CIEM: Contrastive Instruction Evaluation Method for Better Instruction Tuning
This addresses hallucination issues in VLMs for applications like captioning, though it appears incremental as it builds on existing instruction tuning approaches.
The paper tackles hallucination in Vision-Language Models (VLMs) by introducing CIEM, an automatic evaluation pipeline using annotated data and LLMs to generate factual/contrastive question-answer pairs, and CIT, a tuning method that produces such pairs to reduce hallucination, showing superiority of CIT-tuned VLMs over existing methods and datasets.
Nowadays, the research on Large Vision-Language Models (LVLMs) has been significantly promoted thanks to the success of Large Language Models (LLM). Nevertheless, these Vision-Language Models (VLMs) are suffering from the drawback of hallucination -- due to insufficient understanding of vision and language modalities, VLMs may generate incorrect perception information when doing downstream applications, for example, captioning a non-existent entity. To address the hallucination phenomenon, on the one hand, we introduce a Contrastive Instruction Evaluation Method (CIEM), which is an automatic pipeline that leverages an annotated image-text dataset coupled with an LLM to generate factual/contrastive question-answer pairs for the evaluation of the hallucination of VLMs. On the other hand, based on CIEM, we further propose a new instruction tuning method called CIT (the abbreviation of Contrastive Instruction Tuning) to alleviate the hallucination of VLMs by automatically producing high-quality factual/contrastive question-answer pairs and corresponding justifications for model tuning. Through extensive experiments on CIEM and CIT, we pinpoint the hallucination issues commonly present in existing VLMs, the disability of the current instruction-tuning dataset to handle the hallucination phenomenon and the superiority of CIT-tuned VLMs over both CIEM and public datasets.