Label Studio 1.12.0 🚀Automate & Evaluate Labeling Predictions Using LLMs & ML Models

December 2023 Community News

Label Studio Team

🎊 Happy New Year from Label Studio!

Thank you to everyone in the Label Studio Community for helping to make Label Studio the most popular open-source data labeling and annotation platform. 2023 was a year of highlights for the community, including:

🪄 The release of the community-developed Magic Wand Tool.

🔌 The launch of the Label Studio Integrations Directory.

🙋‍♀️ Publishing new labeling templates and tools for RLHF workflows.

🤗 The availability of Label Studio in Hugging Face Spaces.

🧩 Integration of Segment Anything with Label Studio.

🚀 Creating a new open source project: Adala, an agent-based data labeling platform.

You can read about these announcements, and many more, in the Label Studio blog archives. As a special holiday treat, we’ve also prepared two new articles for you to read and demo between your end-of-year festivities.

🦕 Grounding DINO x Label Studio

In this article, we explore how to integrate Grounding DINO (and, as a bonus, Grounding SAM) into your image labeling workflow. Grounding DINO is a zero-shot object detection model that combines a DINO transformer architecture with grounded pre-training. The result is a model that responds to text prompts for object identification within images.

We’ll show you how to integrate this powerful language-vision model into your Label Studio image annotation workflow!

Text to Image Labels with Grounding DINO

🧮 Mastering Math Reasoning with Adala

The second article, Mastering Math Reasoning with Adala, is a technical exploration of how agent-based techniques can substantially improve 30% over the baseline performance of GPT-4 in the GSM8K benchmark. Impressively, this leap in performance was achieved without the necessity of manual prompt engineering, highlighting the efficacy of agents to automatically guide prompt engineering to optimized results.

We’ll show you how to accomplish these results on your own with Adala and give you some insights into how you can apply Adala to your own data labeling projects.

Mastering Math Reasong with Adala

🌟 Annotations

4️⃣ MIT’s The Algorithm Newsletter wraps up 2023 with a list of the Four Trends that Changed AI in 2023. One theme that holds steady across all of their top trends is how disruptive Generative AI and Foundation Models have been across the AI landscape and technology in general.

🪙 For centuries, coin flipping has been the most simple and popular way of generating random events. A new research paper from František Bartoš et al. calls into question how fair these seemingly fair flips are, with their research suggesting that “Fair coins tend to land on the same side they started.” The authors back this claim with evidence from more 350,000 flips, and have published the entire dataset — including videos of study participants flipping coins.

🎇 Thank You For Being Part of the Community!

Everyone from the Label Studio team wishes you the most joyous of Holiday Seasons and a Happy New Year!

Happy Labeling!

Join The Community Slack Channel!

Related Content