Kedis ku xidhan hagaha · Adag Heerka

DeepSeek V3 iyo R1 Kediska Sababta

Tijaabi DeepSeek-V3 iyo R1 qaab dhismeedka MoE, xisaabinta tababarka la soo sheegay, dhuumaha RL, iyo sii daynta kala duwan ee warbixinada asalka ah.

Wadooyinka haga ee la xidhiidhaDeepseek V3 iyo R1 Sababta
Su'aal 1 ee 8

Halbeeg noocee ah ayaa warbixinta farsamada ee DeepSeek-V3 bixisay?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Halbeeg noocee ah ayaa warbixinta farsamada ee DeepSeek-V3 bixisay?
    • 671B total parameters, with 37B activated per token
    • 37B total parameters, all active per token
    • 671B active parameters for every token
    • A dense model with no routed experts
  2. Maxaa xusid mudan tababbarka sababaynta iyo sii deynta DeepSeek-R1?
    • It used no training data
    • The report describes RL-based reasoning and publicly released model weights
    • It was only a closed cloud API
    • It was a rule-based program with no learned model
  3. Waa maxay tirada xisaabinta ay bixisay warbixinta DeepSeek-V3 si ay u hesho tababarkeeda buuxa?
    • 2.788 thousand CPU-hours
    • 2.788 million H100 GPU-hours for R1 only
    • 2.788 million H800 GPU-hours
    • No accelerator-hours were reported
  4. Calaamadee tababar ayaa ka caawisay DeepSeek-R1 horumarinta dabeecadaha sababaynta?
    • Only manually written rules with no learned model
    • Image-label classification
    • Random rewards unrelated to answers
    • Reinforcement learning rewards for correct, verifiable outcomes
  5. Waa maxay hal farsamo oo hufan oo la xidhiidha tababbarka DeepSeek-V3?
    • Feejignaan qarsoon oo madax-badan ah iyo dheellitirka culeyska kaaliyaha- khasaara-la'aanta
    • Ku daabacida qaabka warqad
    • Ka saarida dhammaan lakabyada dareenka
    • Tababar hal tusaale ah
  6. Which types of tasks did the DeepSeek-R1 paper evaluate?
    • Kala soocida sawirka kaliya
    • Mathematics, coding competitions, and STEM tasks
    • Only speech transcription
    • Only spreadsheet formatting
  7. What supervised-learning step does the DeepSeek-R1 pipeline use before its first RL stage?
    • A period of training with no text or prompts
    • A final model-compression step only
    • Cold-start supervised fine-tuning on curated reasoning examples
    • A computer-vision pretraining stage
  8. Which model families were used as bases for the released R1-distilled models?
    • BERT and CLIP
    • Tacotron and WavLM
    • NeRF and Instant-NGP
    • Qwen and Llama