Amazon Web Services published a walkthrough for running the open-source SkyRL reinforcement learning framework on Amazon SageMaker HyperPod, training a Qwen3-VL-8B vision-language model to navigate vi…
Researchers introduced StreamScout, an adaptive inference framework for streaming video understanding that decides how deeply to access memory per query, escalating through up to three visual views on…
Tencent researchers found that non-thinking inference modes in hybrid multimodal AI models increase response failures by up to 48.64%, according to a new paper on arXiv introducing the PatternEval ben…
Researchers introduce Space Tokens, a lightweight, architecture-agnostic framework that equips vision-language models (VLMs) with explicit continuous spatial representations without additional inferen…
A new arXiv study (2608.00077v1) audits visual-token pruning in OCR-critical multimodal large language models (MLLMs), finding that accuracy alone misses failures where correct answers lack local toke…
Researchers introduce SLAPBench, the first benchmark for evaluating multimodal large language models (MLLMs) on four-finger SLAP fingerprint verification, built from NIST SD302b with 7,832 pairs. Unde…
A new agent evaluation method that grades models against real server states rather than their own reports found that model size does not determine success, with a 27B reasoning-tuned Qwen derivative a…
Researchers introduced the MRCL benchmark to test reinforcement learning's effectiveness in continual post-training for vision-language models, finding that RL alone does not prevent catastrophic forg…
Researchers have developed a lightweight multimodal large language model framework for defect grading of power transmission equipment, achieving state-of-the-art performance while reducing manual anno…