TATR improves multi-task model merging by restricting task-vector additions to parameter dimensions that are orthogonal to task-specific loss gradients.
Just pick a sign: Optimizing deep multitask models with gradient sign dropout
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Task Arithmetic in Trust Region: A Training-Free Model Merging Approach to Navigate Knowledge Conflicts
TATR improves multi-task model merging by restricting task-vector additions to parameter dimensions that are orthogonal to task-specific loss gradients.