Seeing is Believing: Rich-Context Hallucination Detection for MLLMs via Backward Visual Grounding

Documenti analoghi

• TagOOD: A Novel Approach to Out-of-Distribution Detection via Vision-Language Representations and Class Center Learning
  di: Li, Jinglun, et al.
  Pubblicazione: (2024)
• DeTrack: In-model Latent Denoising Learning for Visual Object Tracking
  di: Zhou, Xinyu, et al.
  Pubblicazione: (2025)
• Unified Multimodal Visual Tracking with Dual Mixture-of-Experts
  di: Hong, Lingyi, et al.
  Pubblicazione: (2026)
• LingoLoop Attack: Trapping MLLMs via Linguistic Context and State Entrapment into Endless Loops
  di: Fu, Jiyuan, et al.
  Pubblicazione: (2025)
• X-Prompt: Multi-modal Visual Prompt for Video Object Segmentation
  di: Guo, Pinxue, et al.
  Pubblicazione: (2024)