anhtld commited on
Commit
da81a41
·
verified ·
1 Parent(s): 84b30d7

auto-sync 2026-07-03T16:53:36Z workspace (part 2)

Browse files
workspace/latex/main.fdb_latexmk CHANGED
@@ -1,16 +1,17 @@
1
  # Fdb version 4
2
- ["bibtex main"] 1783096787 "main.aux" "main.bbl" "main" 1783096788 0
3
  "./references.bib" 1783037215 3572 027cd2cf27b344368db784320d235a8c ""
4
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/bibtex/bst/base/plain.bst" 1292289607 20613 bd3fbfa9f64872b81ac57a0dd2ed855f ""
5
- "main.aux" 1783096788 14301 740249504be2f28b8c99628a336fc931 "pdflatex"
6
  (generated)
7
  "main.bbl"
8
  "main.blg"
9
  (rewritten before read)
10
- ["pdflatex"] 1783096787 "main.tex" "main.pdf" "main" 1783096788 0
11
  "../paper/sections/theory.tex" 1783037322 4023 e8c0485606fdfea0020faae044bf1189 ""
12
  "../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex" 1783094939 342 289b441acb02d293402c3de06a7e2f7a ""
13
  "../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex" 1783096583 342 20aaeaf8942b48fd162703dc2da041d7 ""
 
14
  "../runs/ctt_base_context_obs_test_rollout_comparison/table.tex" 1783094772 1571 f49c51d0d4b2c7c13fa84fd347a0ef88 ""
15
  "../runs/ctt_base_context_obs_val_rollout_comparison/table.tex" 1783094577 1571 0214b06f8c4f37b139d6bb48248ef861 ""
16
  "../runs/ctt_dominance_val_to_test/table.tex" 1783050265 334 9ae614959acc0004c8ac706ce66d7d3e ""
@@ -137,10 +138,10 @@
137
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/web2c/texmf.cnf" 1692823885 39639 818dd8e5a0dce3a72597a7e6bad45d55 ""
138
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/var/lib/texmf/fonts/map/pdftex/updmap/pdftex.map" 1692821535 5031293 07973d5e761f645996ac32ebf7b8d1f3 ""
139
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/var/lib/texmf/web2c/pdftex/pdflatex.fmt" 1692821530 1386016 abd9ba6eb53c47bc5de8d672b6ac87d3 ""
140
- "main.aux" 1783096788 14301 740249504be2f28b8c99628a336fc931 "pdflatex"
141
- "main.bbl" 1783096787 3120 92555a174bae4ac684748bbe3fae0b8c "bibtex main"
142
- "main.out" 1783096788 2397 c088c8cc9362c16f4a98ebf8fe3ab830 "pdflatex"
143
- "main.tex" 1783096713 35884 95d914568772edd7603e2bbd1b4153b4 ""
144
  "tables/car_decomposition.tex" 1783032219 662 bac24c0014c2ba573860ce199b776934 ""
145
  "tables/main_results.tex" 1783004655 1061 19bfdcc844614aadefc921dae15c700d ""
146
  "tables/selector_calibration.tex" 1783011198 788 1b54a8eb88519c50862f4acf718798ab ""
 
1
  # Fdb version 4
2
+ ["bibtex main"] 1783097658 "main.aux" "main.bbl" "main" 1783097699 0
3
  "./references.bib" 1783037215 3572 027cd2cf27b344368db784320d235a8c ""
4
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/bibtex/bst/base/plain.bst" 1292289607 20613 bd3fbfa9f64872b81ac57a0dd2ed855f ""
5
+ "main.aux" 1783097699 15245 f6629e0989e6efe4e3f8e2623a10fe12 "pdflatex"
6
  (generated)
7
  "main.bbl"
8
  "main.blg"
9
  (rewritten before read)
10
+ ["pdflatex"] 1783097699 "main.tex" "main.pdf" "main" 1783097699 0
11
  "../paper/sections/theory.tex" 1783037322 4023 e8c0485606fdfea0020faae044bf1189 ""
12
  "../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex" 1783094939 342 289b441acb02d293402c3de06a7e2f7a ""
13
  "../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex" 1783096583 342 20aaeaf8942b48fd162703dc2da041d7 ""
14
+ "../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex" 1783097340 344 b41a694c7be179a8b2e13d9326613f99 ""
15
  "../runs/ctt_base_context_obs_test_rollout_comparison/table.tex" 1783094772 1571 f49c51d0d4b2c7c13fa84fd347a0ef88 ""
16
  "../runs/ctt_base_context_obs_val_rollout_comparison/table.tex" 1783094577 1571 0214b06f8c4f37b139d6bb48248ef861 ""
17
  "../runs/ctt_dominance_val_to_test/table.tex" 1783050265 334 9ae614959acc0004c8ac706ce66d7d3e ""
 
138
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/web2c/texmf.cnf" 1692823885 39639 818dd8e5a0dce3a72597a7e6bad45d55 ""
139
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/var/lib/texmf/fonts/map/pdftex/updmap/pdftex.map" 1692821535 5031293 07973d5e761f645996ac32ebf7b8d1f3 ""
140
  "/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/var/lib/texmf/web2c/pdftex/pdflatex.fmt" 1692821530 1386016 abd9ba6eb53c47bc5de8d672b6ac87d3 ""
141
+ "main.aux" 1783097699 15245 f6629e0989e6efe4e3f8e2623a10fe12 "pdflatex"
142
+ "main.bbl" 1783097658 3120 92555a174bae4ac684748bbe3fae0b8c "bibtex main"
143
+ "main.out" 1783097699 2397 c088c8cc9362c16f4a98ebf8fe3ab830 "pdflatex"
144
+ "main.tex" 1783097691 37307 d810e4b0228934db06130af13e988da3 ""
145
  "tables/car_decomposition.tex" 1783032219 662 bac24c0014c2ba573860ce199b776934 ""
146
  "tables/main_results.tex" 1783004655 1061 19bfdcc844614aadefc921dae15c700d ""
147
  "tables/selector_calibration.tex" 1783011198 788 1b54a8eb88519c50862f4acf718798ab ""
workspace/latex/main.fls CHANGED
@@ -789,6 +789,21 @@ INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.t
789
  INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex
790
  INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex
791
  INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
792
  INPUT ./main.bbl
793
  INPUT main.bbl
794
  INPUT ./main.bbl
 
789
  INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex
790
  INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex
791
  INPUT ../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex
792
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
793
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
794
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
795
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
796
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
797
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
798
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
799
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
800
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
801
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
802
+ INPUT ../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex
803
+ INPUT ../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex
804
+ INPUT ../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex
805
+ INPUT ../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex
806
+ INPUT ../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex
807
  INPUT ./main.bbl
808
  INPUT main.bbl
809
  INPUT ./main.bbl
workspace/latex/main.log CHANGED
@@ -1,4 +1,4 @@
1
- This is pdfTeX, Version 3.141592653-2.6-1.40.22 (TeX Live 2021 Gentoo Linux) (preloaded format=pdflatex 2023.8.23) 3 JUL 2026 12:39
2
  entering extended mode
3
  restricted \write18 enabled.
4
  %&-line parsing enabled.
@@ -542,61 +542,56 @@ oke[]seed0[]v3\OT1/cmr/m/n/10 ),
542
  (../runs/ctt_learned_dominance_val_to_test/table.tex) [8]
543
  (../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex)
544
  (../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex)
545
- [9]
546
- Overfull \hbox (31.94885pt too wide) in paragraph at lines 620--624
 
547
  []\OT1/cmtt/m/n/10 scripts/export[]cil[]charts.py\OT1/cmr/m/n/10 , \OT1/cmtt/m/
548
  n/10 scripts/build[]data[]accounting.py\OT1/cmr/m/n/10 , and \OT1/cmtt/m/n/10 s
549
  cripts/audit[]cil[]charts.py
550
  []
551
 
552
 
553
- Overfull \hbox (127.11275pt too wide) in paragraph at lines 626--630
554
  []\OT1/cmtt/m/n/10 cil/chart[]features.py \OT1/cmr/m/n/10 cen-tral-izes deploym
555
  ent-visible chart fea-ture con-struc-tion, and \OT1/cmtt/m/n/10 scripts/audit[]
556
  chart[]feature[]sources.py
557
  []
558
 
559
 
560
- Overfull \hbox (104.78587pt too wide) in paragraph at lines 630--633
561
  []\OT1/cmtt/m/n/10 scripts/slurm/render[]six[]task[]chart[]observations.sbatch
562
  \OT1/cmr/m/n/10 and \OT1/cmtt/m/n/10 scripts/slurm/reexport[]rgb[]ref[]cil[]cha
563
  rts.sbatch
564
  []
565
 
566
 
567
- Overfull \hbox (7.97675pt too wide) in paragraph at lines 637--642
568
  []\OT1/cmtt/m/n/10 scripts/eval[]ctt[]generated[]rollout.py \OT1/cmr/m/n/10 and
569
  \OT1/cmtt/m/n/10 scripts/slurm/eval[]ctt[]generated[]rollout.sbatch
570
  []
571
 
572
 
573
- Overfull \hbox (20.61057pt too wide) in paragraph at lines 637--642
574
  \OT1/cmr/m/n/10 train-split cal-i-bra-tion and meta-data load-ing for deploymen
575
  t-visible chart fea-tures such as \OT1/cmtt/m/n/10 base[]context[]obs\OT1/cmr/m
576
  /n/10 .
577
  []
578
 
579
  [10]
580
- Overfull \hbox (0.38913pt too wide) in paragraph at lines 650--655
581
- []\OT1/cmtt/m/n/10 scripts/eval[]learned[]dominance[]selector.py \OT1/cmr/m/n/1
582
- 0 trains a small validation-calibrated dom-i-nance model
583
- []
584
-
585
-
586
- Overfull \hbox (6.0625pt too wide) in paragraph at lines 655--658
587
  []\OT1/cmtt/m/n/10 scripts/eval[]chart[]positive[]memory[]proxy.py \OT1/cmr/m/n
588
  /10 and \OT1/cmtt/m/n/10 scripts/build[]ctt[]proxy[]comparison.py \OT1/cmr/m/n/
589
  10 gen-
590
  []
591
 
592
 
593
- Overfull \hbox (20.3325pt too wide) in paragraph at lines 660--663
594
  []\OT1/cmtt/m/n/10 scripts/train[]utility[]energy.py \OT1/cmr/m/n/10 and \OT1/c
595
  mtt/m/n/10 scripts/calibrate[]dominance.py \OT1/cmr/m/n/10 im-ple-ment the util
596
  -ity/scoring
597
  []
598
 
599
- [11] [12] (./main.bbl) [13] [14] (./main.aux)
600
 
601
  LaTeX Font Warning: Some font shapes were not available, defaults substituted.
602
 
@@ -604,13 +599,13 @@ Package rerunfilecheck Info: File `main.out' has not changed.
604
  (rerunfilecheck) Checksum: C088C8CC9362C16F4A98EBF8FE3AB830;2397.
605
  )
606
  Here is how much of TeX's memory you used:
607
- 10130 strings out of 480884
608
- 151542 string characters out of 5900692
609
- 507452 words of memory out of 5000000
610
- 27112 multiletter control sequences out of 15000+600000
611
  414597 words of font info for 70 fonts, out of 8000000 for 9000
612
  36 hyphenation exceptions out of 8191
613
- 71i,10n,74p,608b,398s stack positions out of 5000i,500n,10000p,200000b,80000s
614
  {/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/font
615
  s/enc/dvips/cm-super/cm-super-ts1.enc}</cvmfs/soft.computecanada.ca/gentoo/2023
616
  /x86-64-v3/usr/share/texmf-dist/fonts/type1/public/amsfonts/cm/cmbx10.pfb></cvm
@@ -649,10 +644,10 @@ are/texmf-dist/fonts/type1/public/amsfonts/cm/cmtt10.pfb></cvmfs/soft.computeca
649
  nada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/fonts/type1/public/amsfonts/
650
  symbols/msbm10.pfb></cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/shar
651
  e/texmf-dist/fonts/type1/public/cm-super/sfrm1000.pfb>
652
- Output written on main.pdf (14 pages, 347589 bytes).
653
  PDF statistics:
654
- 312 PDF objects out of 1000 (max. 8388607)
655
- 268 compressed objects within 3 object streams
656
- 72 named destinations out of 1000 (max. 500000)
657
  129 words of extra memory for PDF output out of 10000 (max. 10000000)
658
 
 
1
+ This is pdfTeX, Version 3.141592653-2.6-1.40.22 (TeX Live 2021 Gentoo Linux) (preloaded format=pdflatex 2023.8.23) 3 JUL 2026 12:54
2
  entering extended mode
3
  restricted \write18 enabled.
4
  %&-line parsing enabled.
 
542
  (../runs/ctt_learned_dominance_val_to_test/table.tex) [8]
543
  (../runs/ctt_base_context_obs_learned_dominance_context_val_to_test/table.tex)
544
  (../runs/ctt_base_context_obs_learned_dominance_train_to_test/table.tex)
545
+ (../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/
546
+ table.tex) [9]
547
+ Overfull \hbox (31.94885pt too wide) in paragraph at lines 641--645
548
  []\OT1/cmtt/m/n/10 scripts/export[]cil[]charts.py\OT1/cmr/m/n/10 , \OT1/cmtt/m/
549
  n/10 scripts/build[]data[]accounting.py\OT1/cmr/m/n/10 , and \OT1/cmtt/m/n/10 s
550
  cripts/audit[]cil[]charts.py
551
  []
552
 
553
 
554
+ Overfull \hbox (127.11275pt too wide) in paragraph at lines 647--651
555
  []\OT1/cmtt/m/n/10 cil/chart[]features.py \OT1/cmr/m/n/10 cen-tral-izes deploym
556
  ent-visible chart fea-ture con-struc-tion, and \OT1/cmtt/m/n/10 scripts/audit[]
557
  chart[]feature[]sources.py
558
  []
559
 
560
 
561
+ Overfull \hbox (104.78587pt too wide) in paragraph at lines 651--654
562
  []\OT1/cmtt/m/n/10 scripts/slurm/render[]six[]task[]chart[]observations.sbatch
563
  \OT1/cmr/m/n/10 and \OT1/cmtt/m/n/10 scripts/slurm/reexport[]rgb[]ref[]cil[]cha
564
  rts.sbatch
565
  []
566
 
567
 
568
+ Overfull \hbox (7.97675pt too wide) in paragraph at lines 658--663
569
  []\OT1/cmtt/m/n/10 scripts/eval[]ctt[]generated[]rollout.py \OT1/cmr/m/n/10 and
570
  \OT1/cmtt/m/n/10 scripts/slurm/eval[]ctt[]generated[]rollout.sbatch
571
  []
572
 
573
 
574
+ Overfull \hbox (20.61057pt too wide) in paragraph at lines 658--663
575
  \OT1/cmr/m/n/10 train-split cal-i-bra-tion and meta-data load-ing for deploymen
576
  t-visible chart fea-tures such as \OT1/cmtt/m/n/10 base[]context[]obs\OT1/cmr/m
577
  /n/10 .
578
  []
579
 
580
  [10]
581
+ Overfull \hbox (6.0625pt too wide) in paragraph at lines 679--682
 
 
 
 
 
 
582
  []\OT1/cmtt/m/n/10 scripts/eval[]chart[]positive[]memory[]proxy.py \OT1/cmr/m/n
583
  /10 and \OT1/cmtt/m/n/10 scripts/build[]ctt[]proxy[]comparison.py \OT1/cmr/m/n/
584
  10 gen-
585
  []
586
 
587
 
588
+ Overfull \hbox (20.3325pt too wide) in paragraph at lines 684--687
589
  []\OT1/cmtt/m/n/10 scripts/train[]utility[]energy.py \OT1/cmr/m/n/10 and \OT1/c
590
  mtt/m/n/10 scripts/calibrate[]dominance.py \OT1/cmr/m/n/10 im-ple-ment the util
591
  -ity/scoring
592
  []
593
 
594
+ [11] [12] (./main.bbl [13]) [14] (./main.aux)
595
 
596
  LaTeX Font Warning: Some font shapes were not available, defaults substituted.
597
 
 
599
  (rerunfilecheck) Checksum: C088C8CC9362C16F4A98EBF8FE3AB830;2397.
600
  )
601
  Here is how much of TeX's memory you used:
602
+ 10136 strings out of 480884
603
+ 151941 string characters out of 5900692
604
+ 507836 words of memory out of 5000000
605
+ 27114 multiletter control sequences out of 15000+600000
606
  414597 words of font info for 70 fonts, out of 8000000 for 9000
607
  36 hyphenation exceptions out of 8191
608
+ 71i,10n,74p,608b,450s stack positions out of 5000i,500n,10000p,200000b,80000s
609
  {/cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/font
610
  s/enc/dvips/cm-super/cm-super-ts1.enc}</cvmfs/soft.computecanada.ca/gentoo/2023
611
  /x86-64-v3/usr/share/texmf-dist/fonts/type1/public/amsfonts/cm/cmbx10.pfb></cvm
 
644
  nada.ca/gentoo/2023/x86-64-v3/usr/share/texmf-dist/fonts/type1/public/amsfonts/
645
  symbols/msbm10.pfb></cvmfs/soft.computecanada.ca/gentoo/2023/x86-64-v3/usr/shar
646
  e/texmf-dist/fonts/type1/public/cm-super/sfrm1000.pfb>
647
+ Output written on main.pdf (14 pages, 348856 bytes).
648
  PDF statistics:
649
+ 318 PDF objects out of 1000 (max. 8388607)
650
+ 273 compressed objects within 3 object streams
651
+ 73 named destinations out of 1000 (max. 500000)
652
  129 words of extra memory for PDF output out of 10000 (max. 10000000)
653
 
workspace/latex/main.pdf CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f511286398b71b7d8610e2bb81dfa67cbeb5001d4afb6bb4ccc3455ac08e0e2e
3
- size 345628
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edc7ba7e13f09defcd2592970097e2ad85cf1cacae02602756814c76d24cb3b1
3
+ size 347589
workspace/latex/main.tex CHANGED
@@ -59,7 +59,7 @@ test success at 25.69\% coverage. We keep the metric boundary explicit:
59
  \outcomeptr{} is reported only after generated candidates are rolled out, while
60
  distance-only support diagnostics are \pptc{}. These runs are therefore
61
  diagnostic evidence for the Atlas thesis and the next selector/generator gate,
62
- not an unqualified SOTA claim.
63
  \end{abstract}
64
 
65
  \section{Introduction}
@@ -611,6 +611,27 @@ success selector gap. The paper should therefore treat train-only dominance as
611
  evidence that the current selector is underpowered, not as a solved deployment
612
  method.
613
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
614
  \section{Reproducibility Artifacts}
615
 
616
  The current draft is backed by explicit artifacts:
@@ -648,10 +669,13 @@ The current draft is backed by explicit artifacts:
648
  measured rows; it can use rollout row scores or recompute scores from a
649
  train-only utility-energy checkpoint.
650
  \item \texttt{scripts/eval\_learned\_dominance\_selector.py} trains a small
651
- validation-calibrated dominance model over deployment-visible candidate
652
- features and evaluates it on held-out measured test rows; it also logs
653
- feature/target ablations, including context metadata and tangent-code
654
  variants, for selector diagnostics.
 
 
 
655
  \item \texttt{scripts/eval\_chart\_positive\_memory\_proxy.py} and
656
  \texttt{scripts/build\_ctt\_proxy\_comparison.py} generate the local-atlas
657
  baseline and validation proxy gate table.
@@ -668,7 +692,7 @@ The current draft is backed by explicit artifacts:
668
 
669
  \section{Limitations and Next Steps}
670
 
671
- This draft is not an unqualified SOTA claim. The current \ctt{} evidence
672
  includes validation and test measured generated-candidate rollouts, but the
673
  selected action fails the internal paper gate even when the test proposal oracle
674
  passes 50\%. The missing method component is calibrated dominance and a better
 
59
  \outcomeptr{} is reported only after generated candidates are rolled out, while
60
  distance-only support diagnostics are \pptc{}. These runs are therefore
61
  diagnostic evidence for the Atlas thesis and the next selector/generator gate,
62
+ not a final deployment claim.
63
  \end{abstract}
64
 
65
  \section{Introduction}
 
611
  evidence that the current selector is underpowered, not as a solved deployment
612
  method.
613
 
614
+ \begin{table}[t]
615
+ \centering
616
+ \caption{Nonlinear train-calibrated selector diagnostic. The model and
617
+ threshold are selected only on held-out train-calibration rows, then evaluated
618
+ once on the held-out test rollout rows. The best nonlinear row does not beat
619
+ Table~\ref{tab:ctt-base-context-obs-learned-train-dominance}, so the current
620
+ bottleneck is not just linear separability in the dominance selector.}
621
+ \label{tab:ctt-base-context-obs-nonlinear-train-dominance}
622
+ \scriptsize
623
+ \resizebox{\linewidth}{!}{\input{../runs/ctt_base_context_obs_nonlinear_dominance_basic_positive_train_to_test/table.tex}}
624
+ \end{table}
625
+
626
+ Table~\ref{tab:ctt-base-context-obs-nonlinear-train-dominance} tests the
627
+ obvious stronger-selector hypothesis with histogram-gradient and random-forest
628
+ diagnostics. The best nonlinear train-calibrated row reaches 30.56\% selected
629
+ test success, below the 31.25\% clean ridge row above, while using the same
630
+ test proposal set and leakage-audited train calibration source exclusion. This
631
+ negative result supports the next engineering target: richer visual-language or
632
+ object-centric chart features and outcome-vector utility modeling, not merely a
633
+ more flexible shallow selector over the current feature set.
634
+
635
  \section{Reproducibility Artifacts}
636
 
637
  The current draft is backed by explicit artifacts:
 
669
  measured rows; it can use rollout row scores or recompute scores from a
670
  train-only utility-energy checkpoint.
671
  \item \texttt{scripts/eval\_learned\_dominance\_selector.py} trains a small
672
+ validation- or train-calibrated dominance model over deployment-visible
673
+ candidate features and evaluates it on held-out measured test rows; it also
674
+ logs feature/target ablations, including context metadata and tangent-code
675
  variants, for selector diagnostics.
676
+ \item \texttt{scripts/eval\_nonlinear\_dominance\_selector.py} runs the
677
+ train-calibrated nonlinear selector diagnostics in
678
+ \texttt{runs/ctt\_base\_context\_obs\_nonlinear\_dominance\_*}.
679
  \item \texttt{scripts/eval\_chart\_positive\_memory\_proxy.py} and
680
  \texttt{scripts/build\_ctt\_proxy\_comparison.py} generate the local-atlas
681
  baseline and validation proxy gate table.
 
692
 
693
  \section{Limitations and Next Steps}
694
 
695
+ This draft is not yet a final deployment method. The current \ctt{} evidence
696
  includes validation and test measured generated-candidate rollouts, but the
697
  selected action fails the internal paper gate even when the test proposal oracle
698
  passes 50\%. The missing method component is calibrated dominance and a better