Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2409.06927 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915032018714624 |
|---|---|
| author | Ackerman, Christopher M. |
| author_facet | Ackerman, Christopher M. |
| contents | Activation engineering is becoming increasingly popular as a means of online control of large language models (LLMs). In this work, we extend the idea of inference-time steering with vectors that represent a behavioral direction of interest to tuning those vectors directly into the model, obviating the need for online control. First, we identify activation vectors related to honesty in an open-source LLM (Llama-2-13b-chat). Next, we demonstrate that model output can be made more or less honest by adding positive or negative multiples of these vectors to residual stream activations during generation. Then, we show that a similar effect can be achieved by fine-tuning the vectors directly into the model, by use of a dual loss function based on the cosine similarity of residual stream activations to the vectors combined with a standard token-based loss ("representation tuning"). Finally, we compare the generations in response to honesty-probing prompts from the resulting models to those from models fine-tuned with a token-based loss alone, and to those from the untuned model subjected to online steering. Overall, fine-tuning the vectors into the models using the cosine similarity plus token loss showed a stronger effect than online steering, and generalized better than using the standard loss, suggesting the potential utility of this approach as a safety measure. Code and data are available at https://github.com/cma1114/representation_tuning. Tuned models are available at https://huggingface.co/collections/cackerman/representation-tuning-66da1e5ab41cd1b824687d9f. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_06927 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Representation Tuning Ackerman, Christopher M. Machine Learning Computation and Language Activation engineering is becoming increasingly popular as a means of online control of large language models (LLMs). In this work, we extend the idea of inference-time steering with vectors that represent a behavioral direction of interest to tuning those vectors directly into the model, obviating the need for online control. First, we identify activation vectors related to honesty in an open-source LLM (Llama-2-13b-chat). Next, we demonstrate that model output can be made more or less honest by adding positive or negative multiples of these vectors to residual stream activations during generation. Then, we show that a similar effect can be achieved by fine-tuning the vectors directly into the model, by use of a dual loss function based on the cosine similarity of residual stream activations to the vectors combined with a standard token-based loss ("representation tuning"). Finally, we compare the generations in response to honesty-probing prompts from the resulting models to those from models fine-tuned with a token-based loss alone, and to those from the untuned model subjected to online steering. Overall, fine-tuning the vectors into the models using the cosine similarity plus token loss showed a stronger effect than online steering, and generalized better than using the standard loss, suggesting the potential utility of this approach as a safety measure. Code and data are available at https://github.com/cma1114/representation_tuning. Tuned models are available at https://huggingface.co/collections/cackerman/representation-tuning-66da1e5ab41cd1b824687d9f. |
| title | Representation Tuning |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2409.06927 |