SOTAVerified

Visual Entailment

Visual Entailment (VE) - is a task consisting of image-sentence pairs whereby a premise is defined by an image, rather than a natural language sentence as in traditional Textual Entailment tasks. The goal is to predict whether the image semantically entails the text.

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–10 of 56 papers

TitleStatusHype
Defeasible Visual Entailment: Benchmark, Evaluator, and Reward-Driven OptimizationCode1
Advancing Grounded Multimodal Named Entity Recognition via LLM-Based Reformulation and Box-Based SegmentationCode1
Understanding Figurative Meaning through Explainable Visual EntailmentCode1
MoPE: Mixture of Prompt Experts for Parameter-Efficient and Scalable Multimodal FusionCode1
LLMs as Bridges: Reformulating Grounded Multimodal Named Entity RecognitionCode1
Do LVLMs Understand Charts? Analyzing and Correcting Factual Errors in Chart CaptioningCode1
Good Questions Help Zero-Shot Image ReasoningCode1
I Spy a Metaphor: Large Language Models and Diffusion Models Co-Create Visual MetaphorsCode1
Benchmarking Robustness of Multimodal Image-Text Models under Distribution ShiftCode1
Harnessing the Power of Multi-Task Pretraining for Ground-Truth Level Natural Language ExplanationsCode1
Show:102550
← PrevPage 1 of 6Next →

No leaderboard results yet.