Localized Symbolic Knowledge Distillation for Visual Commonsense Models

TL;DR

This paper introduces a method to enhance vision-language models with localized reasoning capabilities by distilling knowledge from large language models, enabling precise referencing of image regions for improved commonsense understanding.

Contribution

It presents a novel approach to train localized visual commonsense models that support reference-based inputs, improving zero-shot reasoning in vision-language tasks.

Findings

Localized models outperform baseline in zero-shot reasoning

Training with localized commonsense improves model precision

Human evaluations favor the proposed distillation approach

Abstract

Instruction following vision-language (VL) models offer a flexible interface that supports a broad range of multimodal tasks in a zero-shot fashion. However, interfaces that operate on full images do not directly enable the user to "point to" and access specific regions within images. This capability is important not only to support reference-grounded VL benchmarks, but also, for practical applications that require precise within-image reasoning. We build Localized Visual Commonsense models, which allow users to specify (multiple) regions as input. We train our model by sampling localized commonsense knowledge from a large language model (LLM): specifically, we prompt an LLM to collect commonsense knowledge given a global literal image description and a local literal region description automatically generated by a set of VL models. With a separately trained critic model that selects high-quality examples, we find that training on the localized commonsense corpus can successfully distill existing VL models to support a reference-as-input interface. Empirical results and human evaluations in a zero-shot setup demonstrate that our distillation method results in more precise VL models of reasoning compared to a baseline of passing a generated referring expression to an LLM.