SafeInferCom improves robotic task planning with safe mid-generation checks
SafeInferCom: Safe Inference-Time Compute via Verifier-Guided Mid-Generation Intervention for Robotic Task Planning
RoboticsArtificial IntelligenceComputation and LanguageLogic in Computer Science
Summary
Robots often need to plan multiple steps to complete tasks correctly, but errors can happen during this planning, wasting time and causing failures. The authors created SafeInferCom, a system that checks and fixes planning errors while the robot is still thinking, without stopping the process. This approach helps robots avoid mistakes earlier and finish tasks more successfully and efficiently. They tested SafeInferCom with various robot models and simulations, including a real robot arm demonstration.
What this means in practice
- •For robotics engineers: Improve robot planning reliability and efficiency by integrating SafeInferCom to verify and correct plans during inference.
- •For automation system developers: Develop safer automated systems that use mid-generation verification to reduce errors in task execution sequences for complex workflows.
Authors
Weizhe Xu, Jialiang Fan, Mengyu Liu, Fanxin Kong
Abstract
Large Reasoning Language Models (LRLMs) enable multi-step reasoning for robotic task planning, but continued reasoning can overwrite valid intermediate plans or leave constraint violations unresolved, reducing planning reliability and wasting inference-time computation. We develop an inference-time monitor that exposes and verifies intermediate plans without disrupting the original decoding trajectory. Building on this monitor, we propose SafeInferCom, a formal verifier-guided framework that preserves valid intermediate plans and directs error correction during generation. Experiments across multiple LRLMs and planning domains reveal reasoning-response inconsistency and limited self-correction under one-shot inference. SafeInferCom improves planning success and accelerates error correction relative to one-shot inference. When combined with iterative refinement, it further improves success while reducing token usage compared with refinement alone. We additionally evaluate SafeInferCom in VirtualHome and provide a real-world robotic-arm demonstration.