REVIEW 8 cited by
Twin-Merging: Dynamic Integration of Modular Expertise in Model Merging
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
abstract
In the era of large language models, model merging is a promising way to combine multiple task-specific models into a single multitask model without extra training. However, two challenges remain: (a) interference between different models and (b) heterogeneous data during testing. Traditional model merging methods often show significant performance gaps compared to fine-tuned models due to these issues. Additionally, a one-size-fits-all model lacks flexibility for diverse test data, leading to performance degradation. We show that both shared and exclusive task-specific knowledge are crucial for merging performance, but directly merging exclusive knowledge hinders overall performance. In view of this, we propose Twin-Merging, a method that encompasses two principal stages: (1) modularizing knowledge into shared and exclusive components, with compression to reduce redundancy and enhance efficiency; (2) dynamically merging shared and task-specific knowledge based on the input. This approach narrows the performance gap between merged and fine-tuned models and improves adaptability to heterogeneous data. Extensive experiments on $20$ datasets for both language and vision tasks demonstrate the effectiveness of our method, showing an average improvement of $28.34\%$ in absolute normalized score for discriminative tasks and even surpassing the fine-tuned upper bound on the generative tasks. Our implementation is available in \url{https://github.com/LZY-the-boys/Twin-Merging}
Forward citations
Cited by 8 Pith papers
-
Continual Learning in Transition
Continual learning is reframed as continual capability evolution along three axes: when learning happens, where capability is stored, and how updates are driven.
-
Unraveling LoRA Interference: Orthogonal Subspaces for Robust Model Merging
Initializing LoRA's A matrix with the smallest-eigenvalue eigenvectors of other tasks' feature covariance reduces interference when merging task-specific LoRA models, improving average merged accuracy.
-
Decouple and Orthogonalize: A Data-Free Framework for LoRA Merging
A data-free LoRA merging framework that decouples weight magnitude from direction and orthogonalizes directions to reduce task interference, outperforming existing merging methods across vision, language and multimoda...
-
Task Vector Bases: A Unified and Scalable Framework for Compressed Task Arithmetic
Task Vector Bases compresses T task vectors into M softmax-mixed basis vectors that preserve task arithmetic operations, with empirical gains over PCA and random selection.
-
FREE-Merging: Fourier Transform for Efficient Model Merging
High-pass filtering of task vectors in the Fourier domain reduces task interference in model merging, and lightweight rescaling experts recover lost performance.
-
Intrinsic Strain-Driven Topological Evolution in SrRuO3 via Flexural Strain Engineering
The abstract reports a 21% anomalous Hall conductivity increase in flexurally strained SrRuO3, but the submitted full text belongs to a different machine learning paper.
-
CITER: Collaborative Inference for Efficient Large Language Model Decoding with Token-Level Routing
CITER trains a token-level router with preference optimization to route non-critical tokens to a small model and critical tokens to a large model, reducing inference cost on QA and math benchmarks.
-
LLaMA-MoE v2: Exploring Sparsity of LLaMA from Perspective of Mixture-of-Experts with Post-Training
A dense instruction-tuned LLaMA-3-8B can be partitioned into Attention-MoE and MLP-MoE and re-trained with about 7B instruction tokens, recovering usable math/code skills with half the activated parameters.
Discussion (0). Continue with ORCID to comment.