A Primer on the Inner Workings of Transformer-based Language Models

2024/04/30 by Javier Ferrando, Gabriele Sarti, Ferrando, Javier +5 · 4 voices · 11 citations
#cs.CL

paper · pdf · doi:10.48550/arxiv.2405.00208

Abstract

The rapid progress of research aimed at interpreting the inner workings of advanced language models has highlighted a need for contextualizing the insights gained from years of work in this area. This primer provides a concise technical introduction to the current techniques used to interpret the inner workings of Transformer-based language models, focusing on the generative decoder-only architecture. We conclude by presenting a comprehensive overview of the known internal mechanisms implemented by these models, uncovering connections across popular approaches and active research directions in this area.

Cited by

• Unifying Learning Dynamics and Generalization in Transformers Scaling Law
• Attention is All You Need to Defend Against Indirect Prompt Injection Attacks in LLMs
• Are language models aware of the road not taken? Token-level uncertainty and hidden state dynamics
• DePass: Unified Feature Attributing by Simple Decomposed Forward Pass
• Explainability of Large Language Models: Opportunities and Challenges toward Generating Trustworthy Explanations
• Direct Multi-Token Decoding
• The Geometry of Reasoning: Flowing Logics in Representation Space
• Toward Mechanistic Explanation of Deductive Reasoning in Language Models
• Model-Agnostic Correctness Assessment for LLM-Generated Code via Dynamic Internal Representation Selection
• Interpreting Attention Heads for Image-to-Text Information Flow in Large Vision-Language Models
• Cross-Attention is Half Explanation in Speech-to-Text Models

Discussions

• Perhaps these ones are useful. A Primer on the Inner Workings of Transformer-based Language Models, by Javier Ferrando et al arxiv.org/abs/2405.00208 Transformer feed-forward layers are key-value m [bsky, 4 points, 1 comments]
• A Primer on the Inner Workings of Transformer-Based Language Models [hn, 4 points, 0 comments]
• A Primer on the Inner Workings of Transformer-Based Language Models [hn, 4 points, 0 comments]
• A Primer on the Inner Workings of Transformer-Based Language Models https://arxiv.org/abs/2405.00208 [bsky, 0 points, 0 comments]

Related

• Generative Pre-trained Transformer: A Comprehensive Review on Enabling Technologies, Potential Applications, Emerging Challenges, and Future Directions
• Why transformers are obviously good models of language
• A Survey on Large Language Models from Concept to Implementation
• Introduction to Transformers: an NLP Perspective
• What Context Features Can Transformer Language Models Use?