Quality assurance · Production

Leading genAI text-to-image app accelerates product development with Labelbox high-quality labeled data

The problem

The company needed to rapidly generate high-quality human preference training data for its text-to-image AI models, but labeling data internally would have diverted resources from core product development. The complexity and labor-intensive nature of generative AI data labeling required dedicated tooling and teams to meet tight timeframes.

Workflow diagram · grounded in source
1
Training data need identified
trigger
“the company was looking for rapid ways to generate high-quality human preference training data”
2
Labelbox labelers engaged
integration
“delivering highly-skilled labelers along with project management experts that allowed them to meet their deadlines”
3
RLHF preference labeling
human_review
“the company used Labelbox's LLM human preference editor to perform reinforcement learning from human feedback (RLHF), which enabled them to easily create preference data for training a reward model based on multiple outputs from a single…”
4
Multimodal model comparison
human_review
“leverage Labelbox's latest multimodal chat solution to improve throughput for creating additional data that would fix where their models were underperforming. This was done by comparing model outputs on text-to-image prompts versus other…”
5
Data engine weekly evaluation
validation
“implement a data engine workflow which allowed them to compare multiple models against each other in live, multi-turn conversations and rank outputs for a more data-driven evaluation of models. This human evaluation approach would then h…”
6
Targeted training iteration
feedback_loop
“target training data creation towards the areas which needed the most improvement and which would have the most impact on model performance, and iteratively fed back and verified their results through another human evaluation run”
Reported outcome

Using Labelbox, the company improved the speed of creating human preference training data by 2x and accelerated product development from months to weeks.

Reported metrics
Human preference training data creation speed (result summary)2x
Human preference training data creation speed (body text)over 50%
Product development cycle timemonths to weeks
Reported stack
LabelboxLabelbox Labeling ServicesLLM human preference editormultimodal chat solution
Source
https://labelbox.com/customers/text-to-image-app-customer-story
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Using Labelbox, the company improved the speed of creating human preference training data by 2x and accelerated product development from months to weeks.

What tools did this team use?

Labelbox, Labelbox Labeling Services, LLM human preference editor, multimodal chat solution.

What results were reported?

Human preference training data creation speed (result summary): 2x; Human preference training data creation speed (body text): over 50%; Product development cycle time: months to weeks (source-reported, not independently verified).

How is this quality assurance AI workflow structured?

Training data need identified → Labelbox labelers engaged → RLHF preference labeling → Multimodal model comparison → Data engine weekly evaluation → Targeted training iteration.

Related quality assurance cases