MAPIE Roadmap 2026

[GBrelurut]

Dear Contributors,

📆 As we approach the end of 2025, we reflect on our accomplishments and look forward to an exciting year ahead. We extend our warmest wishes for a successful conclusion to this year and an inspiring start to 2026.

🗺️ To harness this momentum, we have outlined a roadmap that focuses on key initiatives for the upcoming period. Here’s a glimpse of what lies ahead:

Phase 1: Complete the binary classification control process, culminating in an application of LLM-as-Judge.

Status : nearly done (expected for end of December 2025) ☑️

Phase 2: Strengthen and standardize the risk control aspect of multi-label classification, with a key application focused on image segmentation.

Status : on-going (should be terminated beginning of 2026) 🏃

Phase 3: Add exchangeability tests to enable users to confidently verify the legitimate application of MAPIE

Status : expected for mid 2026 ⏳

Phase 4: Revisit conformal predictions with an emphasis on adaptability (typically with Gibbs or Mondrian methods)

Status : expected for end of 2026 ⌛

We invite you to actively participate in these initiatives, as your expertise and insights have already proven crucial to this project.
Thank you for your dedication and collaboration as we embark on this journey into 2026! 🎊