Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2410.17245 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916448965754880 |
|---|---|
| author | Pres, Itamar Ruis, Laura Lubana, Ekdeep Singh Krueger, David |
| author_facet | Pres, Itamar Ruis, Laura Lubana, Ekdeep Singh Krueger, David |
| contents | Representation engineering methods have recently shown promise for enabling efficient steering of model behavior. However, evaluation pipelines for these methods have primarily relied on subjective demonstrations, instead of quantitative, objective metrics. We aim to take a step towards addressing this issue by advocating for four properties missing from current evaluations: (i) contexts sufficiently similar to downstream tasks should be used for assessing intervention quality; (ii) model likelihoods should be accounted for; (iii) evaluations should allow for standardized comparisons across different target behaviors; and (iv) baseline comparisons should be offered. We introduce an evaluation pipeline grounded in these criteria, offering both a quantitative and visual analysis of how effectively a given method works. We use this pipeline to evaluate two representation engineering methods on how effectively they can steer behaviors such as truthfulness and corrigibility, finding that some interventions are less effective than previously reported. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_17245 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Towards Reliable Evaluation of Behavior Steering Interventions in LLMs Pres, Itamar Ruis, Laura Lubana, Ekdeep Singh Krueger, David Artificial Intelligence Computation and Language Representation engineering methods have recently shown promise for enabling efficient steering of model behavior. However, evaluation pipelines for these methods have primarily relied on subjective demonstrations, instead of quantitative, objective metrics. We aim to take a step towards addressing this issue by advocating for four properties missing from current evaluations: (i) contexts sufficiently similar to downstream tasks should be used for assessing intervention quality; (ii) model likelihoods should be accounted for; (iii) evaluations should allow for standardized comparisons across different target behaviors; and (iv) baseline comparisons should be offered. We introduce an evaluation pipeline grounded in these criteria, offering both a quantitative and visual analysis of how effectively a given method works. We use this pipeline to evaluate two representation engineering methods on how effectively they can steer behaviors such as truthfulness and corrigibility, finding that some interventions are less effective than previously reported. |
| title | Towards Reliable Evaluation of Behavior Steering Interventions in LLMs |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2410.17245 |