Do All Autoregressive Transformers Remember Facts the Same Way? A Cross-Architecture Analysis of Recall Mechanisms
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916945097392128 |
|---|---|
| author | Choe, Minyeong Cho, Haehyun Seo, Changho Kim, Hyunil |
| author_facet | Choe, Minyeong Cho, Haehyun Seo, Changho Kim, Hyunil |
| contents | Understanding how Transformer-based language models store and retrieve factual associations is critical for improving interpretability and enabling targeted model editing. Prior work, primarily on GPT-style models, has identified MLP modules in early layers as key contributors to factual recall. However, it remains unclear whether these findings generalize across different autoregressive architectures. To address this, we conduct a comprehensive evaluation of factual recall across several models -- including GPT, LLaMA, Qwen, and DeepSeek -- analyzing where and how factual information is encoded and accessed. Consequently, we find that Qwen-based models behave differently from previous patterns: attention modules in the earliest layers contribute more to factual recall than MLP modules. Our findings suggest that even within the autoregressive Transformer family, architectural variations can lead to fundamentally different mechanisms of factual recall. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_08778 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Do All Autoregressive Transformers Remember Facts the Same Way? A Cross-Architecture Analysis of Recall Mechanisms Choe, Minyeong Cho, Haehyun Seo, Changho Kim, Hyunil Computation and Language Understanding how Transformer-based language models store and retrieve factual associations is critical for improving interpretability and enabling targeted model editing. Prior work, primarily on GPT-style models, has identified MLP modules in early layers as key contributors to factual recall. However, it remains unclear whether these findings generalize across different autoregressive architectures. To address this, we conduct a comprehensive evaluation of factual recall across several models -- including GPT, LLaMA, Qwen, and DeepSeek -- analyzing where and how factual information is encoded and accessed. Consequently, we find that Qwen-based models behave differently from previous patterns: attention modules in the earliest layers contribute more to factual recall than MLP modules. Our findings suggest that even within the autoregressive Transformer family, architectural variations can lead to fundamentally different mechanisms of factual recall. |
| title | Do All Autoregressive Transformers Remember Facts the Same Way? A Cross-Architecture Analysis of Recall Mechanisms |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2509.08778 |