Sign In

ShopGym: An Integrated Framework for Realistic Simulation and Scalable Benchmarking of E-Commerce Web Agents

์ž‘์„ฑ์ž
  • Haebom
์นดํ…Œ๊ณ ๋ฆฌ
Empty

์ €์ž

Chinmay Savadikar, Mingyu Zhao, Yuanzheng Zhu, Han Li, Shuang Xie, Alberto Castelo, Tianfu Wu, Lingyun Wang

๐Ÿ’ก ๊ฐœ์š”

๋ณธ ๋…ผ๋ฌธ์€ ์ „์ž์ƒ๊ฑฐ๋ž˜ ์›น ์—์ด์ „ํŠธ ๊ฐœ๋ฐœ ๋ฐ ํ‰๊ฐ€๋ฅผ ์œ„ํ•œ ํ˜„์‹ค์ ์ด๊ณ  ์ œ์–ด ๊ฐ€๋Šฅํ•˜๋ฉฐ ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ํ™˜๊ฒฝ์˜ ๋ถ€์žฌ๋ฅผ ์ง€์ ํ•ฉ๋‹ˆ๋‹ค. ์ด๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด, ShopGym์ด๋ผ๋Š” ํ†ตํ•ฉ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์•ˆํ•˜๋ฉฐ, ์ด๋Š” ์‹ค์ œ ์Šคํ† ์–ดํ”„๋ก ํŠธ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๋‹ค์–‘ํ•˜๊ณ  ์ œ์–ด ๊ฐ€๋Šฅํ•œ ์ƒŒ๋“œ๋ฐ•์Šค ํ™˜๊ฒฝ์„ ์ƒ์„ฑํ•˜๊ณ , ์ด๋ฅผ ํ™œ์šฉํ•˜์—ฌ ์—์ด์ „ํŠธ ํ‰๊ฐ€๋ฅผ ์œ„ํ•œ ๊ตฌ์กฐํ™”๋œ ๋ฒค์น˜๋งˆํฌ ์ž‘์—…์„ ํ•ฉ์„ฑํ•ฉ๋‹ˆ๋‹ค. ShopGym์€ ์‹ค์ œ ์›น์‚ฌ์ดํŠธ์˜ ๊ตฌ์กฐ์  ํŠน์„ฑ์„ ์œ ์ง€ํ•˜๋ฉด์„œ ์žฌํ˜„ ๊ฐ€๋Šฅํ•˜๊ณ  ๊ฒ€์‚ฌ ๊ฐ€๋Šฅํ•œ ํ‰๊ฐ€ ํ™˜๊ฒฝ์„ ์ œ๊ณตํ•˜์—ฌ ๊ณผํ•™์  ๋น„๊ต๋ฅผ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ”‘ ์‹œ์‚ฌ์  ๋ฐ ํ•œ๊ณ„

โ€ข
์‹ค์ œ ์ „์ž์ƒ๊ฑฐ๋ž˜ ํ™˜๊ฒฝ์˜ ๋ณต์žก์„ฑ๊ณผ ๋‹ค์–‘์„ฑ์„ ์œ ์ง€ํ•˜๋ฉด์„œ๋„ ์ œ์–ด ๊ฐ€๋Šฅํ•˜๊ณ  ์žฌํ˜„ ๊ฐ€๋Šฅํ•œ ๋ฒค์น˜๋งˆํ‚น ํ™˜๊ฒฝ์„ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๋Š” ๋ฐฉ๋ฒ•๋ก ์„ ์ œ์‹œํ•ฉ๋‹ˆ๋‹ค.
โ€ข
ShopGym ํ”„๋ ˆ์ž„์›Œํฌ๋Š” ์‹ค์ œ ์Šคํ† ์–ดํ”„๋ก ํŠธ์˜ ํŠน์ง•์„ ๋ณด์กดํ•˜๋ฉด์„œ๋„ ์—์ด์ „ํŠธ ํ‰๊ฐ€์— ์œ ์šฉํ•œ ๋‹ค์–‘ํ•œ ๋ฒค์น˜๋งˆํฌ ์ž‘์—…์„ ์ž๋™์œผ๋กœ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ์Œ์„ ๋ณด์—ฌ์ค๋‹ˆ๋‹ค.
โ€ข
์ œ์•ˆ๋œ ์ƒŒ๋“œ๋ฐ•์Šค ํ™˜๊ฒฝ์—์„œ์˜ ์—์ด์ „ํŠธ ์„ฑ๋Šฅ์ด ์‹ค์ œ ์Šคํ† ์–ดํ”„๋ก ํŠธ์—์„œ์˜ ์„ฑ๋Šฅ๊ณผ ์–‘์˜ ์ƒ๊ด€๊ด€๊ณ„๋ฅผ ๊ฐ€์ง์„ ์ž…์ฆํ•˜์—ฌ, ShopGym์˜ ์œ ํšจ์„ฑ์„ ํ™•์ธํ•ฉ๋‹ˆ๋‹ค.
โ€ข
ShopGym์€ ๋‹ค์–‘ํ•œ ์ „์ž์ƒ๊ฑฐ๋ž˜ ์‹œ๋‚˜๋ฆฌ์˜ค์— ๋Œ€ํ•œ ์—์ด์ „ํŠธ์˜ ๋Šฅ๋ ฅ ๋ฐ ํ–‰๋™์„ ์ฒด๊ณ„์ ์œผ๋กœ ํ‰๊ฐ€ํ•  ์ˆ˜ ์žˆ๋Š” ๊ธฐ๋ฐ˜์„ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.
โ€ข
(ํ•œ๊ณ„์  ๋˜๋Š” ํ–ฅํ›„ ๊ณผ์ œ) ์ƒ์„ฑ๋œ ์ƒŒ๋“œ๋ฐ•์Šค ํ™˜๊ฒฝ์ด ๋ชจ๋“  ์‹ค์ œ ์ „์ž์ƒ๊ฑฐ๋ž˜ ํ™˜๊ฒฝ์˜ ๋ฏธ๋ฌ˜ํ•œ ์ฐจ์ด์ ์„ ์™„๋ฒฝํ•˜๊ฒŒ ํฌ์ฐฉํ•˜์ง€ ๋ชปํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ํ•ฉ์„ฑ๋œ ๋ฐ์ดํ„ฐ์˜ ํ˜„์‹ค๊ฐ ๋ฐ ๋‹ค์–‘์„ฑ์„ ์ง€์†์ ์œผ๋กœ ํ–ฅ์ƒ์‹œํ‚ค๊ธฐ ์œ„ํ•œ ์—ฐ๊ตฌ๊ฐ€ ํ•„์š”ํ•ฉ๋‹ˆ๋‹ค. ๋˜ํ•œ, ๋‹ค์–‘ํ•œ ์—์ด์ „ํŠธ ์•„ํ‚คํ…์ฒ˜ ๋ฐ ํ•™์Šต ๋ฐฉ๋ฒ•๋ก ์— ๋Œ€ํ•œ ShopGym์˜ ์ ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ์ถ”๊ฐ€์ ์œผ๋กœ ๊ฒ€์ฆํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
๐Ÿ‘