Mastering Task Arithmetic: τJp as a Key Indicator for Weight Disentanglement

Kotaro Yoshida, Yuji Naraki, Takafumi Horie, Ryosuke Yamaki, Ryotaro Shimizu +3 more
2/12/2026

No official code found. Best implementation: katoro8989/tau-Jp_Task_Arithmetic (3 stars, Apache-2.0, last commit 2025-08-01).

Semantic Scholar

Code Implementations(1)

katoro8989/tau-Jp_Task_ArithmeticCommunity70%
Apache-2.0

Mastering Task Arithmetic: Jp as a Key Indicator for Weight Disentanglement (ICLR2025)

30PythonSep 25, 20241 years agoApache-2.0

Similar Papers

Mixed Samples as Probes for Unsupervised Model Selection in Domain Adaptation

91% match2026

Variance as a Catalyst: Efficient and Transferable Semantic Erasure Adversarial Attack for Customized Diffusion Models

90% match1 repo62026

Pfeife: Automatic Pipeline Parallelism for PyTorch

88% match1 repo62026

Diffusion Posterior Sampling for Linear Inverse Problem Solving: A Filtering Perspective

88% match1 repo502026

MODeL: Memory Optimizations for Deep Learning

88% match2026

DISCS: A Benchmark for Discrete Sampling

88% match2026

Cite this paper

@article{yoshida2026mastering,
  title  = {Mastering Task Arithmetic: τJp as a Key Indicator for Weight Disentanglement},
  author = {Kotaro Yoshida and Yuji Naraki and Takafumi Horie and Ryosuke Yamaki and Ryotaro Shimizu and Yuki Saito and Julian J. McAuley and Hiroki Naganuma},
  year   = {2026},
  url    = {https://api.semanticscholar.org/CorpusID:180e8815a8528162d4b950d769bd8eac037ddafe},
  journal = {ICLR 2025 2025}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/180e8815a8528162d4b950d769bd8eac037ddafe.svg)](https://codeofpaper.com/paper/180e8815a8528162d4b950d769bd8eac037ddafe)

Discussion