Improve multimodal evaluations with enhanced arena style experience

We are excited to announce a brand new look-and-feel to our chat arena style capabilities in the Labelbox platform. Based on feedback from leading AI labs and our growing community of highly-skilled Alignerrs, we reimagined the experience when comparing multimodal models. The enhancements help accelerate labeling workflows, increase the efficiency of our labeling services, and further elevate data quality.

Key updates now available to all customers include a simplified layout for model response comparison, quick access to per-message classifications, guided navigation based on your custom ontology, clear identification of any missed labeling tasks, and more. 

These enhancements empower both frontier model builders and enterprise AI teams to rapidly evaluate models and generate high-quality, multimodal datasets for a variety of AI post-training tasks, including RLHF/DPO, SFT, and red teaming.

Multimodal chat evaluation: A central hub for human data generation

The Labelbox multimodal chat editor serves as a dynamic workspace where subject matter experts and labelers can simultaneously interact directly with up to 10 different AI models in real-time, multi-turn conversations. Supported models the latest foundation models from Meta, Google, OpenAI, Anthropic, and more.

The tool helps build post-training GenAI datasets for a wide range of data modalities, including text, images, videos, audio, and documents. It can generate and refine high-quality training data or analyze model performance through activities like preference ranking. Users can rate, rank, rewrite, or classify model responses to fuel AI advancements.

Interested in exploring the tool through a quick, interactive demo? Explore a quick 60-second product tour to learn how to get started with the multimodal chat editor without having to sign-in or set-up anything. 

Redesigned user interface to accelerate and improve data creation

Labelbox's latest updates are the direct result of listening closely to customer needs and feedback. These enhancements are tailored to improve productivity, accuracy, and overall user experience.

New product updates include the following:

• Ergonomic, horizontal layout for response comparison: The revamped editor interface is now more ergonomic and expansive, making it easier to complete tasks and view critical information at the same time. The layout has fewer controls and details distracting the user. Model responses are shown side-by-side for easier comparison and to minimize scrolling.

• Integrated per-response classifications: Per-response classifications are now located right next to the relevant responses, making them easy to complete and providing an intuitive first step. 
• Guided user experience: A "Next" button provides clear prompts, guiding users through the workflow and ensuring no steps are missed.
• Enhanced global classifications: Powerful global classifications are now available to labelers after all tasks and messages are complete to ensure the best quality response.

• Built-in automation: The integration of model-assisted labeling (MAL) and ground truth (GT) imports streamlines the process of adding classifications, rankings, and selections, saving valuable time and effort.

The impact: High-quality data, faster

These enhancements to the Labelbox platform aren't just about incremental improvements; they represent an important change needed to accelerate the delivery of high-quality human data.

Labelbox customers stand to gain significant advantages across several key areas:

• Improved efficiency: By streamlining workflows, providing clear guidance, and integrating intelligent automation, Labelbox empowers labelers and raters to work faster and smarter. This translates to faster project completion times and a quicker path to ROI.
• Enhanced data quality: The redesigned interface promotes more accurate and informed labeling decisions. The ability to easily compare model responses side-by-side leads to higher-quality annotations and, ultimately, more reliable and performant AI models.
• Reduced onboarding time: The intuitive interface and guided experience minimize the learning curve for new labelers. This means less time spent on training and more time spent on generating valuable data.

Labelbox is here to help

Labelbox's commitment to innovation is evident in these powerful enhancements to the multimodal chat editor. By continuously refining its platform based on user feedback, Labelbox remains at the forefront of AI data solutions, enabling organizations to unlock the full potential of their AI initiatives.

Enjoy these two easy (and free) ways to explore these new features for yourself: 

• Explore the Multimodal Chat Editor for free: Experience the enhanced capabilities and intuitive interface of the Labelbox multimodal chat editor with free product access.
• View our on-demand product demos: Take a guided, interactive tour and explore the new features in action.

Contact our team anytime with questions or if you are ready to discuss your AI data needs and discover how Labelbox might be able to help.