CBMAS: Cognitive Behavioral Modeling via Activation Steering
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912813862092800 |
|---|---|
| author | Ismail, Ahmed H. Kuang, Anthony Akinkugbe, Ayo Zhu, Kevin O'Brien, Sean |
| author_facet | Ismail, Ahmed H. Kuang, Anthony Akinkugbe, Ayo Zhu, Kevin O'Brien, Sean |
| contents | Large language models (LLMs) often encode cognitive behaviors unpredictably across prompts, layers, and contexts, making them difficult to diagnose and control. We present CBMAS, a diagnostic framework for continuous activation steering, which extends cognitive bias analysis from discrete before/after interventions to interpretable trajectories. By combining steering vector construction with dense α-sweeps, logit lens-based bias curves, and layer-site sensitivity analysis, our approach can reveal tipping points where small intervention strengths flip model behavior and show how steering effects evolve across layer depth. We argue that these continuous diagnostics offer a bridge between high-level behavioral evaluation and low-level representational dynamics, contributing to the cognitive interpretability of LLMs. Lastly, we provide a CLI and datasets for various cognitive behaviors at the project repository, https://github.com/shimamooo/CBMAS. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_06109 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | CBMAS: Cognitive Behavioral Modeling via Activation Steering Ismail, Ahmed H. Kuang, Anthony Akinkugbe, Ayo Zhu, Kevin O'Brien, Sean Artificial Intelligence Machine Learning Large language models (LLMs) often encode cognitive behaviors unpredictably across prompts, layers, and contexts, making them difficult to diagnose and control. We present CBMAS, a diagnostic framework for continuous activation steering, which extends cognitive bias analysis from discrete before/after interventions to interpretable trajectories. By combining steering vector construction with dense α-sweeps, logit lens-based bias curves, and layer-site sensitivity analysis, our approach can reveal tipping points where small intervention strengths flip model behavior and show how steering effects evolve across layer depth. We argue that these continuous diagnostics offer a bridge between high-level behavioral evaluation and low-level representational dynamics, contributing to the cognitive interpretability of LLMs. Lastly, we provide a CLI and datasets for various cognitive behaviors at the project repository, https://github.com/shimamooo/CBMAS. |
| title | CBMAS: Cognitive Behavioral Modeling via Activation Steering |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2601.06109 |