Research map / Vision-language and multimodal / Visual question answering

Bootstrapping Variational Information Pursuit with Large Language and Vision Models for Interpretable Image Classification

ICLR 2024 · Vision-language and multimodal › Visual question answering · approach: Visual reasoning diagnostics

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Bootstraps Variational Information Pursuit (V-IP) by using large language and vision models to automatically generate the query set and answers for interpretable image classification.

Problem

Applying Variational Information Pursuit to a new task requires manually specifying a query set and collecting query answers for training, which is costly and limits interpretable image classification.

Approach

Use large language models to generate candidate interpretable queries and large vision models to answer them over images, so V-IP can be trained without manual query specification or annotation, yielding short query chains that serve as faithful explanations.

Experimental setup

Builds on: Variational Information Pursuit

Closest papers by meaning

• When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.86)
• Rethinking Token Reduction for Large Vision-Language Models (CVPR 2026 · similarity 0.86)
• Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy (ICLR 2024 · similarity 0.85)
• FRAMES-VQA: Benchmarking Fine-Tuning Robustness across Multi-Modal Shifts in Visual Question Answering (CVPR 2025 · similarity 0.85)
• Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.85)
• MoDA: Modulation Adapter for Fine-Grained Visual Understanding in Instructional MLLMs (ICML 2026 · similarity 0.85)