Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note2 min read

Symbal: Detecting Systematic Misalignments in Model-Generated Captions

Source

arXiv

arxiv.org/abs/2607.15216v1

Read the source

Source

Plain-Language Brief

Researchers present Symbal, a system that identifies systematic misalignments in image-text pairs generated by multimodal large language models (MLLMs). The system uses a structured dual-stage setup and performs well on the introduced benchmark, SymbalBench. Symbal can be used to audit MLLM-generated captions and identify critical errors without requiring access to the underlying MLLM.

What the Document Says

  • The study focuses on systematic misalignments in image-text pairs generated by multimodal large language models (MLLMs).
  • Symbal is presented, which uses a structured dual-stage setup and off-the-shelf foundation models to identify such errors.
  • SymbalBench, a benchmark for evaluating automated methods on systematic misalignment detection, is introduced. It consists of 1.7 million image-text pairs from two domains with annotated systematic misalignments.
  • Symbal exhibits strong performance on SymbalBench, correctly identifying systematic misalignments in 63.8% of datasets.
  • Real-world evaluations show that Symbal can accurately surface systematic misalignments in captions generated by four MLLMs and is a powerful tool for auditing off-the-shelf image-caption datasets.
  • The code for Symbal is available at https://github.com/Stanford-AIMI/Symbal.

Key Concepts Mentioned

  • multimodal large language models (MLLMs)
  • systematic misalignments
  • image-text pairs
  • Symbal
  • SymbalBench
  • benchmark

Claims Made by the Document

  • Symbal correctly identifies systematic misalignments in 63.8% of datasets.
  • Symbal can accurately surface systematic misalignments in captions generated by four MLLMs.

Evidence Used

  • Symbal exhibits strong performance on SymbalBench, correctly identifying systematic misalignments in 63.8% of datasets.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Multimodal large language models (MLLMs) often introduce errors when generating image captions, resulting in misaligned image-text pairs. Our work focuses on a class of captioning errors that we refer to as systematic misalignments, where a recurring error in MLLM-generated captions is closely associated with the presence of a specific visual feature in the paired image. Given a vision-language dataset with MLLM-generated captions, our aim in this work is to detect such errors, a task we refer to as systematic misalignment detection. As our first key contribution, we present Symbal, which utilizes a structured, dual-stage setup with off-the-shelf foundation models to identify systematic misalignments and summarize results in natural language. As our second key contribution, we introduce SymbalBench, a benchmark designed to evaluate automated methods on our proposed task. SymbalBench consists of 1.7 million image-text pairs from two domains (natural and medical images), organized into 420 vision-language datasets with annotated systematic misalignments. Symbal exhibits strong performance on this benchmark, correctly identifying systematic misalignments in 63.8% of datasets, a nearly 4x improvement over the closest baseline. We supplement our evaluations on SymbalBench with real-world evaluations, showing that (1) Symbal can accurately surface systematic misalignments in captions generated by four MLLMs and (2) Symbal is a powerful tool for auditing off-the-shelf image-caption datasets. Ultimately, our novel task, method, and benchmark can aid users with auditing MLLM-generated captions and identifying critical errors, without requiring access to the underlying MLLM. Code is available at https://github.com/Stanford-AIMI/Symbal.