Blog

Building true RL systems: An experiment on solving real business tasks

We tested rubric-based rewards and GRPO on a real-world e-commerce task and found they outperformed sparse rewards by 300%. This helps validate their effectiveness for complex, multi-step business workflows.

Agentic AI: What it takes to build AI that acts

Agentic AI is emerging as a new frontier in autonomy, where models can plan, adapt, and take action independently. In this post we highlight three real-world projects with leading AI labs, from multi step tool use to structured reasoning and dynamic instruction following.

Benchmarking agentic search

Enterprises need search-augmented LLMs that deliver fast, trustworthy, and up-to-date answers—not just polished language. Since public benchmarks rarely test for this, the Labelbox research team conducted its own study across three frontier models: Gemini 2.5 Pro, GPT-4.1, and Claude 4.0 Opus.

Rubric evaluations: Fueling the next wave of reinforcement learning

See how Labelbox utilizes custom rubric-based evaluations to help leading AI labs train and assess advanced frontier models with depth and nuance.

Prompt to production: How to improve AI app generators with rubric evals

Discover how modern rubric-based evaluations and human evaluation are crucial for advancing the capabilities of prompt-to-app and AI app generators.

How to fill your RLVR pipeline with advanced reasoning data

Learn how to fill your Reinforcement Learning with Verifiable Rewards (RLVR) pipelines to teach your models effective reasoning, especially for logic, math, and coding problems with clear solutions.

Reinventing AI evaluation: Discover the simplicity of Labelbox's new form-based MMC editor

Discover Labelbox's redesigned Multimodal Chat editor, offering an intuitive, form-based experience for streamlined AI model evaluation and data generation.

New Complex Reasoning Leaderboard: Gemini 2.5 debuts at the top

Labelbox introduces a new Complex Reasoning Leaderboard, ranking Google's Gemini 2.5 Pro as the top AI model for advanced reasoning tasks.

Introducing a powerful, new interactive Workflow editor

Learn about the new Labelbox Workflow that introduces an interactive, node-based editor to create, manage, and visualize multi-step review workflows.

Q1 spotlight: Accelerating AI development with new products and services

Catch up on Labelbox's latest news from Q1, including expanded Leaderboards, the Alignerr Connect launch, and platform advancements empowering the next generation of AI models.