AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

FALCON: False-Negative Aware Learning of Contrastive Negatives in Vision-Language Alignment

Myunsoo Kim, Seongwoong Shim, Byung-Jun Lee

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 3 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
add FALCON learned negative scheduler instead of GRIT-VLP fixed hardest-negative samplingALBEF · MSCOCO IRTR text retrievalR@1: 60.60 -> 62.28Improved
add FALCON instead of MAFA false-negative relabelingALBEF · MSCOCO IRTR image retrievalR@1: 44.77 -> 46.18Improved
use only easiest negatives (q=0.0) instead of hard onesALBEF · MSCOCO IRTR text retrievalR@1: 62.28 -> 18.92Hurt
add FALCON to BLIP-2BLIP-2 · OKVQA testaccuracy: 17.94 -> 20.96Improved
add GRIT-VLP hard negatives to BLIP-2BLIP-2 · VQA2 valaccuracy: 42.46 -> 39.91Hurt
add FALCON to BLIP-2BLIP-2 · COCO Stage-1 text retrievalR@1: 75.22 -> 75.56Improved
add FALCON to SigLIP-2SigLIP-2 · MSCOCO text retrievalR@1: 69.96 -> 72.96Improved
add FALCON to SigLIP-2SigLIP-2 · MSCOCO image retrievalR@1: 54.21 -> 54.15No clear effect
increase search space size from 480 to 5664 (ablation)FALCON · MSCOCO IRTR text retrievalR@1: 58.48 -> 61.72Improved
increase search space size from 5664 to 28320 (ablation)FALCON · MSCOCO IRTR text retrievalR@1: 61.72 -> 61.94No clear effect
use MLM alone as scheduler proxy instead of ITC+ITM (ablation)FALCON · MSCOCO IRTR text retrievalR@1: 57.64 -> 61.72Improved
use instance-level instead of batch-level scheduling (ablation)FALCON · MSCOCO IRTR text retrievalR@1: 58.78 -> 61.72Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 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