Two nice updates from OmnAI Lab: a paper accepted at COLM 2026, and we are organizing the 3rd Unlearning and Model Editing (U&Me) workshop at ECCV'26 🎉

Paper accepted at COLM 2026

Our work on building efficient reasoning models through evolutionary model merging was accepted at the 3rd Conference on Language Modeling (COLM 2026)! The method, Evo-L2S, casts Long-to-Short reasoning as a multi-objective optimization problem, producing a Pareto front of merged models that trade off accuracy and output length.

• M. Iacobelli, A. Minut, T. Mencattini, D. Crisostomi, A. Santilli, I. Masi, E. Rodolà - Multi-objective Evolutionary Merging Enables Efficient Reasoning Models, COLM 2026

In collaboration with the GLADIA Lab 🤝

Organizing the 3rd U&Me workshop at ECCV'26

We are also excited to organize the 3rd Unlearning and Model Editing (U&Me) workshop, this time at ECCV 2026! After the 1st edition at ECCV'24 (Milan) and the 2nd edition at ICCV'25 (Hawaii), the workshop continues to bring together research on efficient and effective techniques for editing trained models — unlearning, model editing and stitching, model compression, efficient domain adaptation, and responsible U&ME (robustness, fairness, privacy, and regulatory compliance).

More info at https://sites.google.com/view/u-and-me-workshop/