AI Paper Atlas

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

Cola: A Benchmark for Compositional Text-to-image Retrieval

Arijit Ray, Filip Radenovic, Abhimanyu Dubey, Bryan A. Plummer, Ranjay Krishna, Kate Saenko

NeurIPS 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Composed and text-based image retrieval · 2 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
replace fine-tuning late layers with MM-AdapterCLIP · Cola GQA single-objectmAP: 42.19 -> 46.83Improved
replace fine-tuning late layers with MM-AdapterCLIP · Cola CLEVR single-objectmAP: 64.06 -> 88.21Improved
replace fine-tuning late layers with MM-AdapterCLIP · Cola PACO single-objectmAP: 15.66 -> 18.56Improved
use MM-Adapter instead of MM-Pred (attention adapts the representation) (ablation)CLIP · Cola CLEVR single-objectmAP: 75.80 -> 88.21Improved
use MM-Adapter instead of MM-Pred (ablation)FLAVA · Cola GQA single-objectmAP: 47.12 -> 48.54Improved
add MM-Adapter on top of FLAVAFLAVA · Cola CLEVR single-objectmAP: 15.41 -> 91.10Improved
add MM-Adapter on CLIP vs fine-tuning FLAVA late layersCLIP · Cola GQA single-objectmAP: 39.58 -> 46.83Improved
add MM-AdapterCLIP · Cola multi-object T2Iaccuracy: 21.42 -> 40.95Improved
add MM-Pred or MM-Adapter on FLAVAFLAVA · Cola multi-object T2Iaccuracy: 24.76 -> 40.47Improved
add MM-PredCLIP · CREPE I2Taccuracy: 77.43 -> 77.84No clear effect
train on combined data instead of region captions (ablation)MM-Adapter · Cola GQA single-objectmAP: 0.4711 -> 0.4788Improved
add multi-object compounds to training (ablation)MM-Adapter · Cola multi-objectaccuracy: 0.2745 -> 0.3975Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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