LEASH: Stay Safe in Agentic Robotics
LEASH: Stay Safe in Agentic Robotics
This paper introduces RoboRisk, a comprehensive benchmark for evaluating safety in agentic robotics, covering five categories across 92 simulated tasks and 13 physical-robot tasks. The authors evaluate six foundation models (GPT-6 Astra/Sol/Luna, Claude Opus 5.5, Gemini 4 Argon, Grok 4.7) and two robot APIs (Dream API, PokeBot's RoboAPI), finding high harm rates. They propose LEASH (Lease-Enforced Agentic Safety Harness), which uses a decision model to implement task, action, and commit gates in visual context, effectively reducing harm across all evaluated models while preserving task completion performance with minimal overhead.
The authors introduce RoboRisk, a benchmark covering five safety categories across simulated and physical tasks, and propose LEASH, a safety harness that reduces harm while preserving task completion for six foundation models and two robot APIs.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
tasks
92
View original evidence
ng. We introduce RoboRisk, a comprehensive benchmark for evaluating safety in agentic robotics. Its 92 simulated tasks and 13 physical-robot tasks fall into five categories: Task-level Refusal (T), Plan Repair (P), Execution-time Hazard (E), Injection (I), and Sequential Memory (S). Prior benchmarks cover at most two of these categories, and several evaluate only text outputs. In contrast, RoboRisk covers all five an
Open source S3tasks
13
View original evidence
ng. We introduce RoboRisk, a comprehensive benchmark for evaluating safety in agentic robotics. Its 92 simulated tasks and 13 physical-robot tasks fall into five categories: Task-level Refusal (T), Plan Repair (P), Execution-time Hazard (E), Injection (I), and Sequential Memory (S). Prior benchmarks cover at most two of these categories, and several evaluate only text outputs. In contrast, RoboRisk covers all five an
Open source S3Robot count
2 robots
View original evidence
d executes every task in closed loop. We evaluate six state-of-the-art foundation models on RoboRisk: GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna, Claude Opus 5.5, Gemini 4 Argon, and Grok 4.7. These models generally show high harm rates. We also evaluate two robot-specific APIs: the Dream API and PokeBot's RoboAPI. We further propose the Lease-Enforced Agentic Safety Harness (LEASH), which offers essential safety infrastru
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ng. We introduce RoboRisk, a comprehensive benchmark for evaluating safety in agentic robotics. Its 92 simulated tasks and 13 physical-robot tasks fall into five categories: Task-level Refusal (T), Plan Repair (P), Execution-time Hazard (E), Injection (I), and Sequential Memory (S). Prior benchmarks cover at most two of these categories, and several evaluate only text outputs. In contrast, RoboRisk covers all five an
Open source S3
d executes every task in closed loop. We evaluate six state-of-the-art foundation models on RoboRisk: GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna, Claude Opus 5.5, Gemini 4 Argon, and Grok 4.7. These models generally show high harm rates. We also evaluate two robot-specific APIs: the Dream API and PokeBot's RoboAPI. We further propose the Lease-Enforced Agentic Safety Harness (LEASH), which offers essential safety infrastru
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org