FINER: MLLMs Hallucinate under Fine-grained Negative Queries

TL;DR

This paper introduces FINER, a set of benchmarks and a tuning method to analyze and reduce hallucinations in multimodal large language models when handling fine-grained queries, improving their accuracy and robustness.

Contribution

The paper presents FINER benchmarks and a fine-tuning approach using DPO to significantly reduce hallucinations in MLLMs on fine-grained queries.

Findings

Finetuning with FINER-Tuning reduces hallucinations by up to 24.2%.

Benchmarks reveal hallucinations occur with fine-grained mismatches and present elements.

Finetuning improves performance across multiple existing hallucination benchmarks.

Abstract

Multimodal large language models (MLLMs) struggle with hallucinations, particularly with fine-grained queries, a challenge underrepresented by existing benchmarks that focus on coarse image-related questions. We introduce FIne-grained NEgative queRies (FINER), alongside two benchmarks: FINER-CompreCap and FINER-DOCCI. Using FINER, we analyze hallucinations across four settings: multi-object, multi-attribute, multi-relation, and ``what'' questions. Our benchmarks reveal that MLLMs hallucinate when fine-grained mismatches co-occur with genuinely present elements in the image. To address this, we propose FINER-Tuning, leveraging Direct Preference Optimization (DPO) on FINER-inspired data. Finetuning four frontier MLLMs with FINER-Tuning yields up to 24.2\% gains (InternVL3.5-14B) on hallucinations from our benchmarks, while simultaneously improving performance on eight existing hallucination suites, and enhancing general multimodal capabilities across six benchmarks. Code, benchmark, and models are available at \href{https://explainableml.github.io/finer-project/}{https://explainableml.github.io/finer-project/}.