Quiz połączony z przewodnikiem · Ciężko Poziom
Quiz dotyczący rozumowania DeepSeek V3 i R1
Przetestuj architekturę MoE DeepSeek-V3 i R1, zgłoszone obliczenia szkoleniowe, potok RL i wydestylowane wersje z oryginalnych raportów.
Pytanie 1 z 8
What parameter counts did the DeepSeek-V3 technical report give?
Podgląd każdego pytania8 pytania
Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.
- What parameter counts did the DeepSeek-V3 technical report give?
- 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
- What was notable about DeepSeek-R1’s reasoning training and release?
- 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
- What compute figure did the DeepSeek-V3 report give for its full training run?
- 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
- What training signal helped DeepSeek-R1 develop reasoning behaviors?
- Only manually written rules with no learned model
- Image-label classification
- Random rewards unrelated to answers
- Reinforcement learning rewards for correct, verifiable outcomes
- Jaka jest jedna technika zwiększająca efektywność powiązana ze szkoleniem DeepSeek-V3?
- Wielogłowicowa ukryta uwaga i pomocnicze równoważenie obciążenia bez strat
- Drukowanie modelu na papierze
- Usunięcie wszystkich warstw uwagi
- Szkolenie na jednym przykładzie
- Which types of tasks did the DeepSeek-R1 paper evaluate?
- Tylko klasyfikacja obrazu
- Mathematics, coding competitions, and STEM tasks
- Only speech transcription
- Only spreadsheet formatting
- 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
- 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