User-in-the-loop Evaluation of Multimodal LLMs for Activity Assistance

TL;DR

This study evaluates multimodal large language models for activity assistance, focusing on visual grounding, activity forecasting, and user-in-the-loop replanning through offline benchmarks and a pioneering user study.

Contribution

It introduces a comprehensive evaluation framework combining offline benchmarks and the first user-in-the-loop study for multimodal LLMs in activity assistance.

Findings

Socratic Models outperform VCLMs in offline and online tasks.

Grounding long visual history remains challenging for current models.

Offline metrics do not reliably predict online performance.

Abstract

Our research investigates the capability of modern multimodal reasoning models, powered by Large Language Models (LLMs), to facilitate vision-powered assistants for multi-step daily activities. Such assistants must be able to 1) encode relevant visual history from the assistant's sensors, e.g., camera, 2) forecast future actions for accomplishing the activity, and 3) replan based on the user in the loop. To evaluate the first two capabilities, grounding visual history and forecasting in short and long horizons, we conduct benchmarking of two prominent classes of multimodal LLM approaches -- Socratic Models and Vision Conditioned Language Models (VCLMs) on video-based action anticipation tasks using offline datasets. These offline benchmarks, however, do not allow us to close the loop with the user, which is essential to evaluate the replanning capabilities and measure successful activity completion in assistive scenarios. To that end, we conduct a first-of-its-kind user study, with 18 participants performing 3 different multi-step cooking activities while wearing an egocentric observation device called Aria and following assistance from multimodal LLMs. We find that the Socratic approach outperforms VCLMs in both offline and online settings. We further highlight how grounding long visual history, common in activity assistance, remains challenging in current models, especially for VCLMs, and demonstrate that offline metrics do not indicate online performance.