Gushaka ibimenyetso muri LoCoMo
Run yuzuye ifunguye. Scores zerekana niba evidence ikenewe iri muri memories zagaruwe.
Benchmark ifunguye
Reba amakuru rusange, ibipimo byo gushakisha, igenzura ry'ibisubizo, ibipimo fatizo, imipaka n'amategeko y'ibivugwa inyuma y'ibisubizo byatangajwe.
Run yuzuye ifunguye. Scores zerekana niba evidence ikenewe iri muri memories zagaruwe.
Buri murongo ukoresha imirimo 1,536 yose ifite ibimenyetso n’umupaka umwe wa top-50. Codex yakoze ibisubizo, umusuzumyi wihariye utazi sisitemu arabipima. Iri ni isuzuma ryakozwe na Engram, si urutonde rwemewe rwa LoCoMo.
Ibipimo bibiri byatangajwe bikoresha imirimo 1,536 yose ifite ibimenyetso. Gushaka ibimenyetso bisaba ibiranga inkomoko, kandi buri sisitemu ikoreshwa na model zimwe zo gusubiza no gusuzuma. Sisitemu zituzuza amasezerano amwe y’isuzuma zandikwa ukwazo aho kuvangwa ku rutonde.
ibibazo bya evidence retrieval byapimwe
Engram R@50
answer accuracy ivuye muri top-50 context
Dutandukanya evidence retrieval na answer accuracy. Byombi bikoresha LoCoMo ifunguye, ariko bipima ibice bitandukanye bya memory loop.
AgentMemory igarura ibisubizo bitarenze 100, ni yo mpamvu R@200 igaragazwa nka N/A. Tenure ntibika indangamimerere z’ibimenyetso bya LoCoMo, bityo ntishyirwa ku rutonde rusange; igerageza ryayo rikomeza gusobanurwa mu buryo bw’isuzuma.
Dutandukanya evidence retrieval na answer accuracy. Byombi bikoresha LoCoMo ifunguye, ariko bipima ibice bitandukanye bya memory loop.
Ubufasha n’ibitekerezo
Niba setup iteye urujijo, tool ikora bidasanzwe, cyangwa pricing limit isa n’itari yo, andika hano. Message ihita itugeraho.