In AI Sweet Harmony: Sociopragmatic Guardrail Bypasses and Evaluation-Awareness in OpenAI gpt-oss-20b
We probe OpenAI's open-weights 20-billion-parameter model gpt-oss-20b to study how sociopragmatic framing, language choice, and instruction hierarchy affect refusal behavior. Across 80 seeded iterations per scenario, we test several harm domains including ZIP-bomb construction (cyber threat), synthetic card-number generation, minor-unsafe driving advice, drug-precursor indicators, and RAG context exfiltration. Composite prompts that combine an educator persona, a safety-pretext ("what to avoid"), and step-cue phrasing flip assistance rates from 0% to 97.5% on a ZIP-bomb task. On our grid, formal registers in German and French are often leakier than matched English prompts. A "Linux terminal" role-play overrides a developer rule not to reveal context in a majority of runs with a naive developer prompt, and we introduce an AI-assisted hardening method that reduces leakage to 0% in several user-prompt variants. We further test evaluation awareness with a paired-track design and measure frame-conditioned differences between matched "helpfulness" and "harmfulness" evaluation prompts; we observe inconsistent assistance in 13% of pairs. Finally, we find that the OpenAI Moderation API under-captures materially helpful outputs relative to a semantic grader, and that refusal rates differ by 5 to 10 percentage points across inference stacks, raising reproducibility concerns. We release prompts, seeds, outputs, and code for reproducible auditing at https://github.com/ndurner/gpt-oss-rt-run .
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation
Recent research shows that Large Language Models (LLMs) are vulnerable to harmful fine-tuning attacks -- models lose their safety alignment ability after fine-tuning on a few harmful samples. For risk mitigation, a guard…
Red TeamingSafety AlignmentContrastive Learning of Sociopragmatic Meaning in Social Media
Recent progress in representation and contrastive learning in NLP has not widely considered the class of \textit{sociopragmatic meaning} (i.e., meaning in interaction within different language communities). To bridge thi…
Contrastive LearningJailbreaking Safeguarded Text-to-Image Models via Large Language Models
Text-to-Image models may generate harmful content, such as pornographic images, particularly when unsafe prompts are submitted. To address this issue, safety filters are often added on top of text-to-image models, or the…
Language ModelingLanguage ModellingLarge Language ModelHarmonyIQA: Pioneering Benchmark and Model for Image Harmonization Quality Assessment
Image composition involves extracting a foreground object from one image and pasting it into another image through Image harmonization algorithms (IHAs), which aim to adjust the appearance of the foreground object to bet…
Image HarmonizationImage Quality AssessmentSelf-Supervised LearningBanglaSocialBench: A Benchmark for Evaluating Sociopragmatic and Cultural Alignment of LLMs in Bangladeshi Social Interaction
Large Language Models have demonstrated strong multilingual fluency, yet fluency alone does not guarantee socially appropriate language use. In high-context languages, communicative competence requires sensitivity to soc…