Beyond Safe Answers: A Benchmark for Evaluating True Risk Awareness in Large Reasoning Models

• URL: http://arxiv.org/abs/2505.19690v1
• Date: Mon, 26 May 2025 08:49:19 GMT
• Title: Beyond Safe Answers: A Benchmark for Evaluating True Risk Awareness in Large Reasoning Models
• Authors: Baihui Zheng, Boren Zheng, Kerui Cao, Yingshui Tan, Zhendong Liu, Weixun Wang, Jiaheng Liu, Jian Yang, Wenbo Su, Xiaoyong Zhu, Bo Zheng, Kaifu Zhang,
• Abstract summary: We introduce textbfBeyond Safe Answers (BSA) bench, a novel benchmark comprising 2,000 challenging instances organized into three distinct SSA scenario types.<n> Evaluations of 19 state-of-the-art LRMs demonstrate the difficulty of this benchmark, with top-performing models achieving only 38.0% accuracy in correctly identifying risk rationales.<n>Our work provides a comprehensive assessment tool for evaluating and improving safety reasoning fidelity in LRMs, advancing the development of genuinely risk-aware and reliably safe AI systems.
• Score: 29.569220030102986
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Despite the remarkable proficiency of \textit{Large Reasoning Models} (LRMs) in handling complex reasoning tasks, their reliability in safety-critical scenarios remains uncertain. Existing evaluations primarily assess response-level safety, neglecting a critical issue we identify as \textbf{\textit{Superficial Safety Alignment} (SSA)} -- a phenomenon where models produce superficially safe outputs while internal reasoning processes fail to genuinely detect and mitigate underlying risks, resulting in inconsistent safety behaviors across multiple sampling attempts. To systematically investigate SSA, we introduce \textbf{Beyond Safe Answers (BSA)} bench, a novel benchmark comprising 2,000 challenging instances organized into three distinct SSA scenario types and spanning nine risk categories, each meticulously annotated with risk rationales. Evaluations of 19 state-of-the-art LRMs demonstrate the difficulty of this benchmark, with top-performing models achieving only 38.0\% accuracy in correctly identifying risk rationales. We further explore the efficacy of safety rules, specialized fine-tuning on safety reasoning data, and diverse decoding strategies in mitigating SSA. Our work provides a comprehensive assessment tool for evaluating and improving safety reasoning fidelity in LRMs, advancing the development of genuinely risk-aware and reliably safe AI systems.

Related papers

• SafeRBench: A Comprehensive Benchmark for Safety Assessment in Large Reasoning Models [60.8821834954637]
  We present SafeRBench, the first benchmark that assesses LRM safety end-to-end.<n>We pioneer the incorporation of risk categories and levels into input design.<n>We introduce a micro-thought chunking mechanism to segment long reasoning traces into semantically coherent units.
  arXiv  Detail & Related papers  (2025-11-19T06:46:33Z)
• DeepKnown-Guard: A Proprietary Model-Based Safety Response Framework for AI Agents [12.054307827384415]
  Large Language Models (LLMs) have become increasingly prominent, severely constraining their trustworthy deployment in critical domains.<n>This paper proposes a novel safety response framework designed to safeguard LLMs at both the input and output levels.
  arXiv  Detail & Related papers  (2025-11-05T03:04:35Z)
• Towards Safe Reasoning in Large Reasoning Models via Corrective Intervention [53.25106308403173]
  We show that existing methods overlook the unique significance of safe reasoning, undermining their trustworthiness and posing potential risks in applications if unsafe reasoning is accessible for and exploited by malicious users.<n>We propose Intervened Preference Optimization (IPO), an alignment method that enforces safe reasoning by substituting compliance steps with safety triggers and constructing pairs for preference learning with strong signals.
  arXiv  Detail & Related papers  (2025-09-29T07:41:09Z)
• RADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized Collaboration [81.38705556267917]
  Existing safety evaluation methods for large language models (LLMs) suffer from inherent limitations.<n>We introduce a theoretical framework that reconstructs the underlying risk concept space.<n>We propose RADAR, a multi-agent collaborative evaluation framework.
  arXiv  Detail & Related papers  (2025-09-28T09:35:32Z)
• AURA: Affordance-Understanding and Risk-aware Alignment Technique for Large Language Models [6.059681491089391]
  AURA provides comprehensive, step level evaluations across logical coherence and safety-awareness.<n>Our framework seamlessly combines introspective self-critique, fine-grained PRM assessments, and adaptive safety-aware decoding.<n>This research represents a pivotal step toward safer, more responsible, and contextually aware AI, setting a new benchmark for alignment-sensitive applications.
  arXiv  Detail & Related papers  (2025-08-08T08:43:24Z)
• Exploring the Secondary Risks of Large Language Models [17.845215420030467]
  We introduce secondary risks marked by harmful or misleading behaviors during benign prompts.<n>Unlike adversarial attacks, these risks stem from imperfect generalization and often evade standard safety mechanisms.<n>We propose SecLens, a black-box, multi-objective search framework that efficiently elicits secondary risk behaviors.
  arXiv  Detail & Related papers  (2025-06-14T07:31:52Z)
• ReasoningShield: Content Safety Detection over Reasoning Traces of Large Reasoning Models [19.963759799471568]
  Reasoning Models (LRMs) are transforming the AI landscape with advanced reasoning capabilities.<n>While the generated reasoning traces enhance model transparency, they can still contain unsafe content, even when the final answer appears safe.<n>Existing moderation tools, primarily designed for question-answer (QA) pairs, are empirically ineffective at detecting hidden risks embedded in reasoning traces.<n>We propose ReasoningShield, the first safety detection model tailored to identify potential risks in the reasoning trace before reaching the final answer.
  arXiv  Detail & Related papers  (2025-05-22T19:44:41Z)
• Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model [30.774446187857475]
  We conduct a safety evaluation of 11 Multimodal Large Reasoning Models (MLRMs) across 5 benchmarks.<n>Our analysis reveals distinct safety patterns across different benchmarks.<n>It is a potential approach to address safety issues in MLRMs by leveraging the intrinsic reasoning capabilities of the model to detect unsafe intent.
  arXiv  Detail & Related papers  (2025-05-10T06:59:36Z)
• SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models [50.34706204154244]
  Acquiring reasoning capabilities catastrophically degrades inherited safety alignment.<n>Certain scenarios suffer 25 times higher attack rates.<n>Despite tight reasoning-answer safety coupling, MLRMs demonstrate nascent self-correction.
  arXiv  Detail & Related papers  (2025-04-09T06:53:23Z)
• The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1 [70.94607997570729]
  We present a comprehensive safety assessment of OpenAI-o3 and DeepSeek-R1 reasoning models.<n>We investigate their susceptibility to adversarial attacks, such as jailbreaking and prompt injection, to assess their robustness in real-world applications.
  arXiv  Detail & Related papers  (2025-02-18T09:06:07Z)
• EARBench: Towards Evaluating Physical Risk Awareness for Task Planning of Foundation Model-based Embodied AI Agents [53.717918131568936]
  Embodied artificial intelligence (EAI) integrates advanced AI models into physical entities for real-world interaction.<n>Foundation models as the "brain" of EAI agents for high-level task planning have shown promising results.<n>However, the deployment of these agents in physical environments presents significant safety challenges.<n>This study introduces EARBench, a novel framework for automated physical risk assessment in EAI scenarios.
  arXiv  Detail & Related papers  (2024-08-08T13:19:37Z)
• Safe Inputs but Unsafe Output: Benchmarking Cross-modality Safety Alignment of Large Vision-Language Model [73.8765529028288]
  We introduce a novel safety alignment challenge called Safe Inputs but Unsafe Output (SIUO) to evaluate cross-modality safety alignment.<n>To empirically investigate this problem, we developed the SIUO, a cross-modality benchmark encompassing 9 critical safety domains, such as self-harm, illegal activities, and privacy violations.<n>Our findings reveal substantial safety vulnerabilities in both closed- and open-source LVLMs, underscoring the inadequacy of current models to reliably interpret and respond to complex, real-world scenarios.
  arXiv  Detail & Related papers  (2024-06-21T16:14:15Z)
• ASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models [65.79770974145983]
  ASSERT, Automated Safety Scenario Red Teaming, consists of three methods -- semantically aligned augmentation, target bootstrapping, and adversarial knowledge injection. We partition our prompts into four safety domains for a fine-grained analysis of how the domain affects model performance. We find statistically significant performance differences of up to 11% in absolute classification accuracy among semantically related scenarios and error rates of up to 19% absolute error in zero-shot adversarial settings.
  arXiv  Detail & Related papers  (2023-10-14T17:10:28Z)
• Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements [76.80453043969209]
  This survey presents a framework for safety research pertaining to large models. We begin by introducing safety issues of wide concern, then delve into safety evaluation methods for large models. We explore the strategies for enhancing large model safety from training to deployment.
  arXiv  Detail & Related papers  (2023-02-18T09:32:55Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.