Butter-Bench: Evaluating LLM Controlled Robots for Practical Intelligence

TL;DR

Butter-Bench is a benchmark designed to evaluate the practical intelligence of LLM-controlled robots in real-world scenarios, revealing current limitations in multi-step planning and social understanding.

Contribution

This paper introduces Butter-Bench, a new benchmark for assessing LLMs in robotic tasks, focusing on their reasoning abilities separate from low-level control.

Findings

LLMs score 40% on Butter-Bench

Humans score 95% on Butter-Bench

Fine-tuning LLMs for embodied reasoning does not improve scores

Abstract

We present Butter-Bench, a benchmark evaluating large language model (LLM) controlled robots for practical intelligence, defined as the ability to navigate the messiness of the physical world. Current state-of-the-art robotic systems use a hierarchical architecture with LLMs in charge of high-level reasoning, and a Vision Language Action (VLA) model for low-level control. Butter-Bench evaluates the LLM part in isolation from the VLA. Although LLMs have repeatedly surpassed humans in evaluations requiring analytical intelligence, we find humans still outperform LLMs on Butter-Bench. The best LLMs score 40% on Butter-Bench, while the mean human score is 95%. LLMs struggled the most with multi-step spatial planning and social understanding. We also evaluate LLMs that are fine-tuned for embodied reasoning and conclude that this training does not improve their score on Butter-Bench.