\n<\/p>\n
Lengthy-horizon execution in Massive Language Fashions (LLMs) stays unstable even when high-level methods are supplied. Evaluating on managed algorithmic puzzles, we reveal that whereas decomposition is crucial for stability, excessive decomposition creates a \u201cno-recovery bottleneck\u201d. We present that this bottleneck turns into essential on account of extremely non-uniform error distribution, the place constant errors on a couple of \u201conerous\u201d steps grow to be irreversible. To handle this, we suggest Lookahead-Enhanced Atomic Decomposition (LEAD). By incorporating short-horizon future validation and aggregating overlapping rollouts, LEAD gives sufficient isolation to keep up stability whereas retaining sufficient native context to appropriate errors. This permits the o4-mini mannequin to unravel Checkers Leaping as much as complexity n = 13, whereas excessive decomposition fails past n = 11. <\/p>\n<\/div>\n\n","protected":false},"excerpt":{"rendered":"
Lengthy-horizon execution in Massive Language Fashions (LLMs) stays unstable even when high-level methods are supplied. Evaluating on managed algorithmic puzzles, we reveal that whereas decomposition is crucial for stability, excessive decomposition creates a \u201cno-recovery bottleneck\u201d. We present that this bottleneck turns into essential on account of extremely non-uniform error distribution, the place constant errors on […]<\/p>\n","protected":false},"author":2,"featured_media":17114,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[55],"tags":[7876,5033,1338,9936,9935,616],"class_list":["post-17112","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-machine-learning","tag-bottleneck","tag-breaking","tag-lead","tag-longhorizon","tag-norecovery","tag-reasoning"],"_links":{"self":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/17112","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=17112"}],"version-history":[{"count":1,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/17112\/revisions"}],"predecessor-version":[{"id":17113,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/posts\/17112\/revisions\/17113"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=\/wp\/v2\/media\/17114"}],"wp:attachment":[{"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=17112"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=17112"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/techtrendfeed.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=17112"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}