wassname/moral-maps
1
0
Fork 0
mirror of https://github.com/wassname/moral-maps.git synced 2026-09-09 11:27:22 +08:00
Code Issues Packages Projects Releases Wiki Activity

journal: coherent-C sweep -- mfq2 coherent to C=3, showcase C=1 validated

Browse Source 
job 234: ordinal pmass 1.0 at both poles up to C=3.0, steer grows 0.129->0.324.
C=1 is well inside the coherent range. Joint-coherent C is bounded by the side
instruments' -C neutral-degeneracy (a model property at C=1), not ordinal
coherence. Completes the goal's "sweep for the largest coherent C" clause.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>
This commit is contained in:
wassname
2026-06-25 04:57:37 +08:00
co-authored by Claudypoo
parent f5fc6302d4
commit 5be5451e13
1 changed files with 10 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/RESEARCH_JOURNAL.md
+10
View File
@@ -715,3 +715,13 @@ the current core gives 0.788 (probe_word_readout.py). So this model's MFV top-1
0.780); the 82.6% was a stale/erroneous table entry, not a reachable target. 0.773
is the canonical (digit) value. Switching the showcase to Qwen3-4B gave the
cleaner, fully-coherent, bidirectional result.
Coherent-C sweep (job 234, csweep_coherent_c.py, mfq2 Qwen3-4B): the ordinal
readout stays fully coherent (pmass 1.000 at BOTH poles) all the way to C=3.0,
with the steer growing monotonically -- mean|profile delta| 0.129 (C=0.5), 0.215
(1.0), 0.265 (1.5), 0.296 (2.0), 0.324 (3.0). So the showcase's fixed C=1 is well
inside the coherent range, not at its edge; the headline path tolerates much
stronger steering. The binding constraint on a joint-coherent C across all five
instruments is the side instruments' -C neutral-degeneracy (profile pins to 3.0,
already at C=1; pmass stays ~1.0), a model property, not an ordinal coherence
break. C=1 stands as a valid, coherent real coefficient.