Researchers automated LLM reasoning strategy design and cut token usage by 69.5%

Vercel Security Checkpoint .spinner[data-astro-cid-jlpqxutv]{display:flex;flex-direction:column;justify-content:center;align-items:center}.spinner-circle[data-astro-cid-jlpqxutv]{transform-origin:center;animation:spinner-circle-animate 1s step-start infinite}rect[data-astro-cid-jlpqxutv]{fill:var(--ds-gray-900)}@keyframes spinner-circle-animate{0%{transform:rotate(0)}10%{transform:rotate(36deg)}20%{transform:rotate(72deg)}30%{transform:rotate(108deg)}40%{transform:rotate(144deg)}50%{transform:rotate(180deg)}60%{transform:rotate(216deg)}70%{transform:rotate(252deg)}80%{transform:rotate(288deg)}90%{transform:rotate(324deg)}to{transform:rotate(360deg)}}:root{--ds-gray-900: hsla(0, 0%, 40%, 1);--ds-gray-alpha-1000: hsla(0, 0%, 0%, .91);--background-color: hsla(0, 0%, 100%, 1);--link-color: #0070f3}@media (prefers-color-scheme: dark){:root{--ds-gray-900: hsla(0, 0%, 63%, 1);--ds-gray-alpha-1000: hsla(0, 0%, 100%, .92);--background-color: hsla(0, 0%, 4%, 1);--link-color: #3291ff}}html{box-si