AI Paper Atlas

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

Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models

Yuriel Ryan, Ip Hei Man, Adriel Kuek, Paul Pu Liang, Roy Ka-Wei Lee

ICML 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multimodal fusion with missing modalities · 0 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 VLM-generated captions to the text modalityHateful Memes training data · Hateful Memes trainredundancy R: 0.0553 -> 0.2319 (+319%)Improved
add VLM-generated captions to the text modalityHateful Memes training data · Hateful Memes trainunique visual U_V: 0.3465 -> 0.1710 (-51%)Improved
use meaningful captions instead of random text (ablation)interaction transfer · Hateful Memes testredundancy R: 0.0653 -> 0.0939Improved
use larger captioner (Qwen2.5 32B) instead of SmolVLM-2B (ablation)interaction transfer · Hateful Memes testredundancy R: 0.0752 -> 0.0939Improved
caption videos in a synergy-dominant datasetUR-FUNNY · UR-FUNNY testunique text U_T: 0.0004 -> 0.0034 (+750.0%)Hurt
caption videos in a synergy-dominant datasetUR-FUNNY · UR-FUNNY testsynergy S: 0.0043 -> 0.0030 (-30.2%)Hurt
add 50% extra redundancy via self-captioningSmolVLM · macro averagevisual-induced errors: -38.3%Improved
add 50% extra redundancy via self-captioningSmolVLM · macro averageaccuracy and consistency: +4.0% accuracy, +16.8% consistencyImproved
add 50% extra redundancy via self-captioningSmolVLM · macro averagelanguage-induced and mixed errors: +15.2% LI, +42.9% MixHurt
add 25% extra redundancy via self-captioningLLaVA-OneVision · macro averagevisual-induced errors: -34.4%Improved
add 25-50% redundancySmolVLM and LLaVA-OneVision · corruption severityperformance stabilityImproved
caption 100% of training samples (ablation)SmolVLM-2B · hate speech detectionperformance stability: DeltaP approx -4.07Hurt

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