The returned CFO graph from self-evolution reaches 85.0% test survival against the baseline's 0.0% (Fisher's exact test p = 2.6e-8). The best single round observed during the search reached 95.0% test survival (Round 7), but the author's protocol commits the validation-selected checkpoint (85.0% at Round 8/9). · CiteArk
The returned CFO graph from self-evolution reaches 85.0% test survival against the baseline's 0.0% (Fisher's exact test p = 2.6e-8). The best single round observed during the search reached 95.0% test survival (Round 7), but the author's protocol commits the validation-selected checkpoint (85.0% at Round 8/9).
