Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele
Iṣiro-ara-ẹni ni adanwo Awọn Yipo Aṣoju
Checks how agent reflection uses action history, external feedback, episodic memory and critique to guide another attempt.
Ibeere 1 ti 8
Kini iṣaro ara ẹni ṣe afikun si lupu aṣoju boṣewa?
Awotẹlẹ gbogbo ibeere8 ibeere
Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.
- Kini iṣaro ara ẹni ṣe afikun si lupu aṣoju boṣewa?
- Igbesẹ ti o mọọmọ nibiti awoṣe ṣe atako iṣẹ tirẹ ṣaaju ki o to tẹsiwaju
- Awoṣe yiyara ti o fo awọn ipe irinṣẹ
- Ferese ọrọ ọrọ ti o tobi ju laifọwọyi
- Ayẹwo eniyan fun gbogbo igbesẹ
- Ni awọn ilana Reflexion, nibo ni a ti fipamọ awọn igbero ara ẹni ti awoṣe?
- Ni awọn awoṣe ká itanran-aifwy òṣuwọn
- Ninu ifipamọ iranti episodic ti a lo kọja awọn igbiyanju
- Ni lọtọ database awoṣe ko le ka
- Wọn ti sọnu lẹhin igbesẹ kọọkan
- Awọn ifihan agbara esi wo fun iṣaro duro lati jẹ igbẹkẹle julọ?
- Awọn awoṣe ile ti ara igbekele nikan
- Awọn ipari ti awọn ti tẹlẹ idahun
- Awọn ifihan agbara ita bi awọn abajade idanwo tabi awọn ifiranṣẹ aṣiṣe irinṣẹ
- Akoko ti o gba lati dahun
- Kini ipo ikuna ti a mọ ti iṣaro-ara ẹni?
- Nigbagbogbo o jẹ ki awoṣe lọra ju eniyan lọ
- O ba data ikẹkọ jẹ patapata
- O le ṣee lo fun awọn iṣẹ-ṣiṣe aworan nikan
- Awoṣe naa le ni igboya jẹrisi pe idahun ti ko tọ jẹ deede
- Bawo ni ọna Isọdọtun-ara-ẹni ṣe n ṣe agbejade esi?
- Nipa lilo awoṣe kanna lati ṣofintoto ati lẹhinna atunkọ iṣelọpọ tirẹ
- Nipa bibeere API ile-iṣẹ ti o yatọ
- Nipa bibeere olumulo lati ṣe oṣuwọn ami-ami kọọkan
- Nipa piparẹ gbogbo lilo irinṣẹ
- Kilode ti iṣaro ṣe pataki paapaa fun awọn iṣẹ-ṣiṣe ti ọpọlọpọ-igbesẹ gẹgẹbi ifaminsi tabi lilọ kiri wẹẹbu?
- Nitoripe awọn iṣẹ-ṣiṣe wọnyi kii ṣe awọn aṣiṣe
- Nitoripe awọn igbiyanju ọkan-kọja nigbagbogbo kuna ṣugbọn ibawi-ati-gbiyanju ṣaṣeyọri
- Nitori iṣaro yọ awọn nilo fun eyikeyi irinṣẹ
- Nitoripe o ṣe imukuro iwulo fun ibi-afẹde kan
- In a reflection loop, why is a failing unit test useful evidence for the next attempt?
- It proves the model’s self-critique is correct
- It updates model weights automatically
- It provides an external, concrete signal about which behavior did not meet the requirement
- It guarantees the entire program is wrong
- If an agent stores a critique in memory but does not fine-tune its model, how can that critique affect the next attempt?
- It changes the base model’s parameters permanently
- It changes the programming language
- It disables the external tool
- It can be retrieved and inserted into the next prompt as context