multimodal-ai
57 posts tagged multimodal-ai.
- Fine-tuning 3D generation models is an asset pipeline problem
- HiVe-Mem and the Case for Two-Dimensional Agent Memory
- FlowAlign treats alignment as target-space geometry, not just matching pairs
- What Changed in OpenAI's Images 2.6 for People Who Actually Ship
- Why VLM robot reward models hallucinate progress, and a prompt fix that works
- LARC moves image reasoning out of token space
- What LLM Group Memory Keeps, and Why It Still Gets the Answer Wrong
- FLORE shows why humanoids need force, not just motion
- Speech-only conversation models move the hard part to turn-taking
- Medical exam scores are not enough for clinical LLMs
- NVIDIA’s JCA turns generated video into rig motion
- RefCraft shows web agents still struggle with visual references
- What DeepSeek V4.1 Flash on Dwarfstar tells us about SSD-streaming local models
- Time-series models need two views, not one prettier chart
- PrismaCoder's real trick: a rubric that both filters data and rewards training
- Gemini 4 Argon looks less like a chatbot race and more like a tool-use bet
- Can vision-language models actually spot persuasion in memes?
- Reading 4,000 Comic Books at Once: What the PAGE Framework Actually Does
- ViSkill and the case for keeping agent memory visual
- What SpaceCast-Bench reveals about the gap between seeing space and predicting it
- EmbeddingGemma 2 and the push toward smaller multimodal retrieval
- GPT-6 and Intelligent UI: What OpenAI Actually Announced
- Diffusion models may be sketching meaning before pixels
- MemPilot and the Case for Deciding Memory at Query Time, Not Ingest Time
- Robot agents need indexed video memory, not bigger prompts
- A simulated canvas turns vision models into operators
- MatLoom makes generated materials editable, not just pretty
- Prompt optimization should chase ranking, not accuracy, in imbalanced clinical AI
- Grounded Entity Biographies: A Different Fix for Long-Video Memory
- Audio models do not automatically share speech and text concepts
- The Modality Gap in Speech Fact-Checking, and Why Retrieval Alone Doesn't Fix It
- DiaVLo Turns Vision-Language Model Failures Into Named Behaviours
- Full-Duplex Voice With Tool Calls: What NemotronLabs VoiceChat Actually Ships
- Qwen-Image-2.1 puts open image editing closer to production work
- AI posters get better when the model stops being the designer
- Paint-Anything makes hex colors a first-class diffusion control
- Camera-free smart glasses would test what AI wearables are really for
- MUSE tests vision-language models where classroom context gets messy
- LACE compresses speech tokens one codec layer at a time
- Slip Detection Is Where Robot Hands Stop Dropping Things
- JPEG XL Is a Workflow Question, Not a Format War
- MindTopo Tests Whether Vision Models Grasp Topology, Not Just Distances
- Instruction vs. Example: How Vision-Language Models Actually Moderate Content
- NOAH models patient records as timelines, not snapshots
- What ChatGPT Images 2.5 Changes for People Who Actually Ship Images
- VBVR-Pro makes visual reasoning a training loop, not a demo
- OmniScientist argues that AI scientists need eyes, not just workflows
- AMIE’s video consult result is about perception, not replacement
- Video deep research agents need to look before they search
- MODUS brings any-to-any multimodal modeling to decoder-only systems
- Multimodal AI needs a plan for missing inputs
- MIRROR trains vision models by making each modality teach the others
- Sarcasm detection needs the mismatch, not just the meme
- Visual pretraining is a bet against text extraction
- Claude-real-video points to video as an adapter problem
- Speaker recognition is a better agent test than another chat demo
- Multimodal models still change answers when you shuffle the evidence