paper-with-me

홈 › Papers

Decoupling Knowledge and Reasoning in Transformers: A Modular Architecture with Generalized Cross-Attention

2025-01-01 · Zhenyu Guo, WenGuang Chen

Transformers have achieved remarkable success across diverse domains, but their monolithic architecture presents challenges in interpretability, adaptability, and scalability. This paper introduces a novel modular Transformer architecture that explicitly decouples knowledge and reasoning through a generalized cross-attention mechanism to a globally shared knowledge base with layer-specific transformations, specifically designed for effective knowledge retrieval. Critically, we provide a rigorous mathematical derivation demonstrating that the Feed-Forward Network (FFN) in a standard Transformer is a specialized case (a closure) of this generalized cross-attention, revealing its role in implicit knowledge retrieval and validating our design. This theoretical framework provides a new lens for understanding FFNs and lays the foundation for future research exploring enhanced interpretability, adaptability, and scalability, enabling richer interplay with external knowledge bases and other systems.

📄 PDF Abstract BibTeX arXiv:2501.00823

Code (0)

등록된 구현이 없습니다.

Tasks

Retrieval

Methods 이 논문이 사용한 방법론

Attention 설명 없음
BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…
BASE 설명 없음
Absolute Position Encodings Absolute Position Encodings are a type of position embeddings for [Transformer-based models] where positional encodings are…
Dropout Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is…

Similar Papers 제목 키워드 기반

Disentangling Reasoning Capabilities from Language Models with Compositional Reasoning Transformers

2022-10-20 · Wanjun Zhong, Tingting Ma, Jiahai Wang, Jian Yin 외

This paper presents ReasonFormer, a unified reasoning framework for mirroring the modular and compositional reasoning process of humans in complex decision-making. Inspired by dual-process theory in cognitive science, th…

Decision Making

CausalR: Causal Reasoning over Natural Language Rulebases

2021-11-16 · ACL ARR November 2021 11 · Anonymous

Transformers have been shown to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language. Recent works show that such models can also produce the reasoning steps (i.e.…

Fact SelectionLogical Reasoning

FaiRR: Faithful and Robust Deductive Reasoning over Natural Language

2022-01-16 · ACL ARR January 2022 1 · Anonymous

Transformers have been shown to be able to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language. Recent works show that such models can also produce the reasoning …

Fact SelectionLogical Reasoning

FaiRR: Faithful and Robust Deductive Reasoning over Natural Language

2022-03-19 · ACL 2022 5 · Soumya Sanyal, Harman Singh, Xiang Ren

Transformers have been shown to be able to perform deductive reasoning on a logical rulebase containing rules and statements written in natural language. Recent works show that such models can also produce the reasoning …

Fact SelectionLogical Reasoning

Hydra: A Modular Architecture for Efficient Long-Context Reasoning

2025-08-20 · Siddharth Chaudhary, Dev Patel, Maheep Chaudhary, Bennett Browning arxiv

The quadratic complexity of transformers fundamentally limits reasoning system deployment in resource-constrained and long-context settings. We introduce Hydra, a modular architecture based upon a state-space backbone wh…