Arize AI Community Icon

Home
Events

Arize AX Releases
Arize AX Support
Arize News
Arize Observe
Discussions
Introductions
Phoenix Support

Powered by Tightknit

Phoenix Support

Phoenix Support

How to Add Tags to Your Dataset for Better Experiment Analytics

·Aug 25, 2025 05:42 PM

Howdy! Is there an effective way to add labels/tags to my dataset cases used in experiments? I want to add a tag, such as "adversarial", so I can run analytics on which types of cases perform well vs. still need work.

4 comments

· Sorted by Oldest

RunLLM
APP
·
Yes, Phoenix supports adding tags to dataset examples for cohort analysis. You can add a tag column (e.g., "adversarial") to your pandas DataFrame and specify it in the schema using the tag_column_names parameter. This allows you to filter, group, and analyze performance by tag in the Phoenix UI and during experiments. For example, add a column like adversarial to your DataFrame, then set tag_column_names=["adversarial"] in your schema. Tags are then available for analytics and cohort filtering in the app as described here. Would you like a code example?Sources: - Phoenix: How to define your schema
💡 Hint: Mention RunLLM in the thread for followups.
Jacob M.
·
Amazing! Thank you. 😊
🙌1

·

One follow up question - what's the best way to report correctness/success of the experiment by these labels? Right now, I found docs that point me to this:

import phoenix as px
client = px.Client()

# Get the current dataset version
dataset = client.get_dataset(id="...", version_id="...")

df = dataset.as_dataframe()

df.head()

experiment = client.get_experiment(experiment_id="...")

Jacob M.
·
It seemed to work for me! Thanks!