mlsanityData to 14 Oct 2026

Decompose, Mix, Adapt: A Unified Framework for Parameter-Efficient Neural Network Recombination and Compression

arXiv 2603.27383 ↗First posted 28 Mar 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Parameter Recombination (PR) methods aim to efficiently compose the weights of a neural network for applications like Parameter-Efficient FineTuning (PEFT) and Model Compression (MC), among others. Most methods typically focus on one application of PR, which can make composing them challenging. For example, when deploying a large model you may wish to compress the model and also quickly adapt to new settings. However, PEFT methods often can still contain millions of parameters. This may be small compared to the original model size, but can be problematic in resource constrained deployments like edge devices, where they take a larger portion of the compressed model's parameters. To address this, we present Coefficient-gated weight Recombination by Interpolated Shared basis Projections (CRISP), a general approach that seamlessly integrates multiple PR tasks within the same framework. CRISP accomplishes this by factorizing pretrained weights into basis matrices and their component mixing projections. Sharing basis matrices across layers and adjusting its size enables us to perform MC, whereas the mixer weight's small size (fewer than 200 in some experiments) enables CRISP to support PEFT. Experiments show CRISP outperforms methods from prior work capable of dual-task applications by 4-5% while also outperforming the state-of-the-art in PEFT by 1.5% and PEFT+MC combinations by 1%. Our code is available on the repository: https://github.com/appledora/CRISP-CVPR26.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 71 of its 103 references that are in the map. First posted on arXiv 28 Mar 2026.

Builds on

71 of 103 references.

References · 71

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  4. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,886
  5. 2019HellaSwag: Can a Machine Really Finish Your Sentence?cited by 4,100
  6. 2015Fast R-CNNcited by 2,933
  7. 2019PIQA: Reasoning about Physical Commonsense in Natural Languagecited by 2,883
  8. 2019BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questionscited by 2,183
  9. 2018Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answeringcited by 2,164
  10. 2013Fine-Grained Visual Classification of Aircraftcited by 1,726
  11. 2017Sigmoid-Weighted Linear Units for Neural Network Function Approximation in Reinforcement Learningcited by 950
  12. 2021Learning to Prompt for Continual Learningcited by 944
38 more
  1. 2022Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learningcited by 896
  2. 2024DoRA: Weight-Decomposed Low-Rank Adaptationcited by 758
  3. 2022AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognitioncited by 719
  4. 2024An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Modelscited by 642
  5. 2023LLaMA-VID: An Image is Worth 2 Tokens in Large Language Modelscited by 546
  6. 2019A Large-scale Study of Representation Learning with the Visual Task Adaptation Benchmarkcited by 425
  7. 2024SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inferencecited by 391
  8. 2023SVDiff: Compact Parameter Space for Diffusion Fine-Tuningcited by 338
  9. 2023LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Modelscited by 324
  10. 2022Scaling & Shifting Your Features: A New Baseline for Efficient Model Tuningcited by 294
  11. 2024SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compressioncited by 243
  12. 2022DyLoRA: Parameter Efficient Tuning of Pre-trained Models using Dynamic Search-Free Low-Rank Adaptationcited by 233
  13. 2024InfLoRA: Interference-Free Low-Rank Adaptation for Continual Learningcited by 198
  14. 2020Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluationcited by 188
  15. 2023HyperDiffusion: Generating Implicit Neural Fields with Weight-Space Diffusioncited by 152
  16. 2024LaCo: Large Language Model Pruning via Layer Collapsecited by 126
  17. 2023Parameter-Efficient Orthogonal Finetuning via Butterfly Factorizationcited by 114
  18. 2022MiniViT: Compressing Vision Transformers with Weight Multiplexingcited by 88
  19. 2024DeCo: Decoupling Token Compression from Semantic Abstraction in Multimodal Large Language Modelscited by 75
  20. 2021Learning Efficient Vision Transformers via Fine-Grained Manifold Distillationcited by 60
  21. 2024SVFT: Parameter-Efficient Fine-Tuning with Singular Vectorscited by 58
  22. 2023UPop: Unified and Progressive Pruning for Compressing Vision-Language Transformerscited by 54
  23. 2023LoRAPrune: Structured Pruning Meets Low-Rank Parameter-Efficient Fine-Tuningcited by 53
  24. 2024VB-LoRA: Extreme Parameter Efficient Fine-Tuning with Vector Bankscited by 52
  25. 2024RoSA: Accurate Parameter-Efficient Fine-Tuning via Robust Adaptationcited by 48
  26. 2023Adversarial Prompt Tuning for Vision-Language Modelscited by 44
  27. 2024SlimGPT: Layer-wise Structured Pruning for Large Language Modelscited by 44
  28. 2021A Unified Pruning Framework for Vision Transformerscited by 40
  29. 2023MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learningcited by 40
  30. 2024Basis Sharing: Cross-Layer Parameter Sharing for Large Language Model Compressioncited by 35
  31. 2023Visual Tuningcited by 34
  32. 2023MeLo: Low-rank Adaptation is Better than Fine-tuning for Medical Image Diagnosiscited by 33
  33. 2024Revisiting the Power of Prompt for Visual Tuningcited by 32
  34. 2024Visual Fourier Prompt Tuningcited by 30
  35. 2019Learning Implicitly Recurrent CNNs Through Parameter Sharingcited by 27
  36. 2023LION: Implicit Vision Prompt Tuningcited by 22
  37. 2024Isomorphic Pruning for Vision Modelscited by 21
  38. 2024Small Scale Data-Free Knowledge Distillationcited by 20

50 most cited of 71 All 71 (JSON)

Built on by

Not cited yet.