Pith. sign in

REVIEW 6 cited by

Why are Visually-Grounded Language Models Bad at Image Classification?

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2405.18415 v2 pith:LYDLVNYM submitted 2024-05-28 cs.CV cs.AIcs.CLcs.LG

classification cs.CVcs.AIcs.CLcs.LG
keywords classificationimagetrainingvlmsdatamodelscapabilitiesclip
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Image classification is one of the most fundamental capabilities of machine vision intelligence. In this work, we revisit the image classification task using visually-grounded language models (VLMs) such as GPT-4V and LLaVA. We find that existing proprietary and public VLMs, despite often using CLIP as a vision encoder and having many more parameters, significantly underperform CLIP on standard image classification benchmarks like ImageNet. To understand the reason, we explore several hypotheses concerning the inference algorithms, training objectives, and data processing in VLMs. Our analysis reveals that the primary cause is data-related: critical information for image classification is encoded in the VLM's latent space but can only be effectively decoded with enough training data. Specifically, there is a strong correlation between the frequency of class exposure during VLM training and instruction-tuning and the VLM's performance in those classes; when trained with sufficient data, VLMs can match the accuracy of state-of-the-art classification models. Based on these findings, we enhance a VLM by integrating classification-focused datasets into its training, and demonstrate that the enhanced classification performance of the VLM transfers to its general capabilities, resulting in an improvement of 11.8% on the newly collected ImageWikiQA dataset.

Discussion (0). Sign in to comment.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Unpacking Hateful Memes: Presupposed Context and False Claims

    cs.CL 2025-10 conditional novelty 6.0 of 10

    A hateful-meme detector that combines presupposed-context fusion, LLM-based social perception, and cross-modal reference graphs outperforms prior models on three benchmarks and transfers to fake news.

  2. Multi-modal Mutual-Guidance Conditional Prompt Learning for Vision-Language Models

    cs.CV 2025-07 conditional novelty 6.0 of 10

    MuGCP adapts CLIP by decoding instance-specific prompts from a frozen MLLM's KV cache and fusing them with visual prompts, achieving state-of-the-art few-shot classification on 14 datasets.

  3. AIGI-Holmes: Towards Explainable and Generalizable AI-Generated Image Detection via Multimodal Large Language Models

    cs.CV 2025-07 conditional novelty 6.0 of 10

    AIGI-Holmes combines visual expert pretraining, SFT on explanation data, and direct preference optimization to deliver human-verifiable explanations and top detection accuracy on unseen AI generators.

  4. Decomposing Visual Classification: Assessing Tree-Based Reasoning in VLMs

    cs.CV 2025-09 conditional novelty 5.0 of 10

    Tree-based reasoning prompts consistently underperform standard zero-shot prompting for VLM image classification on GTSRB and CIFAR-10 across three models.

  5. Filter-And-Refine: A MLLM Based Cascade System for Industrial-Scale Video Content Moderation

    cs.LG 2025-07 conditional novelty 4.0 of 10

    A cascade of an embedding router and a fine-tuned multimodal LLM ranker is claimed to improve content moderation F1 by 66.5% while using 1.5% of the compute of direct LLM deployment.

  6. Foundation Models for Astrophysics

    astro-ph.IM 2026-08 conditional novelty 3.0 of 10

    Astronomical 'foundation models' largely reuse transformers and self-supervised pretraining, but evidence of transfer to new instruments, populations, or tasks remains rare; the paper argues such evidence, not archite...

Pith tools