Idanwo Ilana Aṣoju ReAct

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

1. Awọn agbara meji wo ni ilana ReAct darapọ?
   • Idi (ero) ati ṣiṣe (lilo awọn irinṣẹ)
   • Aworan iran ati awọn iwe ohun kolaginni
   • Ìsekóòdù ati funmorawon
   • Titẹ sita ati Antivirus
2. Kini awọn eroja atunwi mẹta ti lupu ReAct?
   • Iṣagbewọle, Ijade, Tun bẹrẹ
   • Ero, Ise, akiyesi
   • Encode, Yiyipada, Ifihan
   • Wọle, Jade, Sọtun
3. Kini 'Akiyesi' ni loop ReAct?
   • Awọn awoṣe ká ti abẹnu ètò
   • Ọrọigbaniwọle olumulo kan
   • Abajade naa pada lẹhin ṣiṣe iṣe kan, bii abajade wiwa
   • A ik Lakotan imeeli
4. Kini idi ti ReAct le dinku hallucination ti a fiwera pẹlu ero-ero mimọ?
   • Ko jẹ ki awoṣe idi ni gbogbo
   • O paarẹ data ikẹkọ awoṣe
   • O ṣiṣẹ offline nikan
   • O da ero ni awọn akiyesi gidi lati awọn irinṣẹ ati data ita
5. What improvement did the ReAct paper report over imitation/RL baselines on ALFWorld?
   • An absolute success-rate improvement of 34 percentage points in its evaluated setup
   • A 100% success rate across all tasks
   • A 34% reduction in tool calls on every benchmark
   • Ko si iyipada iwọnwọn
6. Which external source did the ReAct paper use to retrieve observations in its knowledge-intensive QA experiments?
   • A private medical database
   • A Wikipedia API
   • A spreadsheet stored in the model weights
   • An offline dictionary only
7. How can an observation change the next step in a ReAct trajectory?
   • It is always ignored after an action
   • It guarantees the next action is correct
   • The model can revise its plan based on the tool or environment result
   • It automatically becomes the final answer
8. Which task types did the original ReAct paper evaluate?
   • Only image compression
   • Only speech recognition
   • Only spreadsheet formatting
   • Knowledge-intensive question answering and interactive decision-making