summaryrefslogtreecommitdiff
path: root/artifacts/synth_v1/match_battery.json
diff options
context:
space:
mode:
authorYurenHao0426 <Blackhao0426@gmail.com>2026-08-01 18:53:29 -0500
committerYurenHao0426 <Blackhao0426@gmail.com>2026-08-01 18:53:29 -0500
commitde827a42e10ede662f4bd2893c4f8b6d54be45dc (patch)
treed7c2c30f5e36102572cacc70e4296d80459d992f /artifacts/synth_v1/match_battery.json
parent4f7ee05cc3b072478062e53645af016861c4b529 (diff)
Amplification beats the instance that defeated fifteen solvers
The caption-omitted field returned under 5% from every solver tried: Umeyama, GRAMPA, five Gromov-Wasserstein variants, FAQ with and without restarts, PATH convex-concave, a moment ladder, semirelaxed GW. The measurements said the answer was not a sixteenth solver. Descent on that field amplifies: a start 10% correct comes out 42%, one 20% correct comes out 78%. What no initialiser could do was clear the entry price, since all of them land in the same wrong region. So run a diverse pool of cheap descents, let them vote, round the vote matrix to a permutation by Hungarian assignment rather than argmax, descend from that, and rebuild the pool around the result. Each round feeds the amplifier a better start. 0.044 -> 0.72 on the best run, 0.54 mean over two. Competitive elsewhere: 0.977 on the easy field against 0.961 for the best GW variant. Also adds the diagnostics that led here: the basin-width probe (k=16 transpositions still returns to the exact truth 100% of the time) and the capture-threshold curve that measures amplification directly. Co-Authored-By: Claude <noreply@anthropic.com>
Diffstat (limited to 'artifacts/synth_v1/match_battery.json')
-rw-r--r--artifacts/synth_v1/match_battery.json165
1 files changed, 165 insertions, 0 deletions
diff --git a/artifacts/synth_v1/match_battery.json b/artifacts/synth_v1/match_battery.json
new file mode 100644
index 0000000..7e65f80
--- /dev/null
+++ b/artifacts/synth_v1/match_battery.json
@@ -0,0 +1,165 @@
+{
+ "protocol": "Blind: the hidden permutation is generated per trial and read only for scoring. Each solver is scored as returned and again after exact steepest-descent refinement.",
+ "rows": [
+ {
+ "field": "omit-size-BOUND0.997",
+ "method": "grampa (current)",
+ "raw": 0.016927083333333332,
+ "refined": 0.044270833333333336,
+ "seconds": 0.021717230478922527,
+ "failures": []
+ },
+ {
+ "field": "omit-size-BOUND0.997",
+ "method": "umeyama",
+ "raw": 0.005208333333333333,
+ "refined": 0.014322916666666666,
+ "seconds": 0.01471241315205892,
+ "failures": []
+ },
+ {
+ "field": "omit-size-BOUND0.997",
+ "method": "gromov-wasserstein",
+ "raw": 0.018229166666666668,
+ "refined": 0.022135416666666668,
+ "seconds": 0.08475653330485027,
+ "failures": []
+ },
+ {
+ "field": "omit-size-BOUND0.997",
+ "method": "entropic GW annealed",
+ "raw": 0.0078125,
+ "refined": 0.033854166666666664,
+ "seconds": 27.067088921864826,
+ "failures": []
+ },
+ {
+ "field": "omit-size-BOUND0.997",
+ "method": "BAPG GW",
+ "raw": 0.013020833333333334,
+ "refined": 0.01171875,
+ "seconds": 1.8669984340667725,
+ "failures": []
+ },
+ {
+ "field": "rank8-BOUND0.93",
+ "method": "grampa (current)",
+ "raw": 0.045572916666666664,
+ "refined": 0.07552083333333333,
+ "seconds": 0.01863869031270345,
+ "failures": []
+ },
+ {
+ "field": "rank8-BOUND0.93",
+ "method": "umeyama",
+ "raw": 0.01953125,
+ "refined": 0.015625,
+ "seconds": 0.0132447878519694,
+ "failures": []
+ },
+ {
+ "field": "rank8-BOUND0.93",
+ "method": "gromov-wasserstein",
+ "raw": 0.82421875,
+ "refined": 0.8815104166666666,
+ "seconds": 0.04134281476338705,
+ "failures": []
+ },
+ {
+ "field": "rank8-BOUND0.93",
+ "method": "entropic GW annealed",
+ "raw": 0.88671875,
+ "refined": 0.88671875,
+ "seconds": 18.96754988034566,
+ "failures": []
+ },
+ {
+ "field": "rank8-BOUND0.93",
+ "method": "BAPG GW",
+ "raw": 0.10546875,
+ "refined": 0.23307291666666666,
+ "seconds": 1.8909683227539062,
+ "failures": []
+ },
+ {
+ "field": "synth-full-BOUND0.99",
+ "method": "grampa (current)",
+ "raw": 0.5377604166666666,
+ "refined": 0.8841145833333334,
+ "seconds": 0.01822185516357422,
+ "failures": []
+ },
+ {
+ "field": "synth-full-BOUND0.99",
+ "method": "umeyama",
+ "raw": 0.0234375,
+ "refined": 0.5403645833333334,
+ "seconds": 0.014760414759318033,
+ "failures": []
+ },
+ {
+ "field": "synth-full-BOUND0.99",
+ "method": "gromov-wasserstein",
+ "raw": 0.3697916666666667,
+ "refined": 0.3971354166666667,
+ "seconds": 0.10002350807189941,
+ "failures": []
+ },
+ {
+ "field": "synth-full-BOUND0.99",
+ "method": "entropic GW annealed",
+ "raw": 0.9609375,
+ "refined": 0.9609375,
+ "seconds": 3.706202586491903,
+ "failures": []
+ },
+ {
+ "field": "synth-full-BOUND0.99",
+ "method": "BAPG GW",
+ "raw": 0.28515625,
+ "refined": 0.41796875,
+ "seconds": 1.7752447128295898,
+ "failures": []
+ },
+ {
+ "field": "natural-BOUND0.29",
+ "method": "grampa (current)",
+ "raw": 0.0078125,
+ "refined": 0.02734375,
+ "seconds": 0.01630059878031413,
+ "failures": []
+ },
+ {
+ "field": "natural-BOUND0.29",
+ "method": "umeyama",
+ "raw": 0.0234375,
+ "refined": 0.04296875,
+ "seconds": 0.014456351598103842,
+ "failures": []
+ },
+ {
+ "field": "natural-BOUND0.29",
+ "method": "gromov-wasserstein",
+ "raw": 0.0,
+ "refined": 0.0,
+ "seconds": 0.08861684799194336,
+ "failures": []
+ },
+ {
+ "field": "natural-BOUND0.29",
+ "method": "entropic GW annealed",
+ "raw": 0.0078125,
+ "refined": 0.005208333333333333,
+ "seconds": 18.440150499343872,
+ "failures": []
+ },
+ {
+ "field": "natural-BOUND0.29",
+ "method": "BAPG GW",
+ "raw": 0.0,
+ "refined": 0.010416666666666666,
+ "seconds": 1.8776884078979492,
+ "failures": []
+ }
+ ]
+} \ No newline at end of file