AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Unveiling Effective In-Context Configurations for Image Captioning: An External & Internal Analysis

Li Li, Yongliang Wu, Jingze Zhu, Jiawei Peng, Jianfei Cai, and Xu Yang

TPAMI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Unified any-modality language models · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase the number of shotsOpenFlamingo and IDEFICS · MSCOCO captioningCIDEr and CLIPScoreMixed
use similarity-based image retrieval instead of random samplingOpenFlamingo and IDEFICS · MSCOCO captioningCIDEr and CLIPScoreMixed
use low-quality ICE captions (TF-60) with more shotsOpenFlamingo and IDEFICS · MSCOCO captioningCIDEr and CLIPScoreHurt
use high-quality ICE captionsOpenFlamingo · MSCOCO captioningCIDErImproved
use inverse MHL-TF (CIDEr-unbiased) ICE captionsOpenFlamingo · MSCOCO captioningCIDEr and CLIPScoreMixed
use model-generated captions (MHL) matched to human captionsOpenFlamingo and IDEFICS · MSCOCO captioningCIDErImproved
anchor-centric masking at layers 10-30 (4-shot) (ablation)IDEFICS · MSCOCOCIDEr: 97.3 -> 74.1Hurt
anchor-centric masking at layers 15-30 (4-shot) (ablation)IDEFICS · MSCOCOCIDEr: 97.3 -> 94.8No clear effect
context-centric (local window) masking at layers 10-30 (32-shot) (ablation)IDEFICS · MSCOCOCIDEr: 104.8 -> 105.1No clear effect
prune non-anchor tokens from layer 20 (32-shot) (ablation)IDEFICS · MSCOCO RS 32-shotCIDEr: 108.1 -> 108.0No clear effect
prune non-anchor tokens from layer 10 (32-shot), with recovery (ablation)IDEFICS · MSCOCO RS 32-shotCIDEr: 108.1 -> 69.95Hurt
recover pruned token states before prediction (Fast_10, 4-shot) (ablation)IDEFICS · MSCOCO RS 4-shotCIDEr: 55.5 -> 61.1Improved

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