QCRI / MemeLens-VLM

MemeLens is a unified multilingual, multitask Vision-Language Model (VLM) for meme understanding. It is fine-tuned from Qwen3-VL-8B-Instruct using a classify-then-explain training strategy on the MemeLens dataset, which consolidates 38 public meme datasets across 20 tasks and …