DeepSight: An All-in-One LM Safety Toolkit

TL;DR

DeepSight introduces an integrated, open-source safety toolkit for large language models that unifies evaluation and diagnosis, providing transparent insights into internal mechanisms and supporting frontier AI risk assessment.

Contribution

It presents the first unified safety evaluation and diagnosis toolkit for LLMs, enabling transparent, scalable, and cost-effective safety analysis.

Findings

Unifies safety evaluation and diagnosis in one toolkit

Supports frontier AI risk evaluation

Transforms safety analysis from black-box to white-box insights

Abstract

As the development of Large Models (LMs) progresses rapidly, their safety is also a priority. In current Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) safety workflow, evaluation, diagnosis, and alignment are often handled by separate tools. Specifically, safety evaluation can only locate external behavioral risks but cannot figure out internal root causes. Meanwhile, safety diagnosis often drifts from concrete risk scenarios and remains at the explainable level. In this way, safety alignment lack dedicated explanations of changes in internal mechanisms, potentially degrading general capabilities. To systematically address these issues, we propose an open-source project, namely DeepSight, to practice a new safety evaluation-diagnosis integrated paradigm. DeepSight is low-cost, reproducible, efficient, and highly scalable large-scale model safety evaluation project consisting of a evaluation toolkit DeepSafe and a diagnosis toolkit DeepScan. By unifying task and data protocols, we build a connection between the two stages and transform safety evaluation from black-box to white-box insight. Besides, DeepSight is the first open source toolkit that support the frontier AI risk evaluation and joint safety evaluation and diagnosis.