What’s New in Arthur – July 2025 Edition
Smarter Evals, Smoother UX, and More Powerful Insights
July was a big month for the Arthur Platform, packed with powerful updates designed to make continuous evaluation easier, faster, and more insightful for AI teams.
From multimodal CV support to major UX upgrades, here’s a quick roundup of what dropped:
New Features to Explore
- Multimodal Computer Vision Evals: You can now evaluate and visualize image-based model inferences directly in the Arthur Platform complete with custom metrics.
- Smarter Segmentation: Slice metrics by prompt version, model version, or any attribute you care about. Perfect for comparing experiments and tracking regressions.
- OTEL Traces for LLM/Agentic Apps: Arthur now consumes traces from your AI stack, giving you end-to-end observability.
- Non-Docker Engine Install: Expanded install options now make Arthur Engine easier to deploy in more environments.
Enhanced Performance & Usability
- Faster, Smarter PII Detection: We’ve reduced false positives and improved detection accuracy for sensitive content.
- Improved Hallucination Detection: Especially for structured content like numbered lists (more accurate signals, fewer false flags).
- Token Limit Controls for Hallucination Checks: Fine-tune max token thresholds for better context sensitivity.
Whether you're building with vision, text, or agents Arthur is here to help teams stay in control, experiment faster, and gain trust in what you ship. Get started for free.
Stay tuned for the next wave of updates is just around the corner.
SHARE
