The page can now chart selected metrics side by side across experiment runs, including quality, cost, and latency metrics. Users can customize the chart layout.
For a candidate run compared with a baseline, Langfuse now provides summary statistics for each score across runs. The comparison table can filter items below a chosen score threshold within one run or across runs, and users can choose displayed metric columns and row height for long outputs.
The update also adds in-context review actions while comparing runs: a human score, corrected output, or comment can be added without leaving the comparison, with keyboard shortcuts for stepping through items. The changelog does not establish the availability tier, pricing, performance impact, or whether these functions are available in self-hosted deployments.