AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Where Do Long Captions Fail? Position-Aware Diagnosis and Reinforcement Learning for Detailed Image Captioning

Yuanze Hu, Zhichao Yang, Junwei Jing, Xin Yang, Wenxuan Liu, Tinghai Zhang, Jia Fu, Hongzhi Zhang, Xiangyu Wu, Ruiming Tang, Tingting Gao, Han Li

NeurIPS 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Concept-guided image captioning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Diagnoses how detailed image captions fail across sentence positions using a position-aware rubric of atomic visual claims, and uses this to guide reinforcement learning for long-form captioning.

Problem

Detailed image captioning is usually framed as trading off coverage of visual facts against hallucination, which hides that captions are ordered sequences where different failures occur at different positions.

Approach

Decomposes dense human reference captions into atomic visual claims arranged by position in the caption, enabling position-aware diagnosis of failures and a position-aware reinforcement learning signal for long captions.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add middle-position reward for supported fine-grained evidence (R_mid) (ablation)PAGER · diagnostic subsetD_mid: +.103 -> +.036Improved
add tail generic-escape penalty (R_tail) (ablation)PAGER · diagnostic subsetS_Q4: .374 -> .436Improved
add tail generic-escape penalty (R_tail) (ablation)PAGER · diagnostic subsetG_Q4: .378 -> .329Improved
add content-aware length budget (R_len) (ablation)PAGER · diagnostic subsetD_mid: +.047 -> +.036Improved
add PixMo rubric filtering (ablation)PAGER · diagnostic subsetD_mid: +.064 -> +.036Improved
use itemwise evidence-span reward instead of holistic checklist (ablation)PAGER · diagnostic subsetD_mid: +.058 -> +.036Improved
split middle/tail rewards instead of a monotone late-position weight (ablation)PAGER · diagnostic subsetH_Q4: .291 -> .235Improved
monotone late-position weight vs split rewards on D_mid (ablation)PAGER · diagnostic subsetD_mid: +.037 vs +.036No clear effect
apply PAGER GRPO to Qwen3-VL-4B-InstructQwen3-VL-4B · diagnostic subsetD_mid: +.096 -> +.036Improved
apply PAGER GRPOQwen3-VL-4B · diagnostic subsetH_Q4: .298 -> .235Improved
PAGER versus CapRL on DCScore-F1PAGER · IIW-400DCScore-F1: .552 (CapRL) vs .526Hurt
PAGER versus CapRL on tail genericityPAGER · diagnostic subsetD_tail: +.123 (CapRL) vs +.177Hurt

Experimental setup

Datasets: dense human reference captions

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy