Pith. sign in

REVIEW 18 cited by

Benchmarking and Improving Detail Image Caption

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 2405.19092 v4 pith:PVKPIFFD submitted 2024-05-29 cs.CV

Benchmarking and Improving Detail Image Caption

classification cs.CV
keywords captionevaluationdataimagedetaillvlmcaptioningcapture
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Image captioning has long been regarded as a fundamental task in visual understanding. Recently, however, few large vision-language model (LVLM) research discusses model's image captioning performance because of the outdated short-caption benchmarks and unreliable evaluation metrics. In this work, we propose to benchmark detail image caption task by curating high-quality evaluation datasets annotated by human experts, GPT-4V and Gemini-1.5-Pro. We also design a more reliable caption evaluation metric called CAPTURE (CAPtion evaluation by exTracting and coUpling coRE information). CAPTURE extracts visual elements, e.g., objects, attributes and relations from captions, and then matches these elements through three stages, achieving the highest consistency with expert judgements over other rule-based or model-based caption metrics. The proposed benchmark and metric provide reliable evaluation for LVLM's detailed image captioning ability. Guided by this evaluation, we further explore to unleash LVLM's detail caption capabilities by synthesizing high-quality data through a five-stage data construction pipeline. Our pipeline only uses a given LVLM itself and other open-source tools, without any human or GPT-4V annotation in the loop. Experiments show that the proposed data construction strategy significantly improves model-generated detail caption data quality for LVLMs with leading performance, and the data quality can be further improved in a self-looping paradigm. All code and dataset will be publicly available at https://github.com/foundation-multimodal-models/CAPTURE.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 18 Pith papers

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

  1. MentalThink: Shaping Thoughts in Mental SVG World

    cs.AI 2026-07 conditional novelty 7.0

    MLLMs that generate and render SVG sketches as multi-turn intermediate reasoning steps reach 55.1% on VSIBench and 76.0% on MindCube, far above the Qwen2.5-VL-7B backbone.

  2. Mitigating Mask Prior Drift and Positional Attention Collapse in Large Diffusion Vision-Language Models

    cs.CV 2026-05 unverdicted novelty 7.0

    Mask prior drift and positional attention collapse cause failures in LDVLMs for long generations, fixed by training-free Mask Prior Suppression and Monotonic RoPE Scaling.

  3. Evaluating Remote Sensing Image Captions Beyond Metric Biases

    cs.CV 2026-04 unverdicted novelty 7.0

    Unfine-tuned MLLMs outperform fine-tuned models on remote sensing image captioning when captions are scored by their ability to reconstruct the source image, and a training-free self-correction method achieves SOTA pe...

  4. DetailVerifyBench: A Benchmark for Dense Hallucination Localization in Long Image Captions

    cs.CV 2026-04 unverdicted novelty 7.0

    DetailVerifyBench supplies 1,000 images and densely annotated long captions to evaluate precise hallucination localization in multimodal large language models.

  5. CaptionQA: Is Your Caption as Useful as the Image Itself?

    cs.CV 2025-11 conditional novelty 7.0

    CaptionQA is a new benchmark with 33,027 questions across natural, document, e-commerce, and embodied AI domains that measures how much utility model-generated captions retain compared to original images when used by ...

  6. VGR: Visual Grounded Reasoning

    cs.CV 2025-06 unverdicted novelty 7.0

    VGR introduces a visual-grounded reasoning MLLM that detects and replays image regions during inference, achieving gains on visual benchmarks with 30% fewer image tokens than the LLaVA-NeXT-7B baseline.

  7. Holo-Captioning: Toward the Text Equivalent of 3D Scenes

    cs.CV 2026-07 conditional novelty 6.5

    HoloScribe generates pure-text structured captions of all entities, boxes, attributes and relations in 3D indoor scenes and outperforms dense captioners and 3D LLMs on a new 15K-scene benchmark.

  8. Reliability-Prioritized Fine-Grained Generation in Multimodal Large

    cs.CV 2026-06 unverdicted novelty 6.0

    Presents GranFact benchmark with expert annotations and a reliability-prioritized DPO method to improve fine-grained yet reliable generation in MLLMs.

  9. Seeing Time: Benchmarking Chronological Reasoning and Shortcut Biases in Vision-Language Models

    cs.AI 2026-06 unverdicted novelty 6.0

    Introduces ChronoVision benchmark with three datasets showing VLMs rely on superficial cues such as color filters rather than genuine chronological reasoning.

  10. ChartFI: Benchmarking Faithfulness and Insightfulness of Chart Descriptions from Multimodal Large Language Models

    cs.CL 2026-05 unverdicted novelty 6.0

    ChartFI-Bench supplies 896 chart-description pairs and four metrics (Faithfulness, Coverage, Informativeness, Acuity) to evaluate MLLM-generated chart descriptions on faithfulness and insightfulness.

  11. ChartFI: Benchmarking Faithfulness and Insightfulness of Chart Descriptions from Multimodal Large Language Models

    cs.CL 2026-05 unverdicted novelty 6.0

    ChartFI-Bench supplies 896 chart-description pairs from visually complex charts and defines four metrics (Faithfulness, Coverage, Informativeness, Acuity) aligned to four quality dimensions to evaluate MLLM-generated ...

  12. ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison

    cs.LG 2026-05 unverdicted novelty 6.0

    ClaimDiff-RL replaces holistic scalar rewards with reference-conditioned atomic claim differences verified by a multimodal judge to improve the hallucination-missing-fact tradeoff in long-form image captioning.

  13. ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison

    cs.LG 2026-05 unverdicted novelty 6.0

    ClaimDiff-RL introduces reference-conditioned atomic claim differences verified by a multimodal judge as the reward signal for fine-grained RL in long-form image captioning.

  14. Mitigating Mask Prior Drift and Positional Attention Collapse in Large Diffusion Vision-Language Models

    cs.CV 2026-05 unverdicted novelty 6.0

    Diagnoses mask prior drift and positional attention collapse in LDVLMs and introduces two plug-and-play decoding interventions that raise long-form generation quality without retraining.

  15. Learning to See What You Need: Gaze Attention for Multimodal Large Language Models

    cs.CV 2026-05 unverdicted novelty 6.0

    Gaze Attention groups visual embeddings into selectable regions and dynamically restricts attention to task-relevant ones, matching dense baselines with up to 90% fewer visual KV entries via added context tokens.

  16. MSD-Score: Multi-Scale Distributional Scoring for Reference-Free Image Caption Evaluation

    cs.CV 2026-05 unverdicted novelty 6.0

    MSD-Score introduces multi-scale distributional scoring on von Mises-Fisher mixtures to evaluate image captions without references and reports state-of-the-art correlation with human judgments.

  17. Reliability-Prioritized Fine-Grained Generation in Multimodal Large

    cs.CV 2026-06 unverdicted novelty 5.0

    Proposes GranFact benchmark with coarse-to-fine annotations and a DPO variant that penalizes unreliable fine-grained claims to improve reliable specificity in MLLM outputs.

  18. Mitigating Entangled Steering in Large Vision-Language Models for Hallucination Reduction

    cs.CV 2026-04 unverdicted novelty 5.0

    MESA reduces hallucinations in LVLMs via controlled selective latent intervention that preserves the original token distribution.