Emergent Modularity in Pre-trained Transformers
This work examines the presence of modularity in pre-trained Transformers, a feature commonly found in human brains and thought to be vital for general intelligence. In analogy to human brains, we consider two main characteristics of modularity: (1) functional specialization of neurons: we evaluate whether each neuron is mainly specialized in a certain function, and find that the answer is yes. (2) function-based neuron grouping: we explore finding a structure that groups neurons into modules by function, and each module works for its corresponding function. Given the enormous amount of possible structures, we focus on Mixture-of-Experts as a promising candidate, which partitions neurons into experts and usually activates different experts for different inputs. Experimental results show that there are functional experts, where clustered are the neurons specialized in a certain function. Moreover, perturbing the activations of functional experts significantly affects the corresponding function. Finally, we study how modularity emerges during pre-training, and find that the modular structure is stabilized at the early stage, which is faster than neuron stabilization. It suggests that Transformers first construct the modular structure and then learn fine-grained neuron functions. Our code and data are available at https://github.com/THUNLP/modularity-analysis.
Code (1)
Tasks
Mixture-of-ExpertsMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
Unlocking Emergent Modularity in Large Language Models
Modular Neural Networks (MNNs) demonstrate various advantages over monolithic models. Existing MNNs are generally $\textit{explicit}$: their modular architectures are pre-defined, with individual modules expected to impl…
Domain GeneralizationTransfer LearningNo Clustering, No Routing: How Transformers Actually Process Rare Tokens
Large language models struggle with rare token prediction, yet the mechanisms driving their specialization remain unclear. Prior work identified specialized ``plateau'' neurons for rare tokens following distinctive three…
Engineering Verifiable Modularity in Transformers via Per-Layer Supervision
Transformers resist surgical control. Ablating an attention head identified as critical for capitalization produces minimal behavioral change because distributed redundancy compensates for damage. This Hydra effect rende…
A Survey of Generative Categories and Techniques in Multimodal Large Language Models
Multimodal Large Language Models (MLLMs) have rapidly evolved beyond text generation, now spanning diverse output modalities including images, music, video, human motion, and 3D objects, by integrating language with othe…
Mixture-of-ExpertsSelf-Supervised LearningSurveyText GenerationAn Efficient General-Purpose Modular Vision Model via Multi-Task Heterogeneous Training
We present a model that can perform multiple vision tasks and can be adapted to other downstream tasks efficiently. Despite considerable progress in multi-task learning, most efforts focus on learning from multi-label da…
Continual LearningMixture-of-ExpertsMulti-Task Learning