Exposing and Addressing Cross-Task Inconsistency in Unified Vision-Language Models

TL;DR

This paper introduces CocoCon, a benchmark dataset for measuring cross-task consistency in vision-language models, revealing significant inconsistency issues and proposing a training method to improve consistency without sacrificing accuracy.

Contribution

The paper presents CocoCon, a novel benchmark for evaluating cross-task consistency, and proposes a rank correlation-based training objective to enhance model consistency across heterogeneous tasks.

Findings

State-of-the-art models exhibit high inconsistency across tasks.

The proposed training method improves multi-task consistency.

Models retain their original accuracy after applying the new training objective.

Abstract

As general purpose vision models get increasingly effective at a wide set of tasks, it is imperative that they be consistent across the tasks they support. Inconsistent AI models are considered brittle and untrustworthy by human users and are more challenging to incorporate into larger systems that take dependencies on their outputs. Measuring consistency between very heterogeneous tasks that might include outputs in different modalities is challenging since it is difficult to determine if the predictions are consistent with one another. As a solution, we introduce a benchmark dataset, CocoCon, where we create contrast sets by modifying test instances for multiple tasks in small but semantically meaningful ways to change the gold label and outline metrics for measuring if a model is consistent by ranking the original and perturbed instances across tasks. We find that state-of-the-art vision-language models suffer from a surprisingly high degree of inconsistent behavior across tasks, especially for more heterogeneous tasks. To alleviate this issue, we propose a rank correlation-based auxiliary training objective, computed over large automatically created cross-task contrast sets, that improves the multi-task consistency of large unified models while retaining their original accuracy on downstream tasks.