AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Demystifying Mergeability: Interpretable Properties to Predict Model Merging Success

Luca Zhou, Bo Zhao, Rose Yu, Emanuele Rodolà

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add gradient-based metrics (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/16avg change in Pearson r: -0.149Improved
add gradient-based metrics (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/32avg change in Pearson r: -0.131Improved
add subspace-overlap metrics (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/16avg change in Pearson r: -0.045Improved
add effective-rank metrics (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/16avg change in Pearson r: +0.020No clear effect
add activation metrics (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/32avg change in Pearson r: -0.059Improved
add L1 regularization (lambda=1.0) (ablation)linear mergeability predictor · LOTO validation, ViT-B/16Pearson r: 0.525 -> 0.544Improved
add L1 regularization (lambda=1.0) (ablation)linear mergeability predictor · LOTO validation, ViT-B/32Pearson r: 0.676 -> 0.648Hurt
add L1 regularization (lambda=1.0) (ablation)linear mergeability predictor · fold-to-fold std of validation r, ViT-B/16std of Pearson r: 0.254 -> 0.159Improved
use linear L1 predictor instead of MLP (ablation)mergeability predictor · LOTO validation, ViT-B/16Pearson r: 0.493 -> 0.544Improved
use Pearson-correlation objective instead of MSE (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/16Pearson r: 0.147 -> 0.544Improved
use Pearson-correlation objective instead of MSE (ablation)L1 linear mergeability predictor · LOTO validation, ViT-B/32Pearson r: 0.143 -> 0.648Improved
fine-tune with gradient-magnitude penaltytask-specific ViT-B/16 models · post-merge normalized accuracyaverage delta: +0.41Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy