Flat-Pack Bench: Evaluating Spatio-Temporal Understanding in Large Vision-Language Models through Furniture Assembly

2026-05-20

AI Summary

Existing vision-language model benchmarks primarily focus on coarse-grained video understanding, making them inadequate for evaluating fine-grained spatiotemporal reasoning in complex real-world scenarios. To address this gap, this work proposes Flat-Pack Bench—the first fine-grained video understanding benchmark centered on furniture assembly. It systematically assesses large vision-language models through multiple-choice question answering augmented with visual prompting, targeting capabilities such as action sequence ordering, state localization, part pairing, and object tracking. Emphasizing spatiotemporal interactions and physical contact understanding, the benchmark reveals significant shortcomings in current state-of-the-art models, highlighting limitations in temporal modeling, object tracking, and comprehension of spatial interactions.

Abstract

The emergence of Large Vision-Language Models (LVLMs) has significantly advanced video understanding capabilities. However, existing benchmarks focus predominantly on coarse-grained tasks such as action segmentation, classification, captioning, and retrieval. Furthermore, these benchmarks often rely on entities that can be easily identified verbally, like household objects, animals, human subjects, etc., limiting their applicability to complex, in-the-wild video scenarios. But, many applications such as furniture assembly, cooking, etc., require step-by-step fine-grained spatio-temporal understanding of the video, which is not sufficiently evaluated in current benchmarks. To address this gap, we introduce Flat-Pack Bench, a novel benchmark centered on furniture assembly tasks. Our benchmark evaluates LVLMs on nuanced tasks, including temporal ordering of assembly actions, temporal localization of assembly state, understanding part mating, and tracking, using multiple-choice questions paired with visual prompts highlighting relevant parts as references for fine-grained questions. Our experiments reveal that state-of-the-art LVLMs struggle significantly with fine-grained spatio-temporal reasoning, highlighting their limitations in effectively leveraging temporal information from videos, limited tracking ability, and understanding of spatial interactions like physical contact.

Problem

• spatio-temporal understanding
• Large Vision-Language Models
• fine-grained video reasoning
• furniture assembly
• video benchmark

Innovation

• spatio-temporal understanding
• Large Vision-Language Models
• fine-grained video reasoning
• furniture assembly benchmark
• temporal localization