Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2410.02472 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913573487247360 |
|---|---|
| author | Costarelli, Anthony Allen, Mat Field, Severin |
| author_facet | Costarelli, Anthony Allen, Mat Field, Severin |
| contents | As Large Language Models (LLMs) become increasingly integrated into our daily lives, the potential harms from deceptive behavior underlie the need for faithfully interpreting their decision-making. While traditional probing methods have shown some effectiveness, they remain best for narrowly scoped tasks while more comprehensive explanations are still necessary. To this end, we investigate meta-models-an architecture using a "meta-model" that takes activations from an "input-model" and answers natural language questions about the input-model's behaviors. We evaluate the meta-model's ability to generalize by training them on selected task types and assessing their out-of-distribution performance in deceptive scenarios. Our findings show that meta-models generalize well to out-of-distribution tasks and point towards opportunities for future research in this area. Our code is available at https://github.com/acostarelli/meta-models-public . |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_02472 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Meta-Models: An Architecture for Decoding LLM Behaviors Through Interpreted Embeddings and Natural Language Costarelli, Anthony Allen, Mat Field, Severin Machine Learning Artificial Intelligence As Large Language Models (LLMs) become increasingly integrated into our daily lives, the potential harms from deceptive behavior underlie the need for faithfully interpreting their decision-making. While traditional probing methods have shown some effectiveness, they remain best for narrowly scoped tasks while more comprehensive explanations are still necessary. To this end, we investigate meta-models-an architecture using a "meta-model" that takes activations from an "input-model" and answers natural language questions about the input-model's behaviors. We evaluate the meta-model's ability to generalize by training them on selected task types and assessing their out-of-distribution performance in deceptive scenarios. Our findings show that meta-models generalize well to out-of-distribution tasks and point towards opportunities for future research in this area. Our code is available at https://github.com/acostarelli/meta-models-public . |
| title | Meta-Models: An Architecture for Decoding LLM Behaviors Through Interpreted Embeddings and Natural Language |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2410.02472 |