add icml 2026 news and update demyst figure and abstract

content/news/2026-05-13/index.md

@@ -0,0 +1,38 @@
+---
+title: "Papers accepted at ICML 2026"
+
+event: ""
+event_url: ""
+
+authors:
+- rodola
+
+summary: ""
+abstract: ""
+
+
+date: '2026-05-13T17:00:00Z'
+
+all_day: true
+
+publishDate: '2026-05-13T17:00:00Z'
+
+tags: []
+
+featured: false
+
+url_code: ''
+url_pdf: ''
+url_slides: ''
+url_video: ''
+slides: ""
+
+# Projects (optional).
+#   Associate this post with one or more of your projects.
+#   Simply enter your project's folder or file name without extension.
+#   E.g. `projects = ["internal-project"]` references `content/project/deep-learning/index.md`.
+#   Otherwise, set `projects = []`.
+projects: []
+---
+
+Our three papers ["*PHALAR: Phasors for Learned Musical Audio Representations*"](https://gladia.di.uniroma1.it/publication/2026-marincione-phalar/), ["*Demystifying Mergeability: Interpretable Properties to Predict Model Merging Success*"](https://gladia.di.uniroma1.it/publication/2026-zhou-demyst/), and ["*Multi-Way Representation Alignment*"](https://gladia.di.uniroma1.it/publication/2026-achara-multiway/) were accepted at **ICML 2026** main conference!

content/publication/2026-zhou-demyst/index.md

@@ -28,7 +28,7 @@ projects: []
 publishDate: '2025-27-02T:26:44'
 publication_types:
 - '1'
-abstract: "Model merging combines knowledge from separately fine-tuned models, yet success factors remain poorly understood. While recent work treats mergeability as an intrinsic property, we show with an architecture-agnostic framework that it fundamentally depends on both the merging method and the partner tasks. Using linear optimization over a set of interpretable pairwise metrics (e.g., gradient L2 distance), we uncover properties correlating with post-merge performance across four merging methods. We find substantial variation in success drivers (46.7% metric overlap; 55.3% sign agreement), revealing method-specific 'fingerprints'. Crucially, however, subspace overlap and gradient alignment metrics consistently emerge as foundational, method-agnostic prerequisites for compatibility. These findings provide a diagnostic foundation for understanding mergeability and motivate future fine-tuning strategies that explicitly encourage these properties."
+abstract: "Model merging combines knowledge from separately fine-tuned models, yet the factors driving its success remain poorly understood. While recent work treats mergeability as an intrinsic property of the models, we show with an architecture-agnostic framework that it fundamentally depends on both the merging method and the partner tasks. Using L1-regularized linear optimization over a set of interpretable pairwise metrics (e.g., gradient L_2 distance), we uncover properties correlating with post-merge normalized accuracy across five merging methods. We find architecture- and method-specific variation in success drivers (64.0% average top-5 metric overlap; 79.3% sign agreement), with certain methods, notably TIES, exhibiting distinct ``fingerprints'' that diverge from the broader consensus. Crucially, however, gradient alignment metrics consistently emerge as the most fundamental signals of compatibility. These findings provide a diagnostic foundation for understanding mergeability and motivate future merge-aware fine-tuning strategies."
 
 links:
 - name: arXiv