Records found: 6
'VISTA is a multi-agent test-time loop that iteratively refines prompts and selects better videos, improving visual, audio, and contextual alignment and earning strong automatic and human preference gains.'
'OpenTSLM integrates time series as a native modality in pretrained LLMs to interpret ECGs, EEGs, and wearable data, achieving major performance and scalability gains over VLM and text-based approaches'
'M3-Agent is a new multimodal framework that creates entity-centric episodic and semantic long-term memories from continuous video and audio, enabling improved multi-turn reasoning and stateful behavior over time.'
'Alibaba's Ovis2.5 (9B and 2B) advances multimodal AI with a native-resolution vision transformer and an optional thinking mode, achieving top scores for open-source models under 40B and improved OCR and chart understanding.'
'Mistral Medium 3.1 improves multimodal reasoning, coding accuracy, and enterprise deployability while cutting operational costs, making it a practical option for organizations needing powerful, affordable LLMs.'
'VL-Cogito uses a staged curriculum RL and adaptive length rewards to significantly boost multimodal reasoning on math, science and chart benchmarks, outperforming several prior MLLMs.'