Weksa
2b29efbc2d
map: recalibration wording after Soul s3 (ja effect, R, mild arms, noise stratum)
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-13 07:33:08 +00:00
Weksa
915863b5b3
map: recalibration on the 35-item Validation corpus; the Train-guide gain was in-sample
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-13 07:24:51 +00:00
44b66f94dc
validation-growth: 42 blind-written Validation items and scenes (35 per language) ( #75 )
2026-11-13 05:29:01 +00:00
43ec24877f
map: validation growth plan and the smallest measurable distil slice ( #74 )
2026-11-13 05:04:52 +00:00
2dcc90796f
cycle floor: paired_floor over round 1 and three prompted replicates ( #73 )
2026-11-13 05:02:31 +00:00
516eb4b885
round: per-item Validation scores in the round record (cycle-floor-scores) ( #71 )
2026-11-13 03:47:53 +00:00
98ab798ca5
paired floor: one composite formula and its pin; docs state the power measure ( #72 )
2026-11-13 03:39:17 +00:00
353dd2138f
relower: paired per-item noise floor over >=3 prompted repeats; panel gain clears the noise of the sum ( #70 )
2026-11-13 03:01:52 +00:00
97aa026498
weksa loop relower: fixed-packet re-lowering measure with a lowering-only floor ( #69 )
2026-11-13 02:05:41 +00:00
ee3e4a2e38
withheld-candidate: candidate round stops on a withheld lowering prompt ( #68 )
2026-11-13 01:29:57 +00:00
8325d7e0b2
test-text-withheld-arms: withheld prompts stop every model call ( #67 )
2026-11-13 01:23:08 +00:00
a0aafaff19
delete-cultural-ontology: data/cultural-ontology deleted ( #66 )
2026-11-13 01:03:20 +00:00
736b6a4acb
test-text-guard: AnswerFiles refuses a prompt that holds the text of an item in a Test group ( #65 )
2026-11-13 00:57:32 +00:00
887e79ef44
mint-prompt-door: the mint prompt teaches the shape the admit door admits ( #64 )
2026-11-12 23:54:01 +00:00
754322252b
ontology-mint-probe: weksa loop mint-ontology proposes entries from Train items only, never writes the run ontology ( #63 )
2026-11-12 22:20:44 +00:00
af0a56d45f
ja-anger-mirror: ja-JP:3 gets hedge and affection bounds as en-US:3 carries ( #62 )
2026-11-12 21:40:39 +00:00
349fe8e239
en-us-6-8: en-US:8 gets a certainty bound so affection and flat certainty never co-select ( #61 )
2026-11-12 20:56:10 +00:00
0a2d00a25f
ja-surprise-certainty: ja-JP:9 reworded so its question ending is available, not prescribed ( #60 )
2026-11-12 20:27:19 +00:00
ccef7a761f
en-us-seed-pairs: en-US :3/:5/:6 made disjoint by conditions, shared sweep helper ( #59 )
2026-11-12 20:18:45 +00:00
981ad0ea1f
ja-seed-pairs: ja-JP :5/:7/:8 get certainty and restraint bounds (AtMostOrUnset) so the seed pairs never co-select ( #58 )
2026-11-12 19:52:14 +00:00
40469430b9
ja-withdraw-scene-entries: withdraw onto:ja-JP:14 and :15 (net-negative on register in the paired measure) ( #57 )
2026-11-12 19:17:39 +00:00
07c2755b95
ja-item19: encoder prompt states one role per predication and the group shape for coordination ( #56 )
2026-11-12 18:48:40 +00:00
f1f508e159
ja-inventory-6: rev 6 concepts (try, someday, a little, as one pleases, kite flyer) and the admit-cap test ( #55 )
2026-11-12 18:27:34 +00:00
7babf8aadc
ja-inventory: rev 5 concepts and frames for the original bed (model-drafted, Validation-informed), weksa inventory admit, onto:ja-JP:15 ( #54 )
2026-11-12 18:08:04 +00:00
002d3186c3
Merge pull request 'language-ontology-scene: scene features, one scene door, Scene-conditioned ja/en entries, runs read entries at their own head' ( #53 ) from weksa-lowering/language-ontology-scene into main
2026-11-12 17:22:09 +00:00
Weksa
139fa94b29
language-ontology-scene r2: a run reads entries and scenes at its own head, default bed is seed+proof+scene rev 3 (no Dao), politeness keyword guard replaced by model readings, nine seed entries reworded
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 17:11:57 +00:00
Weksa
71b9d0e055
language-ontology-scene fix: default bed carries rev 3-4, scene entries never co-select in a layer, en-US:8 names no register, keeper rank re-derived from situation, README and doc true
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 16:48:52 +00:00
Weksa
31a0c2406e
language-ontology-scene s7: 56 model-annotated keeper scenes and Scene-conditioned ontology entries (ja-JP:10-14, en-US:10-13) at inventory rev 4
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 16:33:36 +00:00
Weksa
966a3678a7
language-ontology-scene s3-6: inventory rev 4 in both beds, empty speaker-state stores committed, scene tests, original corpus checks, lowering map
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 16:31:55 +00:00
Weksa
f7f4e58d78
language-ontology-scene s1-2: the run pins the speaker-state store, assembled is the one scene door, select reads slot Scene, Prompt::lower refuses an unread scene, weksa scene put|dump, speaker_scene_revision(n)
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 16:27:13 +00:00
3cc06693fb
Merge pull request 'language-ontology-seed: model-drafted en-US and ja-JP seed entries; target-language YAML deleted' ( #52 ) from weksa-lowering/language-ontology-seed into main
2026-11-12 16:12:59 +00:00
Weksa
e8ba0f2eae
language-ontology-seed fix s4: ja-JP:6 bans ne-softening and hedges only; front door and lowering docs stop naming a target-language profile as register owner
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 16:09:36 +00:00
Weksa
dc5e649851
language-ontology-seed fix s3: ja-JP:3 and :6 name no politeness level; README claim made true; schema clause dropped
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 15:44:08 +00:00
Weksa
79fcc313ed
language-ontology-seed fix s2: neutral ja-JP:1 (no politeness per line), entry text defects, drafter residue out of the seed and score vectors, front-door and schema/tex prose
...
ontology.cc regenerated by weksa ontology admit, 9+9 entries, same generator string.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 15:39:48 +00:00
Weksa
fc99652f13
language-ontology-seed fix s1: rebuild the seed as 9+9 entries that read only stance and discourse and make true claims; pin the generator exactly; drop deleted store paths from the tex and schema; README says nothing grows the seed yet
...
ontology.cc is rebuilt from nothing (the branch is unmerged); no entry carries
over unchanged in number or text.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 15:29:11 +00:00
Weksa
6f14b00090
language-ontology-seed: model-drafted en-US and ja-JP seed entries; the target-language YAML, its typed copy and their schema prose deleted
...
Entries are Annotator::Model (Anthropic), inventory revision 1, drafted from the
YAML and standard references, not from any corpus text. pt-BR profiles are
deleted unconverted. Citations name data/ontology/ontology.cc.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 15:17:36 +00:00
0911c0198c
Merge pull request 'language-ontology: weksa.ontology_entry, its store, select, wired into the lowering prompt' ( #51 ) from weksa-lowering/language-ontology into main
2026-11-12 15:12:14 +00:00
Weksa
e46f933268
One language-tag owner: is_language_tag is canonical-case; tests for the case arms and Hangul fillers; leak-owner doc
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 14:51:31 +00:00
Weksa
602f7c2cd5
Ontology fix pass s1: refuse format and separator characters, canonical-case language, entries on retired features; kill six surviving mutants; two-language round test
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 14:31:35 +00:00
Weksa
cd435e2a48
Ontology tests: select, admit rules, prompt section, run pin, round prompts, draft door, CLI
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 14:14:44 +00:00
Weksa
0ea07d21fa
Ontology entries per language, selected inside Prompt::lower and pinned by the run
...
OntologyEntry, OntologyStore::admit, ontology::select over the view's stance and discourse, OntologyDraft, run keys 15 and 16 with Drift(Ontology), weksa ontology admit and dump.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 14:11:51 +00:00
Weksa
598cbf62d6
Map: cut 5 language-ontology specced as ontology store and seed cuts
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 14:06:16 +00:00
d9ebbcf889
Merge pull request 'ja-panel-corpus: 28 model-written Japanese passages and 8 Aozora Bunko Test sentences' ( #50 ) from weksa-lowering/ja-panel-corpus into main
2026-11-12 13:52:36 +00:00
Weksa
20863615b3
data/original: ja passages replaced, none a rendering of an en-US original or a Dao line
...
All 28 Train/Validation ja passages are new Japanese sentences on the inventory's
concepts with their own imagery (paddies, charcoal, kites, the bath, the market); none
renders an en-US original, and none restates a data/daodejing line (ch3, 5, 7, 8, 78) or
other canon (the old ones on 'true words', 'long-enduring', 'bellows', 'water vs hard
things' are gone). corpus.cc restored from 21eaa07
2026-11-12 13:51:06 +00:00
Weksa
e30ee3886a
data/original: ja passages rewritten onto the Dao r3 concepts; ja Test sentences pinned verbatim
...
The 28 Train/Validation ja passages are new Japanese sentences on the concepts the
pinned inventory holds (water, mind, belly, bone, sage, will, heaven, earth, fill, empty,
weaken, strengthen, compare, resemble ...), written in Japanese idiom, not translated.
The branch was unmerged, so corpus.cc was restored to 21eaa07
2026-11-12 13:51:06 +00:00
Weksa
4dbaef9437
data/original: ja-JP slice, 28 model-written Japanese passages and 8 Aozora Bunko Test sentences
...
Train and Validation (14 each) are Japanese passages written in Japanese, ModelWritten
Anthropic, generator claude-sonnet-5-5; none sits in a Test group. Test is eight
one-sentence excerpts (Soseki, Kyoka, Akutagawa, Ogai, Arishima, Kenji, Kajii, Kunikida)
taken from the card ruby zips (cp932, ruby and the pipe marker dropped, NFC) and checked
verbatim against each card's XHTML file; licence strings keep 底本, 入力 and 校正 within
the store's 200-byte cap. Entities appended (keeper gains 番人; eight pd-* authors).
Items 1..28 and holdout:1..8 unchanged (digest pinned). Tests per language: counts,
provenance, Test public-domain only, no markup, panel round reads both languages, speaker
forms, recall 56, no Test text in any prompt.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 13:51:06 +00:00
9f203a5604
Merge pull request 'ja-panel-loop: language is the item's, the panel the run's; panel accept rule' ( #49 ) from weksa-lowering/ja-panel-loop into main
2026-11-12 13:51:04 +00:00
Weksa
a1c1d9d599
Map: the regression test runs per language; decide adds the gain test for one language
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 13:50:48 +00:00
Weksa
f05906b1f7
Pin the panel sum rule's liveness, fix stale Ladder docs, name the composite verdict in the map record column
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 13:34:48 +00:00
Weksa
edb7fff339
Accept rule: composite tolerance back in the one regression test, panel accepts on a composite sum past one step; bound 1.5 L / LADDER_STEP
...
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-11-12 13:25:36 +00:00
