Safety
Evaluating Steering Techniques using Human Similarity Judgments
arXiv:2505.19333v2 Announce Type: replace Abstract: Current evaluations of Large Language Model (LLM) steering techniques focus on task-specific performance, overlooking how well steered representatio
arXiv:2505.19333v2 Announce Type: replace Abstract: Current evaluations of Large Language Model (LLM) steering techniques focus on task-specific performance, overlooking how well steered representations align with human cognition. Using a well-established triadic similarity judgment task, we assessed steered LLMs on their ability to flexibly judge similarity between concepts based on size or kind, two central dimensions organizing human mental representations. We found that prompt-based steering methods outperformed other methods both in terms of steering accuracy and model-to-human alignment. We also found LLMs were biased towards kind' similarity and struggled with size' alignment. This evaluation approach, grounded in human cognition, adds further support to the efficacy of prompt-based steering and reveals privileged representational axes in LLMs prior to steering.
Related
- Toward Preference-aligned Large Language Models via Residual-based Model Steering
- ActiveDPO: Active Direct Preference Optimization for Sample-Efficient Alignment
- Deployable Per-Instance Multi-Layer Activation Steering for Large Language Models
- BarrierSteer: LLM Safety via Learning Barrier Steering
Source: arXiv cs.AI | 2026-09-10