AI Paper Atlas

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

Seeing What’s Wrong: A Trajectory-Guided Approach to Caption Error Detection

Gabriel Afriat, Ryan Lucas, Xiang Meng, Yufang Hou, Yada Zhu, Rahul Mazumder

ICLR 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Concept-guided image captioning

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
add TRACED caption trajectory on top of baseline scorerBLIP (ITM) · Flickr30k error detectionaccuracy: 88.5 -> 89.5Improved
add TRACED on top of CLIPCLIP · Flickr30k error detectionaccuracy: 83.8 -> 85.7Improved
use Elimination instead of GCD/Fast GCD for trajectories (ablation)TRACED · MM-IMDb error detectionaccuracy: LEMON FIX 77.6 (GCD) -> 78.3 (Elim)Improved
use the full trajectory instead of only the first step (ablation)TRACED · Flickr30k error detection% change in accuracy: -1.25 for first step onlyImproved
use full trajectory instead of last step only (ablation)TRACED · Flickr30k error detection% change in accuracy: -43.07 for last step onlyImproved
use full trajectory instead of mean trajectory (ablation)TRACED · Flickr30k error detection% change in accuracy: -4.90 for mean onlyImproved
combine alignment s and similarity c instead of either alone (ablation)TRACED · MS-COCO error detection% change in accuracy: -0.41 (s only), -6.03 (c only)Improved
maximize instead of minimize the alignment score when building trajectories (ablation)TRACED · Flickr30k error detection% change in accuracy: -0.76 for minimizing (BLIP ITM)Improved
increase trajectory length T (ablation)TRACED · error detectionaccuracyImproved
filter with TRACED-BLIP instead of BLIP before fine-tuningBLIP-2 · Flickr30k captioningBLEU-4: 30.8 -> 31.8Improved
filter with BLIP baseline instead of keeping noisy dataBLIP-2 · Flickr30k captioningBLEU-4: 31.5 -> 30.8Hurt
add TRACED hints to InternVL3-1B correction (ablation)InternVL3-1B · Flickr30k captioningBLEU-4: 32.0 -> 32.5Improved

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