Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2510.01288 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911188585021440 |
|---|---|
| author | Melo, Rui Abreu, Rui Pasareanu, Corina S. |
| author_facet | Melo, Rui Abreu, Rui Pasareanu, Corina S. |
| contents | We draw inspiration from microsaccades, tiny involuntary eye movements that reveal hidden dynamics of human perception, to propose an analogous probing method for large language models (LLMs). Just as microsaccades expose subtle but informative shifts in vision, we show that lightweight position encoding perturbations elicit latent signals that indicate model misbehaviour. Our method requires no fine-tuning or task-specific supervision, yet detects failures across diverse settings including factuality, safety, toxicity, and backdoor attacks. Experiments on multiple state-of-the-art LLMs demonstrate that these perturbation-based probes surface misbehaviours while remaining computationally efficient. These findings suggest that pretrained LLMs already encode the internal evidence needed to flag their own failures, and that microsaccade-inspired interventions provide a pathway for detecting and mitigating undesirable behaviours. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_01288 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Microsaccade-Inspired Probing: Positional Encoding Perturbations Reveal LLM Misbehaviours Melo, Rui Abreu, Rui Pasareanu, Corina S. Machine Learning Artificial Intelligence We draw inspiration from microsaccades, tiny involuntary eye movements that reveal hidden dynamics of human perception, to propose an analogous probing method for large language models (LLMs). Just as microsaccades expose subtle but informative shifts in vision, we show that lightweight position encoding perturbations elicit latent signals that indicate model misbehaviour. Our method requires no fine-tuning or task-specific supervision, yet detects failures across diverse settings including factuality, safety, toxicity, and backdoor attacks. Experiments on multiple state-of-the-art LLMs demonstrate that these perturbation-based probes surface misbehaviours while remaining computationally efficient. These findings suggest that pretrained LLMs already encode the internal evidence needed to flag their own failures, and that microsaccade-inspired interventions provide a pathway for detecting and mitigating undesirable behaviours. |
| title | Microsaccade-Inspired Probing: Positional Encoding Perturbations Reveal LLM Misbehaviours |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2510.01288 |