Search papers, labs, and topics across Lattice.
This paper introduces RealSafe-R1, a safety-aligned version of the DeepSeek-R1 large reasoning model, addressing concerns about its susceptibility to malicious queries. The authors constructed a 15k-trajectory dataset of safety-aware reasoning examples generated by DeepSeek-R1 under explicit refusal instructions. Experiments demonstrate that RealSafe-R1 exhibits improved safety guardrails against harmful queries and jailbreak attacks while preserving the original reasoning capabilities of DeepSeek-R1.
RealSafe-R1 achieves safety alignment of DeepSeek-R1 without sacrificing reasoning performance, a common trade-off in prior safety alignment efforts.
Large Reasoning Models (LRMs), such as OpenAI o1 and DeepSeek-R1, have been rapidly progressing and achieving breakthrough performance on complex reasoning tasks such as mathematics and coding. However, the open-source R1 models have raised safety concerns in wide applications, such as the tendency to comply with malicious queries, which greatly impacts the utility of these powerful models in their applications. In this paper, we introduce RealSafe-R1 as safety-aligned versions of DeepSeek-R1 distilled models. To train these models, we construct a dataset of 15k safety-aware reasoning trajectories generated by DeepSeek-R1, under explicit instructions for expected refusal behavior. Both quantitative experiments and qualitative case studies demonstrate the models' improvements, which are shown in their safety guardrails against both harmful queries and jailbreak attacks. Importantly, unlike prior safety alignment efforts that often compromise reasoning performance, our method preserves the models' reasoning capabilities by maintaining the training data within the original distribution of generation. Model weights of RealSafe-R1 are open-source at https://huggingface.co/RealSafe.
