SpanCalib-VLM: Calibrated Hallucination Span Detection in Vision-Language Models

Amanuel Gizachew Abebe · hf · 2026-09-01

Introduces SpanCalib-VLM, a hybrid system designed to address hallucinations in Vision-Language Models (VLMs). It combines a multimodal sequence tagger with a generative VLM to improve hallucination span detection and calibration through union-calibrated fusion.

Original post →

More from Multimodal

Multimodal channel →