ResponsibleRobotBench: Benchmarking Responsible Robot Manipulation using Multi-modal Large Language Models

2025-12-03

AI Summary

This work addresses the lack of accountability and reliability evaluation for multimodal large language models (MLLMs) in high-risk robotic manipulation. We introduce the first benchmark dedicated to responsible robotic manipulation, comprising 23 multi-stage tasks spanning electrical, chemical, and personal safety-critical scenarios, with validated sim-to-real transferability. Methodologically, we integrate MLLMs with visual perception, in-context learning, hazard detection, and multi-representation action execution within a reproducible framework, supported by a novel multimodal dataset. Our approach uniquely unifies risk-aware reasoning, moral inference, and physics-grounded planning, and introduces new evaluation metrics—including “safety success rate”—to quantify responsible behavior. Experimental results demonstrate that the benchmark effectively discriminates among agents across safety compliance, robustness to environmental perturbations, and generalization to unseen tasks and hazards.

Technology Category

• Intelligent Robots: Manipulation
• Natural Language Processing: Safety and Robustness
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Recent advances in large multimodal models have enabled new opportunities in embodied AI, particularly in robotic manipulation. These models have shown strong potential in generalization and reasoning, but achieving reliable and responsible robotic behavior in real-world settings remains an open challenge. In high-stakes environments, robotic agents must go beyond basic task execution to perform risk-aware reasoning, moral decision-making, and physically grounded planning. We introduce ResponsibleRobotBench, a systematic benchmark designed to evaluate and accelerate progress in responsible robotic manipulation from simulation to real world. This benchmark consists of 23 multi-stage tasks spanning diverse risk types, including electrical, chemical, and human-related hazards, and varying levels of physical and planning complexity. These tasks require agents to detect and mitigate risks, reason about safety, plan sequences of actions, and engage human assistance when necessary. Our benchmark includes a general-purpose evaluation framework that supports multimodal model-based agents with various action representation modalities. The framework integrates visual perception, context learning, prompt construction, hazard detection, reasoning and planning, and physical execution. It also provides a rich multimodal dataset, supports reproducible experiments, and includes standardized metrics such as success rate, safety rate, and safe success rate. Through extensive experimental setups, ResponsibleRobotBench enables analysis across risk categories, task types, and agent configurations. By emphasizing physical reliability, generalization, and safety in decision-making, this benchmark provides a foundation for advancing the development of trustworthy, real-world responsible dexterous robotic systems. https://sites.google.com/view/responsible-robotbench

Problem

• Evaluating responsible robotic manipulation in hazardous environments
• Assessing risk-aware reasoning and safety decision-making in robots
• Benchmarking multimodal AI for physical reliability and generalization

Innovation

• Benchmark for responsible robot manipulation with multi-modal tasks
• Framework integrates visual perception, hazard detection, and planning
• Supports reproducible experiments with standardized safety and success metrics