[guluchen]

Renames four tier5 stems for lower bounds certified in the GH-217 Tier 5 runs (#217, PR #220). This follows the pattern of #216: same matrices, and only the Gx/Gz header comments change.

Old	New	lb evidence (yfclab2, 6 h, default seed; every level EMPTY on both halves)	Reproduction (relabelled -seed=1, candidate binary)
TN_468_4_17to29	TN_468_4_23to29	main 02879a9 CC default: lb 23 at 19 263 s; GH-217 candidate: 19 518 s	lb 23 at 25 088 s
TN_612_2_17to40	TN_612_2_24to40	main 15 922 s; candidate 15 676 s	lb 24 at 24 335 s
TN_648_14_9to28	TN_648_14_21to28	main 9 661 s; candidate 9 907 s	lb 21 at 12 933 s
TN_684_2_9to29	TN_684_2_22to29	main 12 181 s; candidate 9 035 s (different tree)	lb 22 at 6 279 s

Caveats:

• On TN_468, TN_612 and TN_648_14 the main and candidate runs searched node-identical trees, so the relabelled runs carry the reproduction. Their node counts differ at every high level.
• This is the same CC engine, not an independent second method.
• Upper bounds are unchanged: the [Mac agent] Stern/Dumer collision step in the verified upper-bound finder (large open codes) #146 witnesses are copied under the new stems in optimization/experiments/GH-146/witnesses/renamed-2026-11-05/, and VERIFY.txt reports ALL VALID.

Changes:

• git mv of the Hx/Hz/Gx/Gz files;
• rename-history rows in data/matrices/README.md;
• a dated NOTICE paragraph;
• stem names in optimization/baselines/SERVER.md and main_server_reference.json.

QDistSAT mirror: guluchen/QDistSAT PR (same branch name).

🤖 Generated with Claude Code