Tag: #multimodal-llm14 results
DeepSeek releases GitHub technical report on multimodal model with 'Thinking with Visual Primitives' framework, turning points/bboxes into core reasoning units to solve spatial referencing issues. Overcomes language ambiguity in complex layouts, enabling precise spatial inference. Compact model matches GPT-4o, Claude-Sonnet, Gemini on tough benchmarks.
DeepMind's Hassabis wins 2024 Nobel Chemistry for AlphaFold protein prediction. Gemini 3 tops multimodal benchmarks in Nov 2025; AlphaGenome models 98% genome at 90% accuracy in Jan 2026. Predicts 50% AGI chance by 2030 requiring hypothesis skills.
A trillion-parameter domestic multimodal large model open-sourced as top partner for enterprise OpenClaw. Marks major leap in Chinese AI multimodal tech. Now available for integration.
DeepSeek's image recognition mode enters gray release, with hands-on tests revealing it's likely a new model. Non-thinking mode delivers blazing speed. Users report access via beta rollout.
Huawei launched HarmonySpace 6 featuring Xiao Yi intelligent agent, industry's first full-scene chatting AI for navigation, control, and services while driving. Upgrades include AMS multimodal perception for vital signs and safety, MoLA 2.0 with 100B+ param multimodal model and Agent platform. New hardware like dual 17.2-inch screens, million-pixel lights, and laser projector enhance entertainment.
New-generation Xiaomi SU7 launches fully equipped with Xiaomi XLA cognitive large model for advanced assisted driving, featuring top-tier hardware like 700TOPS Thor chip and multiple sensors. XLA enables multimodal fusion including voice-controlled driving and parking. OTA upgrades will roll out to first-gen SU7 Pro/Max/Ultra and YU7 models.
China's first embodied intelligent engineering robot industry standard compilation started in Chengdu on March 24. It aims to fill industry gaps in evaluation, safety, and scaling. Chengdu Humanoid Robot Innovation Center released the world's first multimodal LLM for these robots.
RideJudge introduces a progressive framework aligning visual and logical reasoning for ride-hailing dispute adjudication, addressing limitations in multimodal LLMs. Key innovations include SynTraj for trajectory synthesis, Adaptive Context Optimization, and Ordinal-Sensitive RL. The 8B model achieves 88.41% accuracy, outperforming 32B baselines.
Chinese AI startup StepFun is eyeing a Hong Kong IPO to raise about $500M, following Zhipu AI and MiniMax. It focuses on AI+terminal strategy with multimodal models integrated into OPPO, Honor phones and Geely cars. Recent B+ round raised over 5B RMB from top investors.
PlotChain introduces a deterministic benchmark for evaluating multimodal LLMs on extracting quantitative values from engineering plots like Bode and FFT. It features 450 plots across 15 families with ground truth and checkpoint diagnostics for failure analysis. Top models score ~80% (Gemini 2.5 Pro leads), but frequency tasks remain weak.