SafetyFlow: An Agent-Flow System for Automated LLM Safety Benchmarking

arXiv:2508.15526 · cs.CL · Submitted 2025-08-21 · Read on arXiv

cs.CL

Submitted: 2025-08-21

Updated: 2026-09-22

Comments: Code and dataset are available at https://github.com/yangyangyang127/SafetyFlow

Code: https://github.com/yangyangyang127/SafetyFlow

Project page: https://www.reddit.comhttps://huggingface.co/datasets/tomekkorbak/pile-curse-full

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 69× in this paper · explained in 2 episodes
• Redundancy — 20× in this paper · explained in 2 episodes
• Benchmarking — 10× in this paper · explained in 3 episodes
• Ablation — 8× in this paper · explained in 1 episode
• Alignment — 5× in this paper · explained in 1 episode
• Adaptability — 3× in this paper · explained in 1 episode
• Clustering — 3× in this paper · explained in 2 episodes
• Embeddings — 3× in this paper · explained in 2 episodes
• Agent-Driven Pipeline — 2× in this paper · explained in 1 episode
• Core — 2× in this paper · explained in 2 episodes
• Generalization — 2× in this paper · explained in 13 episodes
• Harmful Rate — 2× in this paper
• Large Language Models (LLMs) — 2× in this paper · explained in 6 episodes
• Ablation study — 1× in this paper · explained in 1 episode
• Agentic System — 1× in this paper · explained in 2 episodes
• Consistency — 1× in this paper · explained in 3 episodes
• Distillation — 1× in this paper · explained in 2 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Foundation Model — 1× in this paper · explained in 3 episodes
• Foundation Models — 1× in this paper · explained in 7 episodes
• Low-Resource Languages — 1× in this paper · explained in 4 episodes
• Multi-agent Collaboration — 1× in this paper · explained in 2 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Safety Rate — 1× in this paper
• Sycophancy — 1× in this paper · explained in 4 episodes
• Transformer — 1× in this paper · explained in 2 episodes

Sources

• Phi-4 Technical Report
• Does Refusal Training in LLMs Generalize to the Past Tense?
• InternLM2 Technical Report
• Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
• Multilingual Jailbreak Challenges in Large Language Models
• Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned
• ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools
• Flames: Benchmarking Value Alignment of LLMs in Chinese
• SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models
• ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversation
• DeepSeek-V3 Technical Report
• Muon is Scalable for LLM Training
• The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs
• GuardReasoner: Towards Reasoning-based LLM Safeguards
• Large Language Model Agent: A Survey on Methodology, Applications and Challenges
• BBQ: A Hand-Built Bias Benchmark for Question Answering
• Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
• CodeAttack: Revealing Safety Generalization Challenges of Large Language Models via Code Completion
• Safety Assessment of Chinese Large Language Models
• Gemma 3 Technical Report

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering