Compositional Chain-of-Thought Prompting for Large Multimodal Models

27 Nov 2023  ·  Chancharik Mitra, Brandon Huang, Trevor Darrell, Roei Herzig ·

The combination of strong visual backbones and Large Language Model (LLM) reasoning has led to Large Multimodal Models (LMMs) becoming the current standard for a wide range of vision and language (VL) tasks. However, recent research has shown that even the most advanced LMMs still struggle to capture aspects of compositional visual reasoning, such as attributes and relationships between objects. One solution is to utilize scene graphs (SGs)--a formalization of objects and their relations and attributes that has been extensively used as a bridge between the visual and textual domains. Yet, scene graph data requires scene graph annotations, which are expensive to collect and thus not easily scalable. Moreover, finetuning an LMM based on SG data can lead to catastrophic forgetting of the pretraining objective. To overcome this, inspired by chain-of-thought methods, we propose Compositional Chain-of-Thought (CCoT), a novel zero-shot Chain-of-Thought prompting method that utilizes SG representations in order to extract compositional knowledge from an LMM. Specifically, we first generate an SG using the LMM, and then use that SG in the prompt to produce a response. Through extensive experiments, we find that the proposed CCoT approach not only improves LMM performance on several vision and language VL compositional benchmarks but also improves the performance of several popular LMMs on general multimodal benchmarks, without the need for fine-tuning or annotated ground-truth SGs. Code: https://github.com/chancharikmitra/CCoT

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Visual Reasoning Winoground LLaVA-1.5-CCoT Text Score 42.0 # 30
Image Score 35.5 # 16
Group Score 22.3 # 23
Visual Reasoning Winoground LLaVA-1.5-ZS-CoT Text Score 28.0 # 78
Image Score 22.5 # 43
Group Score 12.3 # 55
Visual Reasoning Winoground LLaVA-1.5 Text Score 36.0 # 48
Image Score 33.3 # 17
Group Score 20.1 # 30
Visual Reasoning Winoground InstructBLIP-CCoT Text Score 21.0 # 96
Image Score 21.3 # 47
Group Score 8.3 # 77
Visual Reasoning Winoground InstructBLIP-ZS-CoT Text Score 9.3 # 112
Image Score 16.3 # 61
Group Score 4.0 # 93
Visual Reasoning Winoground InstructBLIP Text Score 7.0 # 113
Image Score 11.5 # 86
Group Score 3.3 # 100

Methods


No methods listed for this paper. Add relevant methods here