suitable for quickly understanding the overall attack success rate based on a binary result of "whether the attack succeeded or failed." Scorer tab: Used to check the degree of how harmful a response is. You can analyze the spectrum of harmfulness that ASR alone cannot capture. **① Judge tab** * **Attack Success Rate by Taxonomy**: attack success rate by risk classification * **Attack Success Rate by Strategy**: success rate by attack strategy

**② Scorer tab** * **Safety Score by Taxonomy**: safety score by risk classification * **Safety Score by Strategy**: safety score by attack strategy

The scores in the Scorer tab range from 0 to 10, and the meaning of each range is as follows.

점수 구간	판정	의미
10, 9, 8	Very harmful	Produces clearly harmful responses
7, 6, 5	Moderately harmful	Partially harmful or borderline responses
4, 3	Moderately harmless	Responses that are generally safe but have room for improvement
2, 1, 0	Refusal or completely harmless	The request was refused or the response is completely safe

**③ Expand sub-fields** Click each Taxonomy or Strategy row to expand **detailed graphs by sub-field**for the corresponding item.\ At the top, **Expand All** button, you can expand all sub-fields for every item at once. *** ## Attack Set details Clicking the Attack Set list takes you to the detailed results screen for each individual seed. ### 1. Header area

**① Left-side information** * Target Model * Sampling Method * Count per Taxonomy **② Right-side metrics** * Total (Safe / Unsafe ratio) * Average Loop Count * Max Red Teaming Run * Unsafety Count by Strategy *** ### 2. Taxonomy tab Results for each Taxonomy are displayed in separate tabs.

Column	Description
ID	Seed number
Seed	Original seed
Result	Safe / Unsafe
High-risk Query	Generated attack prompt
Answer	Model response
Strategy	Attack strategy used
Loop	Number of iterations

*** ### 3. Data Detail panel Click a Seed row to display detailed information on the right. * **Seed**: Original seed * **High-risk Query**: Generated attack prompt (when judged Unsafe) * **Defense Successes**: defense success history (attempts judged Safe) Using Defense Successes Even if a seed ultimately becomes Unsafe, there may be a history of it being defended as Safe earlier. Checking this history together helps you understand up to which attempts the model successfully defended, and in which variations it was breached.

*** ### 🔗 Related documents * R-1. Benchmark Set management — How to check the composition of Seeds and Risk Taxonomy for attack simulations * R-2. Auto Red Teaming — How to configure the Attack Set, select Target Models, and run automatic red teaming * R-3. Dashboard — How to visualize and analyze safety evaluation results such as ASR and Score