mirror of
https://github.com/lightningnetwork/lnd.git
synced 2026-08-18 13:07:58 +02:00
cda40a3dc0
In this commit, we log the code_split2 result, the first verdict backed by paired statistics. The mechanism the corpus was built to elicit did emerge: the winner derives unequal split candidates from estimated corridor sizes and plans route sets with one-step lookahead, reserving each candidate route while scoring the next shard against the remainder -- structurally past reactive laddering, and the first time the speculative in-flight reservation machinery became load bearing. And yet it loses to mx_c3 on every tier, significantly so (paired deltas -0.025 to -0.086, sign-test p <= 0.04), including the splitting corpus itself. The pre-registered resolution caveat stands: near-binary per-file scores gave selection little gradient to polish the new machinery, and the same pattern as exp-008 repeats -- the environment elicits the mechanism, but a 400-eval lineage cannot out-refine a 900-eval champion. The paired statistics also settle an informal claim: hb1 and mx_c3 are indistinguishable on mainnet. Champions of record remain hb1 and mx_c3; the Opus 5 reflection arm is still running. |
||
|---|---|---|
| .. | ||
| exp-001-param-smoke.md | ||
| exp-002-param-run1.md | ||
| exp-003-seed-router-vs-lnd.md | ||
| exp-004-code-evolution.md | ||
| exp-004-sim-audit.md | ||
| exp-005-sim-audit.md | ||
| exp-006-breakthrough.md | ||
| exp-007-mix-followup.md | ||
| exp-008-drift-evolution.md | ||
| exp-008-drift1-best-candidate.go | ||
| exp-008-drift1-best-candidate.md | ||
| exp-009-mainnet-validation.md | ||
| exp-010-split2-best-candidate.go | ||
| exp-010-splitting-pressure.md | ||
| exp-011-code-gen2.md | ||
| exp-011-gen2-best-candidate.go | ||