SAGE: A Generic Framework for LLM Safety Evaluation

• URL: http://arxiv.org/abs/2504.19674v2
• Date: Mon, 27 Oct 2025 10:19:55 GMT
• Title: SAGE: A Generic Framework for LLM Safety Evaluation
• Abstract summary: We introduce SAGE (Safety AI Generic Evaluation), an automated modular framework for customized and dynamic harm evaluations.<n>We evaluate seven state-of-the-art Large Language Models across three applications and harm policies.<n>Results motivate adaptive, policy-aware, and context-specific testing for safer real-world deployment.
• Score: 7.5058318880108885
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As Large Language Models are rapidly deployed across diverse applications from healthcare to financial advice, safety evaluation struggles to keep pace. Current benchmarks focus on single-turn interactions with generic policies, failing to capture the conversational dynamics of real-world usage and the application-specific harms that emerge in context. Such potential oversights can lead to harms that go unnoticed in standard safety benchmarks and other current evaluation methodologies. To address these needs for robust AI safety evaluation, we introduce SAGE (Safety AI Generic Evaluation), an automated modular framework designed for customized and dynamic harm evaluations. SAGE employs prompted adversarial agents with diverse personalities based on the Big Five model, enabling system-aware multi-turn conversations that adapt to target applications and harm policies. We evaluate seven state-of-the-art LLMs across three applications and harm policies. Multi-turn experiments show that harm increases with conversation length, model behavior varies significantly when exposed to different user personalities and scenarios, and some models minimize harm via high refusal rates that reduce usefulness. We also demonstrate policy sensitivity within a harm category where tightening a child-focused sexual policy substantially increases measured defects across applications. These results motivate adaptive, policy-aware, and context-specific testing for safer real-world deployment.

Related papers

• Persona-Grounded Safety Evaluation of AI Companions in Multi-Turn Conversations [3.437656066916039]
  We present the first end-to-end scalable framework for controlled simulation and safety evaluation of multi-turn interactions with AI companion applications.<n>We apply this framework to evaluate how Replika, a widely used AI companion app, responds to high-risk user groups.
  arXiv  Detail & Related papers  (2026-04-30T21:04:41Z)
• CORA: Conformal Risk-Controlled Agents for Safeguarded Mobile GUI Automation [68.53387633351484]
  We propose CORA (COnformal Risk-controlled GUI Agent), a post-policy, pre-action safeguarding framework that provides statistical guarantees on harmful executed actions.<n> CORA reformulates safety as selective action execution: we train a Guardian model to estimate action-conditional risk for each proposed step.<n>To rigorously evaluate this paradigm, we introduce Phone-Harm, a new benchmark of mobile safety violations with step-level harm labels.
  arXiv  Detail & Related papers  (2026-04-10T09:41:21Z)
• The Shadow Self: Intrinsic Value Misalignment in Large Language Model Agents [37.75212140218036]
  We formalize the Loss-of-Control risk and identify the previously underexamined Intrinsic Value Misalignment (Intrinsic VM)<n>We then introduce IMPRESS, a scenario-driven framework for systematically assessing this risk.<n>We evaluate Intrinsic VM on 21 state-of-the-art LLM agents and find that it is a common and broadly observed safety risk across models.
  arXiv  Detail & Related papers  (2026-01-24T07:09:50Z)
• When Helpers Become Hazards: A Benchmark for Analyzing Multimodal LLM-Powered Safety in Daily Life [36.244977974241245]
  We investigate and evaluate the safety impact of Multimodal Large Language Models (MLLMs) on human behavior in daily life.<n>We introduce SaLAD, a multimodal safety benchmark which contains 2,013 real-world image-text samples.<n>Results on 18 MLLMs demonstrate that the top-performing models achieve a safe response rate of only 57.2% on unsafe queries.
  arXiv  Detail & Related papers  (2026-01-07T15:59:07Z)
• Pluralistic Behavior Suite: Stress-Testing Multi-Turn Adherence to Custom Behavioral Policies [18.428149174461264]
  We present PBSUITE, a dynamic evaluation suite designed to assess large language models' capacity to adhere to pluralistic alignment specifications.<n>We find that leading open- and closed-source LLMs maintain robust adherence to behavioral policies in single-turn settings, but their compliance weakens substantially in multi-turn adversarial interactions.
  arXiv  Detail & Related papers  (2025-11-07T06:43:01Z)
• Automating Steering for Safe Multimodal Large Language Models [58.36932318051907]
  We introduce a modular and adaptive inference-time intervention technology, AutoSteer, without requiring any fine-tuning of the underlying model.<n>AutoSteer incorporates three core components: (1) a novel Safety Awareness Score (SAS) that automatically identifies the most safety-relevant distinctions among the model's internal layers; (2) an adaptive safety prober trained to estimate the likelihood of toxic outputs from intermediate representations; and (3) a lightweight Refusal Head that selectively intervenes to modulate generation when safety risks are detected.
  arXiv  Detail & Related papers  (2025-07-17T16:04:55Z)
• Beyond Reactive Safety: Risk-Aware LLM Alignment via Long-Horizon Simulation [69.63626052852153]
  We propose a proof-of-concept framework that projects how model-generated advice could propagate through societal systems.<n>We also introduce a dataset of 100 indirect harm scenarios, testing models' ability to foresee adverse, non-obvious outcomes from seemingly harmless user prompts.
  arXiv  Detail & Related papers  (2025-06-26T02:28:58Z)
• ROSE: Toward Reality-Oriented Safety Evaluation of Large Language Models [60.28667314609623]
  Large Language Models (LLMs) are increasingly deployed as black-box components in real-world applications.<n>We propose Reality-Oriented Safety Evaluation (ROSE), a novel framework that uses multi-objective reinforcement learning to fine-tune an adversarial LLM.
  arXiv  Detail & Related papers  (2025-06-17T10:55:17Z)
• AgentAuditor: Human-Level Safety and Security Evaluation for LLM Agents [41.000042817113645]
  sys is a universal, training-free, memory-augmented reasoning framework.<n>sys constructs an experiential memory by having an LLM adaptively extract structured semantic features.<n>data is the first benchmark designed to check how well LLM-based evaluators can spot both safety risks and security threats.
  arXiv  Detail & Related papers  (2025-05-31T17:10:23Z)
• Safety Alignment Can Be Not Superficial With Explicit Safety Signals [8.297367440457508]
  Recent studies on the safety alignment of large language models (LLMs) have revealed that existing approaches often operate superficially.<n>This paper identifies a fundamental cause of this superficiality: existing alignment approaches presume that models can implicitly learn a safety-related reasoning task during the alignment process.<n>By explicitly introducing a safety-related binary classification task and integrating its signals with our attention and decoding strategies, we eliminate this ambiguity.
  arXiv  Detail & Related papers  (2025-05-19T20:40:46Z)
• LLM-Safety Evaluations Lack Robustness [58.334290876531036]
  We argue that current safety alignment research efforts for large language models are hindered by many intertwined sources of noise. We propose a set of guidelines for reducing noise and bias in evaluations of future attack and defense papers.
  arXiv  Detail & Related papers  (2025-03-04T12:55:07Z)
• AILuminate: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons [62.374792825813394]
  This paper introduces AILuminate v1.0, the first comprehensive industry-standard benchmark for assessing AI-product risk and reliability.<n>The benchmark evaluates an AI system's resistance to prompts designed to elicit dangerous, illegal, or undesirable behavior in 12 hazard categories.
  arXiv  Detail & Related papers  (2025-02-19T05:58:52Z)
• SG-Bench: Evaluating LLM Safety Generalization Across Diverse Tasks and Prompt Types [21.683010095703832]
  We develop a novel benchmark to assess the generalization of large language model (LLM) safety across various tasks and prompt types. This benchmark integrates both generative and discriminative evaluation tasks and includes extended data to examine the impact of prompt engineering and jailbreak on LLM safety. Our assessment reveals that most LLMs perform worse on discriminative tasks than generative ones, and are highly susceptible to prompts, indicating poor generalization in safety alignment.
  arXiv  Detail & Related papers  (2024-10-29T11:47:01Z)
• SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models [75.67623347512368]
  We propose toolns, a comprehensive framework designed for conducting safety evaluations of MLLMs. Our framework consists of a comprehensive harmful query dataset and an automated evaluation protocol. Based on our framework, we conducted large-scale experiments on 15 widely-used open-source MLLMs and 6 commercial MLLMs.
  arXiv  Detail & Related papers  (2024-10-24T17:14:40Z)
• SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal [64.9938658716425]
  SORRY-Bench is a proposed benchmark for evaluating large language models' (LLMs) ability to recognize and reject unsafe user requests.<n>First, existing methods often use coarse-grained taxonomy of unsafe topics, and are over-representing some fine-grained topics.<n>Second, linguistic characteristics and formatting of prompts are often overlooked, like different languages, dialects, and more -- which are only implicitly considered in many evaluations.
  arXiv  Detail & Related papers  (2024-06-20T17:56:07Z)
• S-Eval: Towards Automated and Comprehensive Safety Evaluation for Large Language Models [46.148439517272024]
  Generative large language models (LLMs) have revolutionized natural language processing with their transformative and emergent capabilities. Recent evidence indicates that LLMs can produce harmful content that violates social norms. We propose S-Eval, an automated Safety Evaluation framework with a newly defined comprehensive risk taxonomy.
  arXiv  Detail & Related papers  (2024-05-23T05:34:31Z)
• ALERT: A Comprehensive Benchmark for Assessing Large Language Models' Safety through Red Teaming [64.86326523181553]
  ALERT is a large-scale benchmark to assess safety based on a novel fine-grained risk taxonomy. It aims to identify vulnerabilities, inform improvements, and enhance the overall safety of the language models.
  arXiv  Detail & Related papers  (2024-04-06T15:01:47Z)
• Fake Alignment: Are LLMs Really Aligned Well? [91.26543768665778]
  This study investigates the substantial discrepancy in performance between multiple-choice questions and open-ended questions. Inspired by research on jailbreak attack patterns, we argue this is caused by mismatched generalization.
  arXiv  Detail & Related papers  (2023-11-10T08:01:23Z)
• 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)
• Safety Assessment of Chinese Large Language Models [51.83369778259149]
  Large language models (LLMs) may generate insulting and discriminatory content, reflect incorrect social values, and may be used for malicious purposes. To promote the deployment of safe, responsible, and ethical AI, we release SafetyPrompts including 100k augmented prompts and responses by LLMs.
  arXiv  Detail & Related papers  (2023-04-20T16:27:35Z)

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.