Sign In

BuildArena: A Physics-Aligned Interactive Benchmark of LLMs for Engineering Construction

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

์ €์ž

Tian Xia, Tianrun Gao, Wenhao Deng, Long Wei, Xiaowei Qian, Chenglei Yu, Tailin Wu

๐Ÿ’ก ๊ฐœ์š”

๋ณธ ์—ฐ๊ตฌ๋Š” ์ž์—ฐ์–ด ๋ช…์„ธ๋ฅผ ๋ฌผ๋ฆฌ์ ์œผ๋กœ ์‹คํ˜„ ๊ฐ€๋Šฅํ•œ ๊ตฌ์กฐ๋ฌผ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๊ณตํ•™ ๊ฑด์„ค ์ž๋™ํ™”๋ฅผ ์œ„ํ•œ ์ฒซ ๋ฒˆ์งธ ๋ฒค์น˜๋งˆํฌ์ธ BuildArena๋ฅผ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค. BuildArena๋Š” ์ •์  ๋ฐ ๋™์  ์—ญํ•™์„ ํฌํ•จํ•˜๋Š” ๋‹ค์–‘ํ•œ ๋‚œ์ด๋„์˜ ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ์ž‘์—… ์„ค๊ณ„์™€ ์–ธ์–ด ์ง€์‹œ์— ๊ธฐ๋ฐ˜ํ•œ ๊ฑด์„ค์„ ์ง€์›ํ•˜๋Š” 3D ๊ณต๊ฐ„ ๊ธฐํ•˜ํ•™ ๊ณ„์‚ฐ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ํŠน์ง•์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์ตœ์ฒจ๋‹จ LLM์˜ ์–ธ์–ด ๊ธฐ๋ฐ˜ ๋ฐ ๋ฌผ๋ฆฌ์  ์ œ์•ฝ์ด ์žˆ๋Š” ๊ฑด์„ค ์ž๋™ํ™” ๋Šฅ๋ ฅ์„ ํ‰๊ฐ€ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ”‘ ์‹œ์‚ฌ์  ๋ฐ ํ•œ๊ณ„

โ€ข
LLM์˜ ๋ฌผ๋ฆฌ์  ์ œ์•ฝ ์กฐ๊ฑด ํ•˜์—์„œ์˜ ๊ณตํ•™ ๊ฑด์„ค ์ž๋™ํ™” ๋Šฅ๋ ฅ์„ ํ‰๊ฐ€ํ•  ์ˆ˜ ์žˆ๋Š” ์ตœ์ดˆ์˜ ๋ฒค์น˜๋งˆํฌ๋ฅผ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.
โ€ข
์ •์  ๋ฐ ๋™์  ์—ญํ•™์„ ์•„์šฐ๋ฅด๋Š” ๋‹ค์–‘ํ•œ ๋‚œ์ด๋„์˜ ์ž‘์—… ์„ค๊ณ„๋ฅผ ํ†ตํ•ด LLM์˜ ๊ฑด์„ค ๊ด€๋ จ ์ถ”๋ก  ๋Šฅ๋ ฅ์„ ์ฒด๊ณ„์ ์œผ๋กœ ํ‰๊ฐ€ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
โ€ข
์–ธ์–ด ์ง€์‹œ๋ฅผ 3D ๊ณต๊ฐ„์—์„œ์˜ ๊ฑด์„ค๋กœ ์—ฐ๊ฒฐํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ์ œ๊ณตํ•˜์—ฌ ์‹ค์งˆ์ ์ธ ์—”์ง€๋‹ˆ์–ด๋ง ์ ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ํƒ์ƒ‰ํ•ฉ๋‹ˆ๋‹ค.
โ€ข
ํ˜„์žฌ ๋ฒค์น˜๋งˆํฌ๋Š” LLM์˜ ์ดˆ๊ธฐ ๊ฑด์„ค ์ž๋™ํ™” ๋Šฅ๋ ฅ์„ ํ‰๊ฐ€ํ•˜๋Š” ๋ฐ ์ค‘์ ์„ ๋‘๊ณ  ์žˆ์–ด, ๋” ๋ณต์žกํ•˜๊ณ  ํ˜„์‹ค์ ์ธ ๊ฑด์„ค ์‹œ๋‚˜๋ฆฌ์˜ค์— ๋Œ€ํ•œ ์ถ”๊ฐ€์ ์ธ ํ™•์žฅ ๋ฐ ๊ฒ€์ฆ์ด ํ•„์š”ํ•ฉ๋‹ˆ๋‹ค.
๐Ÿ‘