Document table provenance boundaries

README.md

@@ -41,6 +41,8 @@ Public Hugging Face links:
 data/paper_tables/
   Exact CSV versions of the camera-ready main and appendix tables, plus the
   raw 6-scenario x 50-draw six-family leaderboard used to regenerate Table 2.
+  Appendix robustness/transfer tables are paper snapshots; not all of them
+  are regenerated by the lightweight paper-core runner.
 
 data/paper_figures/
   Paper figure assets used by the leaderboard Space for visual alignment checks.
@@ -68,7 +70,8 @@ site_data/v1.0.0/
   regenerated from the current paper-aligned outputs.
 
 metadata/
-  Release manifest, schema notes, file inventory, and checksums.
+  Release manifest, schema notes, table provenance notes, file inventory, and
+  checksums.
 ```
 
 ## Paper-Aligned Scope
@@ -77,6 +80,12 @@ The primary paper studies standalone decomposition as component recovery under c
 
 The 22-method files are included as benchmark-only expansion and transfer tracks. Several rows correspond to additional method prototypes or benchmark-side mechanism proxies, not camera-ready paper claims. These rows use the same 6-scenario, 50-draw synthetic protocol where applicable, but they should not replace the primary six-family Table 2 / Figure 3 interpretation.
 
+The one-command reproduction path in this repository regenerates the primary
+paper-core synthetic leaderboard. Appendix robustness and transfer checks are
+released as exact camera-ready CSV summaries, with source summaries and run
+metadata where available. They should be used to audit the paper tables, but
+they are not all exposed through the same lightweight runner.
+
 ## Versioned Evidence Layout
 
 The release separates the frozen camera-ready paper snapshot from benchmark extensions:
@@ -136,6 +145,11 @@ Small CSV files matching the camera-ready manuscript tables. These are the safes
 - `paper_core_50draw_leaderboard.csv`: raw 1,800-row paper-core output (6 scenarios x 50 draws x 6 methods).
 - `paper_core_50draw_by_tier.csv`: tier-level means and metric-specific valid-row counts used to derive the tier-balanced Table 2 columns.
 
+Appendix table snapshots include bounded tuning, period robustness, alignment
+robustness, boundary sensitivity, real-data proxy, semi-synthetic transfer, and
+MSSA pilot summaries. Their audit status is summarized in
+`metadata/paper_table_provenance.md`.
+
 ### `synthetic_full22_extension`
 
 This benchmark-only track uses the same 6-scenario synthetic generator and five-metric evaluation protocol with length 512, 50 draws per scenario, true-period-given evaluation, and a 22-method roster. It includes raw rows, overall summaries, by-scenario summaries, by-tier summaries, coverage, protocol matrix, and backend-selection metadata. It is an expanded roster view, while the camera-ready Table 2 / Figure 3 source is the six-family paper table under `data/paper_tables/`.
@@ -148,6 +162,11 @@ This track uses public real time series with known periods, known mechanisms, or
 
 This benchmark-only transfer track injects known mechanisms into six real monthly backgrounds using a 22-method roster, three mechanisms, two background scales, and eight windows per setting. The released five-metric ranking is `data/semisynth_transfer/results/summary/ranking_paper_5metric_overall.csv`. It is a living benchmark extension rather than a camera-ready paper table. Large downloaded source files are intentionally excluded from this release; the canonical small CSV backgrounds and metric tables are included.
 
+The camera-ready paper's semi-synthetic paragraph is the frozen six-family
+snapshot in `data/paper_tables/semisynthetic_transfer_summary.csv`; the
+22-method transfer files in this directory are a later benchmark extension and
+may have different method-level ordering.
+
 ### `post_rebuttal_second_expansion`
 
 This later companion track contains rebuttal-stage real-data evidence. `real_physics_track_b` covers CO2 and tides with mechanism-informed approximate structure. `real_proxy_track_c` covers CH4, GPCC precipitation, NDVI, QBO, Arctic sea ice, and sunspots using proxy and stability diagnostics.

metadata/checksums.sha256

@@ -2,7 +2,7 @@ bde2eac437158b0e4e36581e3877d372cc6809689a6010997d9bd1962d244259  .gitattributes
 d66cca888321e611bd8a7ffa970433bf3a89619030aba1f4d705261202ef9833  .gitignore
 166849e0d6677f116ae778829175b9e333cb3bfed41ce58adfe256271fd304b1  CITATION.cff
 b4a90b60dda4e329cead21d9c212995337ac4ebfebe5365939c4c9c19479f419  LICENSE
-0bf2d72b1d15ca8db58ceb6e306c0806550843569ef9515a5aff981e6cf57736  README.md
+b60637cec88fa3cd3464f2d31a2216e93f2f13a6ed8c421c05bf4f2f32f428d4  README.md
 e5ded48d0eaffb7629beb060617381c3c087e93ab809bc40f9960a18440695b6  code/TSDecompose/README.md
 7a2ee7b8c6408c90f0b61db17068f70bf359a37cb8a83f490325c79003ffc4e5  code/TSDecompose/requirements.txt
 5d617e2292eea02ed79a3b30e26f708790ab54b2e352bb747449c1aa31bd7aee  code/TSDecompose/scripts/run_paper_benchmark.ps1
@@ -194,9 +194,10 @@ f290c28b0f23789e2cc51eb1999f0f63234b792c75a8cded3f68f9a913012ee6  data/synthetic
 a194ec34aa349f7177f89c2ae2b328f9589e7919f003659c6a5f0497ad32e91e  data/synthetic_full22_extension/results/summary/ranking_paper_5metric_scenarios.csv
 06bd8784cd8dfba5c8ce1d15aa12b4b1ce800ae293e012e5d626257c6fcb679b  data/synthetic_full22_extension/results/summary/ranking_paper_5metric_tiers.csv
 e7769cea54e0d0549add85f119aafdb11e933af5fd2cd09ec812f3639f75594a  data/synthetic_full22_extension/results/summary/report.md
-135ebf889b90ffc48f66082a3985ec44ef6c41aac8020e37b744a7a11860bfae  metadata/dataset_schema.json
-982f12bd0797718076ca7fb2c3a87ca2b624ba7292702a764096ecebd27e414b  metadata/file_inventory.csv
-77fa7b899340e197683462b0813f26d8b26a703ef4b27d0054401ffaf597c632  metadata/release_manifest.json
+5002d24ca61975f0e0028f9ba224a3e395b4169c37c965970f6514589cd22634  metadata/dataset_schema.json
+a76b98b9bf6a977fd03e8133a13e0b9a605cc75fdaaef17ce8023d7579562073  metadata/file_inventory.csv
+95a56db9e00d220f26eee2bc4e4355e4d93c522149179c6b84858c3c71c97b49  metadata/paper_table_provenance.md
+cff7af631e213dfa19fd939f2c770baf9ee53c80562f0dd7de4dc86fa7823853  metadata/release_manifest.json
 5b134e900a2308ad5d3a8c61f72909935c5d61de82350210458c2e978140b1d8  site_data/v1.0.0/evaluation_metrics.json
 02d55c80e7d63e5b907d39789614931c3dc0dcc73d97feaa0d25145f08e17637  site_data/v1.0.0/leaderboard_post_rebuttal_real_physics_track_b.csv
 956471fc5e08dad8b9bd2e14defb4a0d9d7561a3714c94340d89760397a09fe6  site_data/v1.0.0/leaderboard_post_rebuttal_real_proxy_track_c.csv

metadata/dataset_schema.json

@@ -7,6 +7,7 @@
       "paper_core_raw": "data/paper_tables/paper_core_50draw_leaderboard.csv",
       "paper_core_by_tier": "data/paper_tables/paper_core_50draw_by_tier.csv",
       "table2_display": "data/paper_tables/global_performance_summary.csv",
+      "table_provenance": "metadata/paper_table_provenance.md",
       "table2_aggregation": "Tier-balanced over valid metric values: stationary columns are the equal-weight average of Tier 1 and Tier 2 means; non-stationary columns are Tier 3 means. Seasonal metrics are undefined for the trend-only scenario, so metric-specific valid-row counts are provided.",
       "recommended_for_citation_checks": true
     },

metadata/file_inventory.csv

@@ -3,7 +3,7 @@
 ".gitignore","54","d66cca888321e611bd8a7ffa970433bf3a89619030aba1f4d705261202ef9833"
 "CITATION.cff","1025","166849e0d6677f116ae778829175b9e333cb3bfed41ce58adfe256271fd304b1"
 "LICENSE","1087","b4a90b60dda4e329cead21d9c212995337ac4ebfebe5365939c4c9c19479f419"
-"README.md","10549","0bf2d72b1d15ca8db58ceb6e306c0806550843569ef9515a5aff981e6cf57736"
+"README.md","11601","b60637cec88fa3cd3464f2d31a2216e93f2f13a6ed8c421c05bf4f2f32f428d4"
 "code/TSDecompose/README.md","3063","e5ded48d0eaffb7629beb060617381c3c087e93ab809bc40f9960a18440695b6"
 "code/TSDecompose/requirements.txt","189","7a2ee7b8c6408c90f0b61db17068f70bf359a37cb8a83f490325c79003ffc4e5"
 "code/TSDecompose/scripts/run_paper_benchmark.ps1","214","5d617e2292eea02ed79a3b30e26f708790ab54b2e352bb747449c1aa31bd7aee"
@@ -195,8 +195,9 @@
 "data/synthetic_full22_extension/results/summary/ranking_paper_5metric_scenarios.csv","10326","a194ec34aa349f7177f89c2ae2b328f9589e7919f003659c6a5f0497ad32e91e"
 "data/synthetic_full22_extension/results/summary/ranking_paper_5metric_tiers.csv","3511","06bd8784cd8dfba5c8ce1d15aa12b4b1ce800ae293e012e5d626257c6fcb679b"
 "data/synthetic_full22_extension/results/summary/report.md","12091","e7769cea54e0d0549add85f119aafdb11e933af5fd2cd09ec812f3639f75594a"
-"metadata/dataset_schema.json","3291","135ebf889b90ffc48f66082a3985ec44ef6c41aac8020e37b744a7a11860bfae"
-"metadata/release_manifest.json","2923","77fa7b899340e197683462b0813f26d8b26a703ef4b27d0054401ffaf597c632"
+"metadata/dataset_schema.json","3355","5002d24ca61975f0e0028f9ba224a3e395b4169c37c965970f6514589cd22634"
+"metadata/paper_table_provenance.md","4978","95a56db9e00d220f26eee2bc4e4355e4d93c522149179c6b84858c3c71c97b49"
+"metadata/release_manifest.json","3103","cff7af631e213dfa19fd939f2c770baf9ee53c80562f0dd7de4dc86fa7823853"
 "site_data/v1.0.0/evaluation_metrics.json","4609","5b134e900a2308ad5d3a8c61f72909935c5d61de82350210458c2e978140b1d8"
 "site_data/v1.0.0/leaderboard_post_rebuttal_real_physics_track_b.csv","2213","02d55c80e7d63e5b907d39789614931c3dc0dcc73d97feaa0d25145f08e17637"
 "site_data/v1.0.0/leaderboard_post_rebuttal_real_proxy_track_c.csv","8146","956471fc5e08dad8b9bd2e14defb4a0d9d7561a3714c94340d89760397a09fe6"

metadata/paper_table_provenance.md

@@ -0,0 +1,47 @@
+# Camera-Ready Table Provenance Audit
+
+This note records the current evidence boundary for the ICML 2026
+camera-ready tables and appendix summaries. It is an audit artifact for the
+release bundle.
+
+## Status Key
+
+- `exact-rerunnable`: the public code path regenerates the paper value.
+- `exact-from-summary`: the paper value is reproduced exactly by aggregating a
+  released or archived summary table.
+- `csv-exact`: the paper value matches the released camera-ready CSV table.
+- `source-partial`: the exact low-level raw rows or merge scripts are not fully
+  present in this checkout.
+- `extension-separate`: the file is a later benchmark extension and should not
+  overwrite the camera-ready paper snapshot.
+
+## Main Paper
+
+| Item | Paper value/source | Public artifact | Status | Notes |
+| --- | --- | --- | --- | --- |
+| Table 2 / Figure 3 core leaderboard | Six-family, 6-scenario, 50-draw benchmark | `data/paper_tables/paper_core_50draw_leaderboard.csv`, `paper_core_50draw_by_tier.csv`, `global_performance_summary.csv` | `exact-rerunnable` | Regenerated by `code/TSDecompose/scripts/run_paper_benchmark.py`. This is the primary one-command reproduction path. |
+| Scenario-level values cited in Section 4 | Values used in diagnostics paragraphs | `data/paper_tables/paper_core_50draw_leaderboard.csv` and tier/scenario aggregations | `exact-rerunnable` | Checked against the HF table and camera-ready manuscript. |
+| CO2 symbolic-regression table | Direct SR vs decompose-then-regress | Paper table and SR outputs | `exact-from-summary` | Runtime is longer than the core leaderboard: approximately 2h with ND2 or 8h with PySR defaults. |
+
+## Appendix C
+
+| Item | Paper table/claim | Public/rebuttal source | Status | Risk handling |
+| --- | --- | --- | --- | --- |
+| Table 4 bounded tuning | Default and tuned aggregate orders | `data/paper_tables/bounded_tuning_order.csv`; rebuttal prose; `rebuttal_experiments/reports/final_rebuttal_result_report.md`; non-SSA `e1_default_vs_tuned_summary.csv` | `csv-exact`, `source-partial` | The camera-ready text calls this a rebuttal-stage robustness-subset summary. It should not be presented as the one-command leaderboard. |
+| Table 5 period robustness | STL/MSTL true, FFT, and +20% rows | `data/paper_tables/period_robustness.csv`; archived `e3_period_summary.csv` | `exact-from-summary` | Aggregating STL/MSTL over the three configured period scenarios reproduces the table exactly after rounding. |
+| Table 6 alignment robustness | Injected, unsupervised, truth-guided suite scores | `data/paper_tables/alignment_robustness.csv`; archived `e2_alignment_summary.csv`; final rebuttal report for SSA row | `exact-from-summary`, `source-partial` | Non-SSA rows reproduce from `e2_alignment_summary.csv`; SSA row matches report/HF table, but complete merged SSA raw rows are not in this checkout. Caption frames rankings as protocol-dependent diagnostics. |
+| Table 7 boundary sensitivity | Mean absolute suite-score delta | `data/paper_tables/boundary_sensitivity.csv`; final rebuttal report; closest archived `e4_boundary_summary.csv` | `csv-exact`, `source-partial` | The exact table matches HF and the final report. Closest detailed source reproduces VMD/Wavelet closely but not all rounded values; text frames it as a sensitivity diagnostic, not a primary ranking metric. |
+| Table 8 real-data proxy panel | Band plausibility, resampling stability, spectrum overlap | `data/paper_tables/real_proxy_panel.csv`; `data/post_rebuttal_second_expansion/real_proxy_track_c/track_c_summary.csv` | `exact-from-summary` | Averaging the six real proxy datasets by method reproduces the table exactly after rounding. |
+| C.4 CO2/tides mechanism-aware claims | CO2 SSA 0.993/1.000; tides MSTL 0.956/1.000/1.000 | `data/post_rebuttal_second_expansion/real_physics_track_b/track_b_summary.csv` | `exact-from-summary` | Values match the source after rounding. |
+| C.4 semi-synthetic transfer paragraph | Six-family real-background transfer order and CEEMDAN diagnostics | `data/paper_tables/semisynthetic_transfer_summary.csv`; archived old rebuttal semi-synthetic raw files | `exact-from-summary`, `extension-separate` | This is the camera-ready paper snapshot. The separate 22-method semi-synthetic track is a later benchmark extension and may have different method-level ordering. |
+| MSSA pilot | Synthetic multi-sequence summary | `data/paper_tables/mssa_pilot_summary.csv` | `csv-exact` | Treat as initial synthetic extension evidence, not a complete real-panel benchmark. |
+
+## Practical Reproduction Boundary
+
+The release should be described as follows:
+
+> The public artifact provides a one-command reproduction path for the primary
+> six-family paper-core synthetic leaderboard, plus exact machine-readable CSV
+> snapshots for all camera-ready tables. Appendix robustness and transfer
+> checks are rebuttal-stage sensitivity summaries with available source
+> summaries and metadata; they do not replace the primary leaderboard.

metadata/release_manifest.json

@@ -42,6 +42,7 @@
   ],
   "alignment_notes": [
     "data/paper_tables contains exact CSV versions of camera-ready manuscript tables plus paper_core_50draw_leaderboard.csv and paper_core_50draw_by_tier.csv for direct Table 2 reproduction.",
+    "metadata/paper_table_provenance.md records which paper tables are one-command rerunnable, which are exact summary-table snapshots, and which are later benchmark extensions.",
     "data/paper_figures contains small paper figure assets for leaderboard visual alignment checks.",
     "site_data/v1.0.0/evaluation_metrics.json records the five paper-core component-recovery metrics, the Table 2 display subset, and the tier-balanced Table 2 aggregation rule.",
     "synthetic_full22_extension is a benchmark-only 22-method track using the same 6-scenario, 50-draw synthetic protocol where applicable; it should not replace the primary six-family Table 2 / Figure 3 interpretation.",