[guluchen]

PI-approved stem renames (2026-10-31) for eight tier5 codes whose CC lower bounds were raised by GH-157 (#157), GH-162 (#162) / GH-183 (#183) and GH-181 (#181). Do not merge before the PI's confirmation.

Old	New	lb	ub	Evidence
TN_360_4_17to24	TN_360_4_23to24	17 -> 23	24 (unchanged)	GH-157 (#157, PR #161)
TN_396_4_17to26	TN_396_4_23to26	17 -> 23	26 (unchanged)	GH-157 (#157, PR #161)
TN_504_2_17to34	TN_504_2_24to34	17 -> 24	34 (unchanged)	GH-157 (#157, PR #161)
TN_648_10_9to32	TN_648_10_22to32	9 -> 22	32 (unchanged)	GH-157 (#157, PR #161)
BB_1080_4_30to48	BB_1080_4_32to48	30 -> 32	48 (unchanged)	GH-162 (#162, PR #180); reproduced by GH-183 (#183, PR #204)
BB_756_16_30to34	BB_756_16_32to34	30 -> 32	34 (unchanged)	GH-183 (#183, PR #204)
BB_864_4_30to40	BB_864_4_32to40	30 -> 32	40 (unchanged)	GH-183 (#183, PR #204)
TN_496_2_23to30	TN_496_2_27to30	23 -> 27	30 (unchanged)	GH-181 (#181, PR #203)

BB_756 and BB_864: the current stems on main were BB_756_16_30to34 and BB_864_4_30to40 (old lb 30 < 32), and the #183 cross-check comment confirms >= 32 for both, so neither was skipped.

Evidence.

• [Mac agent] CC engine: fractional-charge / edge-cover node lower bounds #157 (TN_360/396/504/648_10): CC engine with the charge rule, 6 h, binary c3c315d. Reproduction: same binary c3c315d and labels, different search order (-cc-rules=excl,pack,orbf,parity,leaf,overlap,charge,mrv, 12 h): all four reproduced, every level EMPTY. Record: optimization/experiments/GH-157/REPRODUCTION.md (on main since CC engine: charge node lower bound as a default CC rule (GH-157) #161).
• [Mac agent] CC engine: orbit-compatible even-weight overlap tie-break #162 / [Mac agent] CC engine: MRV branching and BFS row order on top of tiebreak #183 (BB_1080): lb 32 found by tiebreak seed 0 in 19 378 s (binary 97c8629). Reproduced by two [Mac agent] CC engine: MRV branching and BFS row order on top of tiebreak #183 Tier 5 runs with a different search order (tiebreak,mrv, -cc-roworder=bfs, with and without -cc-tbbase=asc; binary d26070d) at 7 127 s and 6 739 s; every level EMPTY, node counts differ at every level from 6 on. Record: [Mac agent] CC engine: orbit-compatible even-weight overlap tie-break #162 comments, PR CC engine: orbit-compatible even-weight overlap tie-break (GH-162, opt-in; do not merge before PI says merge) #180 section 7.
• [Mac agent] CC engine: MRV branching and BFS row order on top of tiebreak #183 (BB_756, BB_864): lb 32 in the 6 h Tier 5 run (10 437 s / 9 951 s); reproduced 9/9 with relabeled seeds 1-3, 8 h (BB_756, BB_864 and TN_496 lb 24), every level EMPTY. Record: [Mac agent] CC engine: MRV branching and BFS row order on top of tiebreak #183 comment and optimization/experiments/GH-183/S3_xcheck/ on experiment/gh-183-cc-roworder.
• [Mac agent] CC engine: redundant sparse checks and XOR unit propagation #181 (TN_496): lb 27 from sparse,sparseov,sparsemrv, seeds 0/1/2 (6 h). Cross-checked by a different binary (sparse,sparseov,sparsemrv,tiebreak scratch merge), seeds 0 and 3, 8 h: lb 27, every level 1-26 EMPTY. Record: optimization/experiments/GH-181/XC27/ (commit f0afded, PR CC engine: redundant sparse checks (sparse, sparseov, sparsemrv) and XOR unit propagation (GH-181, opt-in; do not merge before PI says merge) #203 section 7a).
• Caveat: each lb comes from DistQLDPC's own CC engine and was reproduced with a different search order or binary. That is not an independent second method. Each rule was opt-in when its bound was found. charge became the default CC rule on main when CC engine: charge node lower bound as a default CC rule (GH-157) #161 was merged on 2026-10-30. tiebreak, -cc-roworder=bfs and the sparse* rules are still opt-in and not on main.

Changes (pattern of #163):

• git mv of Hx/Hz/Gx/Gz for each stem; only the Gx/Gz header comment changes.
• data/matrices/README.md: the tier5 list; eight rename-history rows citing the issue, PR, rule, run and reproduction; open items (TN_648_10, the witness copies, a 2026-10-31 lower-bound caveat).
• NOTICE: a 2026-10-31 rename paragraph.
• README.md: the naming example (TN_648_10_22to32, BB_756_16_32to34), as Adopt GH-146 Stern finder as default (both engines); tier5 stems get verified upper bounds #149 did.
• optimization/baselines/SERVER.md and main_server_reference.json: name columns (as Rename six tier5 TN stems with GH-152 CC lower bounds #163 did).
• The upper-bound witnesses for these eight stems are copied under the new stems to optimization/experiments/GH-146/witnesses/renamed-2026-10-31/: seven from GH-146/witnesses/renamed/ and TN_496 from GH-152/witnesses/. VERIFY.txt shows 8/8 VALID with the repo's pure-Python verify_witnesses.py: Hx Hz^T = 0, H_other e = 0, e not in rowspan(H_same), odd overlap with G, and weight = ub. No external solver was used. Older copies are left as they were (history).
• Old stems remain only in rename-history entries (data/matrices/README.md, NOTICE) and in historical experiment records under optimization/experiments/ ([Mac agent] Stern/Dumer collision step in the verified upper-bound finder (large open codes) #146, [Mac agent] MaxSAT path: stabilizer-overlap dominance constraints in every CSS-half probe #148, [Mac agent] CC engine: Tanner-graph automorphism discovery for orbit seeds and the XZ-dual map (open TN codes) #152, [Mac agent] CC engine: fractional-charge / edge-cover node lower bounds #157), which are unchanged. No tests, scripts or web files referenced these stems.
• Rebased onto main c94d390b (CC engine: charge node lower bound as a default CC rule (GH-157) #161 merged).

Checks (Mac, no timing runs, final tree):

• Build: OK.
• make test-cc-flags: 442 checks, 0 failed.
• scripts/smoke_test.sh: passed. scripts/test_partition_soft_literals.py: passed.
• Tier 0 answers: 13 codes x {maxsat, cc}, 26/26 match the stem distance.
• Tier table vs disk: 51 stems listed = 51 on disk, all four files present.
• Bare stems resolve: all eight new stems load from data/matrices/tier5/ under both engines (-cpu-lim=3 -v), 16/16. The old stems give cannot open.

Open PRs that still mention old stems (not touched here):

• CC engine: orbit-compatible even-weight overlap tie-break (GH-162, opt-in; do not merge before PI says merge) #180, CC engine: MRV branching and seed-anchored reversed-BFS row order (GH-183, opt-in; do not merge before PI says merge) #204, CC engine: per-half probe between tiebreak and MRV + reversed-BFS row order (-cc-roworder=auto, GH-206; do not merge before PI says merge) #215: BB_1080_4_30to48, BB_756_16_30to34, BB_864_4_30to40 and TN_496_2_23to30, in the body and diff.
• CC engine: redundant sparse checks (sparse, sparseov, sparsemrv) and XOR unit propagation (GH-181, opt-in; do not merge before PI says merge) #203: TN_496_2_23to30 in the body. Its diff mentions all eight old stems.
• CC engine: charge node lower bound as a default CC rule (GH-157) #161 is already merged; its [Mac agent] CC engine: fractional-charge / edge-cover node lower bounds #157 records on main keep the old stems as history.
• CC: erasure bound (opt-in token) #195 is an issue, not a PR, and mentions none of these stems.

QDistSAT mirror: guluchen/QDistSAT#15

🤖 Generated with Claude Code