PolicyShiftGuard: Benchmarking and Improving Policy-Adaptive Image Guardrails Paper • 2607.05910 • Published 20 days ago • 38
X-Stream: Exploring MLLMs as Multiplexers for Multi-Stream Understanding Paper • 2606.02482 • Published Jun 1 • 36
WBench: A Comprehensive Multi-turn Benchmark for Interactive Video World Model Evaluation Paper • 2605.25874 • Published May 25 • 106
CiteVQA: Benchmarking Evidence Attribution for Trustworthy Document Intelligence Paper • 2605.12882 • Published May 13 • 274