๊ฐ€์ž… ํ›„ ์ดˆ๋Œ€ ๋งํฌ๋ฅผ ๊ณต์œ ํ•˜๋ฉด ๋™์˜์ƒ ์žฌ์ƒ ๋ฐ ์ดˆ๋Œ€ ๋ณด์ƒ์„ ๋ฐ›์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

Shu Lynn Liu
@shulynnliu
CS PhD Student @BerkeleySky ๐Ÿป Prev @UWMadison ๐Ÿฆก | @mpi_sws_ Fan of @FCBayern #MiaSanMia#
๊ฐ€์ž… May 2020
274 ํŒ”๋กœ์ž‰ ์ค‘    1.8K ํŒฌ
Agents let us build systems for different workloads and requirements. Butโ€ฆ can we trust what they build? We release ๐ŸŒŸSkySynth๐ŸŒŸ: an engine for synthesizing high-performance, just-in-time (JIT) systems we can trust, by co-evolving formal proofs and tests alongside the code. Results: ๐Ÿ’ฟ KV stores up to 2.3ร— faster than Redis and FASTER + formally verified stores with 2.9ร— Claude Code's pass rate ๐Ÿš Model routers up to 48% cheaper than a general router โšก Specialized inference engine with 2.2ร— the throughput of vLLM/SGLang ๐Ÿงต๐Ÿ‘‡
๋” ๋ณด๊ธฐ