Click a node or edge for details

Verification — outgoing probability per source state (sums to 100%)

sourceout totalΣ% → targets
start90100%breakthrough 86.7% (78)   improved 11.1% (10)   stuck 2.2% (2)
improved12100%breakthrough 58.3% (7)   end_limit 16.7% (2)   end_submit 8.3% (1)   improved 8.3% (1)   regressed 8.3% (1)
stuck4100%stuck 50.0% (2)   breakthrough 25.0% (1)   improved 25.0% (1)
breakthrough86100%end_submit 97.7% (84)   stay 2.3% (2)
stay5100%stay 60.0% (3)   end_submit 40.0% (2)
regressed1100%end_limit 100.0% (1)

Gap label distribution — for each label, which src→tgt edges carry it

gap labeltotal src→tgt breakdown
debug_failure2regressed→end_limit 50.0% (1)   improved→regressed 50.0% (1)
explore2start→stuck 50.0% (1)   improved→end_limit 50.0% (1)
fix_bug13improved→breakthrough 30.8% (4)   stay→stay 23.1% (3)   stuck→stuck 15.4% (2)   breakthrough→stay 15.4% (2)   stuck→breakthrough 7.7% (1)   improved→improved 7.7% (1)
implement89start→breakthrough 87.6% (78)   start→improved 11.2% (10)   start→stuck 1.1% (1)
rewrite4improved→breakthrough 75.0% (3)   stuck→improved 25.0% (1)
verify88breakthrough→end_submit 95.5% (84)   stay→end_submit 2.3% (2)   improved→end_submit 1.1% (1)   improved→end_limit 1.1% (1)