←Search results

Search “Multimodal”

By relevance · Total 22 results Clear search

Thursday

  1. Rohan Paul

    Google Releases EmbeddingGemma 2 Paper: A 744M-Parameter On-Device Multimodal Embedding Model

    Google has released the EmbeddingGemma 2 paper, an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. The model is built on Gemma 4 and tuned through contrastive training by learning from the larger Gemini Embedding 2; with 744M parameters, it surpasses models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal model takes about 567MB; the context window is expanded 4 times to 8K tokens, accommodating about 5.5 minutes of audio, 29 images, or 58 video frames; directly embedding speech scores 69.54 on the MSEB spoken query retrieval benchmark, higher than the 57.59 of transcribing first and then retrieving, and code retrieval on MTEB Code improves from 68.76 in EmbeddingGemma 1 to 78.68.

    Google has released the EmbeddingGemma 2 paper, an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. The model is built on Gemma 4 and tuned through contrastive training by learning from the larger Gemini Embedding 2; with 744M parameters, it surpasses models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal model t

    Matched fieldsTitle、Summary、Body

    #ai-news#Papers/Research#Multimodal#English text is being prepared.

Sunday

Thursday

Tuesday

Friday

  1. Google Research

    Google Research proposes a multi-agent framework that automatically generates long-video narratives

    Google Research has released a unified multi-agent framework for automatically generating temporally coherent long-video narratives, addressing identity drift and cascading failures in existing linear pipelines. Built on top of Gemini and Veo as an orchestration layer, the framework consists of four parts: AI video co-director uses multi-armed bandits for global creative search, CANVAS plans multi-shot visual continuity through persistent visual memory, A²RD generates segment by segment autoregressively while maintaining multimodal video memory, and VQQA uses visual question answering to generate semantic gradients for closed-loop prompt optimization. Evaluations show that AI video co-director reaches a peak quality score of 81.4 on GenAD-Bench, CANVAS improves scene-reappearance continuity on ST-Bench and HardContinuityBench, A²RD reduces layout drift on VBench-Long and LVBench-C, and VQQA achieves quality gains on T2V-CompBench, VBench2, and VBench-I2V.

    A man and woman sit at a table looking at a glowing blue 3D holographic projection of a house emerging from a blueprint.

    Google Research has released a unified multi-agent framework for automatically generating temporally coherent long-video narratives, addressing identity drift and cascading failures in existing linear pipelines. Built on top of Gemini and Veo as an orchestration layer, the framework consists of four parts: AI video co-director uses multi-armed bandits for global creative search, CANVAS plans multi-shot visual continuity through persistent visual memory, A²RD generates segment by segment autoregr

    Matched fieldsSummary、Body

    #lab#Model releases#Multimodal#Video

Friday

Thursday

Wednesday

Friday

  1. Story1 reports · 1 facts

    Google Releases EmbeddingGemma 2 Paper

    Google has released the EmbeddingGemma 2 paper, introducing an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. Built on Gemma 4, the model learns from the larger Gemini Embedding 2 and is tuned with contrastive training; with 744M parameters, it outperforms models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal model takes about 567MB; the context window is expanded 4 times to 8K tokens, accommodating about 5.5 minutes of audio, 29 images, or 58 video frames. Embedding speech directly scores 69.54 on the MSEB spoken query retrieval benchmark, higher than the 57.59 achieved by transcribing first and then retrieving, and code retrieval on MTEB Code improves from 68.76 in EmbeddingGemma 1 to 78.68. There are currently no further developments.

    Google has released the EmbeddingGemma 2 paper, introducing an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. Built on Gemma 4, the model learns from the larger Gemini Embedding 2 and is tuned with contrastive training; with 744M parameters, it outperforms models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal

    Matched fieldsSummary

Wednesday

  1. Phoenix Technology

    MoleculeMind MMDesign Technical Report: 18 of 19 Projects Successful, All 6 GPCR Targets Hit

    MoleculeMind has released a technical report on its AI protein de novo design platform MMDesign, covering 19 projects across 18 targets and two modalities—nanobodies and macrocyclic peptides—with 18 yielding experimentally confirmed binders, a project success rate of 94.7%; of 453 tested candidates, 139 were hits, a candidate hit rate of 30.7%. All six GPCR targets succeeded, with best affinities reaching the nanomolar level: GPR68 achieved a best KD as low as 12.1 nM, and all 30 candidates for NK3R were confirmed to bind. The report also shows that when redesigning against 6 targets that already had binders, 5 achieved improvements, and it presents results for the structure prediction model MMFold on the FoldBench antibody–antigen interface benchmark.

    MoleculeMind has released a technical report on its AI protein de novo design platform MMDesign, covering 19 projects across 18 targets and two modalities—nanobodies and macrocyclic peptides—with 18 yielding experimentally confirmed binders, a project success rate of 94.7%; of 453 tested candidates, 139 were hits, a candidate hit rate of 30.7%. All six GPCR targets succeeded, with best affinities reaching the nanomolar level: GPR68 achieved a best KD as low as 12.1 nM, and all 30 candidates for

    Matched fieldsBody

    #chinese#Papers/Research#English text is being prepared.#Agents
  2. QbitAI

    MoleculeMind MMDesign Technical Report: 18 of 19 Projects Succeeded, All 6 GPCR Targets Hit

    MoleculeMind has released a technical report on its AI de novo protein design platform MMDesign, covering 19 design projects across 18 targets and two modalities—nanobodies and macrocyclic peptides—of which 18 yielded bindable molecules, for a project success rate of 94.7%; among 453 tested candidates, 139 were confirmed to bind, for a candidate hit rate of 30.7%. For all six GPCR targets CCR7, CCR8, EP2, GPR68, NK3R, and A2AR, experimentally confirmed binding molecules were obtained, with per-target hit rates ranging from 12.1% to 100%, and the best binding molecules all achieved nanomolar affinity. The report also shows that MMFold achieved a Top-1 high-quality success rate of 39.0% in 172 interface evaluations of antibody-antigen complex structure prediction in FoldBench, compared with 19.8% for AlphaFold 3.

    MoleculeMind has released a technical report on its AI de novo protein design platform MMDesign, covering 19 design projects across 18 targets and two modalities—nanobodies and macrocyclic peptides—of which 18 yielded bindable molecules, for a project success rate of 94.7%; among 453 tested candidates, 139 were confirmed to bind, for a candidate hit rate of 30.7%. For all six GPCR targets CCR7, CCR8, EP2, GPR68, NK3R, and A2AR, experimentally confirmed binding molecules were obtained, with per-t

    Matched fieldsBody

    #chinese#Papers/Research#English text is being prepared.#Open-source ecosystem

Friday

  1. Microsoft AI Blog

    Microsoft AI releases the MAI-DxO diagnostic system and the SDBench benchmark, with results published in Science

    Microsoft AI has released the AI diagnostic system MAI-DxO and the step-by-step diagnostic benchmark SDBench, with the related results published in Science. On 304 complex NEJM cases, MAI-DxO achieved a diagnostic accuracy of 83%, compared with 36% for doctors with external reference materials and 22% without reference materials; on common emergency cases in MIMIC-IV-ED, MAI-DxO scored 80%, compared with 47% for doctors. The system selects examination items step by step through multi-round debates among multiple virtual doctors, and still achieves 81% accuracy under a cost-optimized configuration at less than half the cost of comparison systems, and it has open-sourced the benchmark and scoring code for the SDBench subset of NEJM cases.

    Microsoft AI has released the AI diagnostic system MAI-DxO and the step-by-step diagnostic benchmark SDBench, with the related results published in Science. On 304 complex NEJM cases, MAI-DxO achieved a diagnostic accuracy of 83%, compared with 36% for doctors with external reference materials and 22% without reference materials; on common emergency cases in MIMIC-IV-ED, MAI-DxO scored 80%, compared with 47% for doctors. The system selects examination items step by step through multi-round debat

    Matched fieldsBody

    #company#Papers/Research#Agents#Evaluations / benchmarks