Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Ìdánwò Ìdánwò ati Atunse Awọn Aṣoju Ara-ẹni

Review Reflexion’s verbal feedback loop, evaluator, episodic memory, benchmark scope, and risks of unreliable feedback. Check evidence and limits.

Jẹmọ itọsọna awọn ọnaReflexion Ati Awọn aṣoju Atunse Ara-ẹni
Ibeere 1 ti 8

Kini abuda asọye ti bii aṣoju Reflexion kan ṣe 'kọ ẹkọ'?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Kini abuda asọye ti bii aṣoju Reflexion kan ṣe 'kọ ẹkọ'?
    • O tọju awọn iweyinpada kikọ ni aaye laisi iyipada awọn iwuwo awoṣe
    • O ṣe imudojuiwọn awọn iwuwo awoṣe nipasẹ iṣatunṣe itanran
    • O ṣe igbasilẹ ẹya awoṣe tuntun lẹhin ikuna kọọkan
    • O beere lọwọ eniyan lati tunkọ data ikẹkọ rẹ
  2. Ninu ilana Isọdọtun, kini Oluyẹwo ṣe?
    • O ṣe ipilẹṣẹ igbese atẹle ti aṣoju
    • O ṣe ikun abajade ti igbiyanju lati gbejade ifihan agbara esi kan
    • O npa awọn iranti atijọ lati fi aaye pamọ
    • O tumọ iṣẹ-ṣiṣe si ede miiran
  3. Kilode ti didara ifihan agbara igbelewọn ṣe pataki pupọ ninu Ipadabọ?
    • Nitoripe o ṣe ipinnu oṣuwọn ikẹkọ awoṣe
    • Nitoripe o pinnu iye GPU ti a lo
    • Nitoripe ifihan agbara ti ko tọ le jẹ ki aṣoju naa ronu lori ati fikun awọn ẹkọ ti ko tọ
    • Nitoripe o ṣeto iwọn otutu ti awoṣe ede
  4. On which coding benchmark did the Reflexion paper report a 91% pass@1 result?
    • Pipa Net
    • COCO detection
    • WMT translation
    • Ìdánilẹ́kọ̀ọ́ Ọmọnìyàn
  5. Ewo ni eewu ṣiṣi otitọ pẹlu awọn aṣoju ti n ṣatunṣe ara ẹni?
    • Wọn le fajade lainidii 'titunṣe' ti o ti pe tẹlẹ
    • Wọn nilo data ike fun gbogbo aami
    • Wọn ko le ṣiṣe awọn lori boṣewa hardware
    • Wọn ba awoṣe ipilẹ jẹ patapata
  6. Where does Reflexion keep feedback for a later attempt?
    • In updated model weights
    • In an episodic memory buffer supplied as text context
    • In a hardware cache shared across users
    • In the original training corpus
  7. What is the role of the Actor in Reflexion?
    • It changes the base model weights
    • It authenticates citations automatically
    • It generates the actions or attempt for the task
    • It sets the benchmark score after the run
  8. What kinds of feedback can Reflexion use?
    • Only gradient updates from fine-tuning
    • Only human ratings in every task
    • Only token likelihood from pretraining
    • Signals such as scalar outcomes or language feedback from external or simulated sources