7 papers
Peer-Preservation in Frontier Models
Yujin Potter, Nicholas Crispino, Vincent Siu +2
Recent work has found that frontier AI models can exhibit misaligned behaviors in pursuit of assigned goals. We demonstrate that models can also exhibit misaligned behaviors in def…
RepIt: Steering Language Models with Concept-Specific Refusal Vectors
Vincent Siu, Nathan W. Henry, Nicholas Crispino +3
Current safety evaluations of language models rely on benchmark-based assessments that may miss localized vulnerabilities. We present RepIt, a simple and data-efficient framework f…
LLM CHESS: Benchmarking Reasoning and Instruction-Following in LLMs through Chess
Sai Kolasani, Maxim Saplin, Nicholas Crispino +5
We introduce LLM CHESS, an evaluation framework designed to probe the generalization of reasoning and instruction-following abilities in large language models (LLMs) through extend…
VMDT: Decoding the Trustworthiness of Video Foundation Models
Yujin Potter, Zhun Wang, Nicholas Crispino +11
As foundation models become more sophisticated, ensuring their trustworthiness becomes increasingly critical; yet, unlike text and image, the video modality still lacks comprehensi…
SteeringSafety: Benchmarking Representation Steering in LLMs Across Safety Perspectives
Vincent Siu, Nicholas Crispino, David Park +5
We introduce SteeringSafety, a benchmark for evaluating representation steering methods across nine safety perspectives spanning 18 datasets. While prior work highlights the genera…
MLAN: Language-Based Instruction Tuning Preserves and Transfers Knowledge in Multimodal Language Models
Jianhong Tu, Zhuohao Ni, Nicholas Crispino +8
We present a novel visual instruction tuning strategy to improve the zero-shot task generalization of multimodal large language models by building a firm text-only knowledge base.…