SOTAVerified

Task Arithmetic

A task vector specifies a direction in the weight space of a pre-trained model, such that movement in that direction improves performance on the task. We build task vectors by subtracting the weights of a pre-trained model from the weights of the same model after fine-tuning on a task. We show that these task vectors can be modified and combined together through arithmetic operations such as negation and addition, and the behavior of the resulting model is steered accordingly.

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–50 of 61 papers

TitleStatusHype
Task Singular Vectors: Reducing Task Interference in Model MergingCode2
Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task ArithmeticCode2
Editing Models with Task ArithmeticCode2
Localizing Task Information for Improved Model Merging and CompressionCode2
Parameter Efficient Multi-task Model Fusion with Partial LinearizationCode1
NegMerge: Consensual Weight Negation for Strong Machine UnlearningCode1
Model Merging by Uncertainty-Based Gradient MatchingCode1
Merging Multi-Task Models via Weight-Ensembling Mixture of ExpertsCode1
Concrete Subspace Learning based Interference Elimination for Multi-task Model FusionCode1
AdaMerging: Adaptive Model Merging for Multi-Task LearningCode1
Localize-and-Stitch: Efficient Model Merging via Sparse Task ArithmeticCode1
An Empirical Study of Multimodal Model MergingCode1
Knowledge Composition using Task Vectors with Learned Anisotropic ScalingCode1
Have You Merged My Model? On The Robustness of Large Language Model IP Protection Methods Against Model MergingCode1
Fine-Tuning Attention Modules Only: Enhancing Weight Disentanglement in Task ArithmeticCode1
Task Arithmetic in the Tangent Space: Improved Editing of Pre-Trained ModelsCode1
CALM: Consensus-Aware Localized Merging for Multi-Task LearningCode0
Cross-Model Transfer of Task Vectors via Few-Shot Orthogonal AlignmentCode0
Efficient Model Editing with Task-Localized Sparse Fine-tuningCode0
Efficient Model Editing with Task Vector Bases: A Theoretical Framework and Scalable ApproachCode0
Investigating Task Arithmetic for Zero-Shot Information RetrievalCode0
Leveraging Submodule Linearity Enhances Task Arithmetic Performance in LLMsCode0
Multi-Task Model Merging via Adaptive Weight DisentanglementCode0
No Train but Gain: Language Arithmetic for training-free Language Adapters enhancementCode0
Towards Diverse Device Heterogeneous Federated Learning via Task Arithmetic Knowledge IntegrationCode0
Language and Task Arithmetic with Parameter-Efficient Layers for Zero-Shot Summarization—0
DuET: Dual Incremental Object Detection via Exemplar-Free Task Arithmetic—0
Layer-Aware Task Arithmetic: Disentangling Task-Specific and Instruction-Following Knowledge—0
Transferring Visual Explainability of Self-Explaining Models through Task Arithmetic—0
Disentangling Task Interference within Neurons: Model Merging in Alignment with Neuronal Mechanisms—0
CultureMERT: Continual Pre-Training for Cross-Cultural Music Representation Learning—0
Mediator: Memory-efficient LLM Merging with Less Parameter Conflicts and Uncertainty Based Routing—0
CAT Merging: A Training-Free Approach for Resolving Conflicts in Model Merging—0
MetaGPT: Merging Large Language Models Using Model Exclusive Task Arithmetic—0
MCU: Improving Machine Unlearning through Mode Connectivity—0
What Matters for Model Merging at Scale?—0
Bias Vector: Mitigating Biases in Language Models with Task Arithmetic Approach—0
Neural Networks Remember More: The Power of Parameter Isolation and Combination—0
When Domain Generalization meets Generalized Category Discovery: An Adaptive Task-Arithmetic Driven Approach—0
On Fairness of Task Arithmetic: The Role of Task Vectors—0
On Giant's Shoulders: Effortless Weak to Strong by Dynamic Logits Fusion—0
OpenThaiGPT 1.6 and R1: Thai-Centric Open Source and Reasoning Large Language Models—0
Beyond Task Vectors: Selective Task Arithmetic Based on Importance Metrics—0
Scalable Strategies for Continual Learning with Replay—0
Single-Input Multi-Output Model Merging: Leveraging Foundation Models for Dense Multi-Task Learning—0
Soup to go: mitigating forgetting during continual learning with model averaging—0
Subspace-Boosted Model Merging—0
Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition—0
Task Arithmetic for Language Expansion in Speech Translation—0
BADTV: Unveiling Backdoor Threats in Third-Party Task Vectors—0
Show:102550
← PrevPage 1 of 2Next →

No leaderboard results yet.