Skip to main content
New Top-K selection strategies

Build better GenAI for with human feedback

Evaluate your models on metrics that truly matter, using our self-serve platform and ready-to-use subjective studies.

Trusted by researchers at

Image and video benchmarking

Transform opinions into evidence.

Create benchmarks by measuring preferences using our growing library of highly configurable studies.

vs

Pairwise preferences

Collect unbiased preferences through side-by-side comparisons.

Surveys

Build surveys to gather qualitative feedback on your model's outputs.

Rubrics and learboards

Compile the results of your evaluations into a leaderboard to track progress over time.

Generated voice sample

How appropriate is this voice?

Poor
Fair
Good
Excellent
Submit

Localization

Interfaces localized into 13 languages. More on request.

Fluent raters

Raters fluent in many languages and dialects.

Filter out the noise

Find raters who perform well on your task.

Audio evaluation

Know the limits of your audio model.

Discover failure modes of your audio pipeline before your users do.

Evaluation infrastructure

Integrate evaluation into your pipeline.

Upload datasets, configure experiments and retrieve results programmatically, without leaving your existing workflow.

create_experiment.py Experiment created
1
2
3
4
5
6
7
8
9
10
11
12
13
response = requests.post(
    f"https://api.mabyduck.com/projects/
    {project_id}/experiments/",
    json={
        "name": "My experiment",
        "type": "pairwise_video",
        "config": {
            "allowTie": False,
            "hiddenReference": True,
        },
    },
    headers=headers,
)

RLHF

Define large-scale experiments to collect human feedback for model fine-tuning.

Regression tests

Create repeatable evaluations to test model performance over time and across versions.

Automated workflows

Use webhooks to automatically respond when data collection completes.

Embedded experiments

Evaluate interactive experiences.

Embed interactive avatars, 3D worlds, videos and custom interfaces directly into a study to collect structured feedback.

Any content

Display any content by embedding JavaScript or external sites.

Session recordings

Replay interactions to get a deeper understanding of the user experience.

Debugger

Debug implementations directly in the browser.

Ready to get started?

Start collecting human feedback today.

Create your first evaluation in minutes, or talk to us about a study tailored to your needs.