Reka AI Multimodal Modeller Testi

Testi çözdüğünüzde cevaplar ve açıklamalar görünür.

1. Which input types did Reka’s 2024 Core, Flash, and Edge technical report describe?
   • Text, images, video, and audio
   • Only text and spreadsheet cells
   • Only still images
   • Audio and source code but no visual input
2. What does Reka say its Vision system can do with video and image content?
   • Compile it into executable code
   • Search, interpret, and reason across the content
   • Guarantee that every event is detected correctly
   • Convert it only into a spreadsheet
3. Which Reka model is described as an open-weight 7B vision-language model for deployment at the edge?
   • Reka Core
   • Reka Vision
   • Reka Edge
   • Reka Flash
4. What did Reka’s 2024 technical report say about how its models could be deployed?
   • They could only run in a public cloud
   • They were limited to a browser extension
   • They could only run on an audio recorder
   • It described deployment flexibility, including on-premises or on-device options
5. What does Reka’s current API model-list endpoint return?
   • The model IDs available to the authenticated user
   • A guaranteed list of every model ever released
   • Training data for each model
   • Only benchmark rankings
6. How should readers interpret Reka’s published model performance comparisons?
   • As guarantees for every device and workload
   • As results tied to the evaluated models, tasks, and test conditions
   • As proof that every model is equally capable
   • As a substitute for checking current model documentation
7. How did Reka’s 2025 visual-language model description handle video frames?
   • It converted the whole video into one audio waveform
   • It processed only the file name and ignored the frames
   • It sampled frames and combined their visual features with text and audio features
   • It replaced all video input with handwritten labels
8. What factors does Reka say affect actual latency and memory use for an Edge deployment?
   • Only the model’s marketing name
   • The color space of the input photo
   • The number of API users in an unrelated account
   • The model, runtime, and device