Data Party THU

Jun 10, 2026 · Artificial Intelligence

How Visual Para-Thinker Tackles Visual Hallucination with a Clever Parallel Reasoning Design

The article introduces Visual Para-Thinker, a parallel reasoning framework for large vision‑language models that mitigates attention drift and visual hallucination by employing path‑aware attention, learnable parallel rotary position embeddings, and hybrid block‑and‑scan visual token partitions, and validates the approach with extensive multimodal benchmarks.