Official-test scalar attribution inventory

Research Purpose And Boundary

This part preserves the individual scalar attribution and correlation inventory for the frozen Task 43 raw-inclusive hybrid. The root reference owns the benchmark claim, component-head performance, family totals, mapping contrast, population stability, and external comparison. This part owns only the 37 scalar rows and their measured overlap on the fixed 5,042-row training matrix.

Mean absolute TreeSHAP measures use by the existing nonlinear ensemble. The Pearson values measure association. Neither quantity is an additive QWK contribution, causal effect, or feature-removal result.

Evidence And Sources

Canonical component result: .orchestration/context/sessions/01a02917-0817-773e-90d4-43870e2002b7/evidence/task-hule-20-44-03/component-audit-result-v3.json, SHA-256 79e9f8c1d17ec6fa36e3d035cf620d0468462b8190628bc4cd6383733f185bbb.
Ordered scalar inventory: .orchestration/context/sessions/01a02917-0817-773e-90d4-43870e2002b7/evidence/task-hule-20-44-03/scalar-features-v3.jsonl, SHA-256 111b11ea70c85ccaa8f68744aad17ffcfed5ce8205b0f35be82f1aeb35c3166e.
Population: the fixed 3,479 base plus 1,563 candidate training rows used by the frozen raw-inclusive hybrid.
Attribution: member-averaged TreeSHAP from the five frozen 805-column hybrid models. Its ensemble reconstruction maximum absolute error is 0.00000477.
Correlation: Pearson correlation on the same 5,042 rows. Max |r| is the maximum absolute correlation between one scalar and any one of the 768 embedding coordinates.

Findings And Interpretation

Four scalar features share the accepted DeBERTa runtime lineage: sent_similarity_variance, prompt_similarity, intro_prompt_sim, and min_para_relevance. The remaining 33 scalars do not consume that embedding runtime.

Construct family	Scalar feature	Mean absolute TreeSHAP	Max \|r\|
Correction/error	`grammar_errors_per_100_words`	0.001296	0.500
Correction/error	`language_tool_spelling_errors_per_100_words`	0.017286	0.660
Correction/error	`final_spelling_issues_per_100_words`	0.002690	0.678
Correction/error	`punctuation_errors_per_100_words`	0.000000	0.281
Readability	`smog`	0.000183	0.329
Readability	`coleman_liau`	0.001813	0.584
Readability	`ari`	0.003098	0.383
Readability	`avg_sentence_length`	0.001825	0.428
Lexical statistics	`ttr`	0.000103	0.660
Length	`word_count`	0.052873	0.378
Length	`avg_word_length`	0.001016	0.591
Syntax	`parse_tree_depth`	0.000403	0.472
Syntax	`passive_ratio`	0.000367	0.218
Syntax	`dep_distance`	0.000984	0.424
Semantic coherence, embedding-derived	`sent_similarity_variance`	0.000377	0.161
Prompt relevance, embedding-derived	`prompt_similarity`	0.000761	0.305
Prompt relevance, embedding-derived	`intro_prompt_sim`	0.000165	0.240
Prompt relevance, embedding-derived	`min_para_relevance`	0.000000	0.286
Complex nominal	`procedural_or_event_head_complex_np_density_log1p`	0.000095	0.264
RST relation	`tier3_rst_gum_primary_tree_satellite_relation_rate`	0.001632	0.332
Clause rates	`finite_advcl_per_100w`	0.000885	0.259
Clause rates	`finite_relative_clause_per_100w`	0.000128	0.234
Clause rates	`nonfinite_participial_advcl_per_100w`	0.000321	0.205
Clause rates	`nonfinite_participial_acl_per_100w`	0.000000	0.172
Frequency/sophistication	`p_lex_lambda`	0.000175	0.423
Frequency/sophistication	`advanced_guiraud`	0.018843	0.487
Word-range dispersion	`mean_log_contextual_diversity`	0.000213	0.466
Word-range dispersion	`mean_contextual_diversity_percent`	0.000000	0.496
Word-range dispersion	`contextual_diversity_coverage`	0.000000	0.352
Moving-average lexical diversity	`mattr_50`	0.000000	0.566
Concreteness	`mean_concreteness`	0.000390	0.461
Concreteness	`concreteness_coverage`	0.000511	0.304
Word prevalence	`mean_word_prevalence`	0.000231	0.257
Word prevalence	`word_prevalence_coverage`	0.000507	0.283
Lexical decision	`mean_lexical_decision_rt_ms`	0.000728	0.424
Lexical decision	`lexical_decision_rt_coverage`	0.002162	0.252
Causal-predicate complex nominal	`abstract_np_with_causal_predicate`	0.000376	0.244

No scalar is constant, and no two scalar columns are exact duplicates. Eight scalar pairs have absolute Pearson correlation at least 0.8:

Scalar pair	r
`ari` / `avg_sentence_length`	0.988
`mean_log_contextual_diversity` / `mean_contextual_diversity_percent`	0.953
`coleman_liau` / `avg_word_length`	0.945
`concreteness_coverage` / `word_prevalence_coverage`	0.936
`language_tool_spelling_errors_per_100_words` / `final_spelling_issues_per_100_words`	0.935
`ari` / `parse_tree_depth`	0.833
`avg_sentence_length` / `parse_tree_depth`	0.832
`p_lex_lambda` / `mean_log_contextual_diversity`	-0.813

The high correlations include shared denominators and overlapping constructs, but correlation alone does not establish that one feature is removable. The TreeSHAP values likewise describe the fitted ensemble on this matrix; they do not report performance after removing a feature.

Evidence Gaps And Follow-Up

No matched per-feature removal fits were run. They were outside Task 03 and would add 37 post-hoc model-selection comparisons. The three declared reduced heads and the root reference's family attribution are the bounded component evidence for this story.