Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

A Unified Framework for Knowledge Transfer in Bidirectional Model Scaling

About

Transferring pre-trained knowledge from a source model to a target model of a different architectural size is a key challenge for flexible and efficient model scaling. However, current parameter-space methods treat Small-to-Large (S2L) and Large-to-Small (L2S) scaling as separate, incompatible problems, focusing on parameter synthesis and selection, respectively. This fragmented perspective has resulted in specialized tools, hindering a unified, bidirectional framework. In this paper, we propose BoT (Bidirectional knowledge Transfer), the first size-agnostic framework to unify S2L and L2S scaling. Our core insight is to treat model weights as continuous signals, where models of different sizes represent distinct discretizations of the transferable knowledge. This multi-resolution perspective directly casts S2L and L2S scaling as the signal processing operations of upsampling and downsampling, naturally leading to the adoption of the Discrete Wavelet Transform (DWT) and its Inverse (IDWT). BoT leverages the recursive nature of wavelets, using the decomposition level as a dynamic scaling factor to bridge disparate model sizes in a parameter-free and computationally efficient manner. Extensive experiments on DeiT, BERT, and GPT demonstrate significant pre-training FLOPs savings (up to 67.1% for S2L, 52.8% for L2S) and state-of-the-art performance on benchmarks like GLUE and SQuAD.

Jianlu Shen, Fu Feng, Jiaze Xu, Yucheng Xie, Jiaqi Lv, Xin Geng• 2026

Related benchmarks

TaskDatasetResultRank
Image ClassificationStanford Cars
Accuracy88.6
635
Image ClassificationFood-101
Accuracy84.4
542
Image ClassificationCIFAR-100
Accuracy82.4
435
Image ClassificationCUB-200 2011
Accuracy74.1
356
Image ClassificationOxford Flowers 102
Accuracy94.7
234
Image ClassificationiNaturalist
Accuracy67.1
63
Natural Language UnderstandingGLUE
SST-290.6
55
Image ClassificationCIFAR-10
Accuracy97.2
33
Question AnsweringSQuAD v1.1 & v2.0
SQuAD v1.1 F184.7
21
Natural Language UnderstandingGLUE
SST-2 Accuracy88.92
8
Showing 10 of 13 rows

Other info

Follow for update