File size: 16,666 Bytes
ecb8ba4 2ad8b3a ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 4ae31f8 ecb8ba4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 | #!/usr/bin/env python
from __future__ import annotations
import argparse
import json
import re
import subprocess
import sys
from pathlib import Path
from typing import Any
FORBIDDEN_PATTERNS: tuple[tuple[str, str, str], ...] = (
(
"submission_status_language",
r"\bA\*|\bQ1\b|Best Paper",
"Submission draft must not contain internal venue/readiness language.",
),
(
"future_work_should",
r"future work should",
"Main method text must not defer the method to future work.",
),
(
"generic_q_phi_generator",
r"q_phi|q_\s*\\phi|q_\s*\{\\phi\}|q\^\s*\\phi",
"The main generator must be CTT transport, not a generic q_phi noise model.",
),
(
"distance_proxy_called_ptr",
r"(PTR[^.\n]{0,80}(distance|proxy))|((distance|proxy)[^.\n]{0,80}PTR)",
"Distance-only support diagnostics must be called PPTC, not PTR.",
),
)
REQUIRED_PHRASES: tuple[tuple[str, str, str], ...] = (
(
"title_ctt",
"Causal Tangent Transport",
"Paper title/introduction names CTT as the central method.",
),
(
"transport_operator",
"T_{\\phi}(z_s,z_t,\\xi_s^+)",
"Paper defines train-positive source-to-target transport.",
),
(
"outcome_ptr",
"OutcomePTR",
"Paper names measured rollout positive-tangent recall separately.",
),
(
"pptc",
"PPTC",
"Paper names distance-only Proxy Positive Tangent Coverage.",
),
(
"support_gap",
"SupportGap",
"Paper exposes the support part of CAR.",
),
(
"selector_gap",
"SelectorGap",
"Paper exposes the selector part of CAR.",
),
)
REQUIRED_PATHS: tuple[tuple[str, str, str], ...] = (
("metrics_module", "cil/metrics.py", "Canonical measured/proxy metrics."),
("metrics_eval", "scripts/eval_metrics.py", "JSON/TeX metric export and proxy guards."),
("metrics_tests", "tests/test_metrics.py", "Regression tests for metric separation."),
("chart_export", "scripts/export_cil_charts.py", "Chart database export."),
("chart_audit", "scripts/audit_cil_charts.py", "Chart leakage audit."),
("data_accounting", "runs/data_accounting/table.json", "Scripted data accounting."),
("data_accounting_table", "runs/data_accounting/table.tex", "Data accounting table."),
("leakage_audit", "runs/leakage_audit/report.json", "Leakage audit artifact."),
("ctt_model", "cil/models/ctt.py", "Causal Tangent Transport module."),
("tangent_encoder", "cil/models/tangent_encoder.py", "Tangent-code encoder/decoder helpers."),
("chart_encoder", "cil/models/chart_encoder.py", "Chart encoder."),
("utility_energy", "cil/models/utility_energy.py", "Utility energy scorer."),
("ctt_train", "scripts/train_ctt.py", "CTT training script."),
("ctt_proxy_eval", "scripts/eval_ctt_proxy.py", "Proxy support evaluation."),
("ctt_rollout_eval", "scripts/eval_ctt_rollout.py", "Measured rollout evaluation."),
("utility_train", "scripts/train_utility_energy.py", "Utility energy training."),
("dominance_calibration", "scripts/calibrate_dominance.py", "Calibrated dominance rule."),
("selector_diagnostic_sweep", "scripts/build_selector_diagnostic_sweep.py", "Selector diagnostic sweep summary."),
("theory_tex", "paper/sections/theory.tex", "Theory section included by paper."),
("paper_pdf", "latex/main.pdf", "Compiled paper PDF."),
)
REQUIRED_RUN_FILES: tuple[str, ...] = (
"table.tex",
"metrics.json",
"command.txt",
"git_hash.txt",
"data_hash.txt",
"split_hash.txt",
)
ADVISOR_RUN_FILES: tuple[str, ...] = (
"config.yaml",
"train.log",
"eval.log",
"metrics_by_task.json",
"metrics_by_seed.json",
)
INPUT_RE = re.compile(r"\\input\{([^}]+)\}")
ALLOWED_MARKDOWN_FILES: tuple[str, ...] = ("README.md",)
def main(argv: list[str] | None = None) -> int:
parser = argparse.ArgumentParser(
description=(
"Audit the CTT paper against the claim-to-artifact contract. "
"Outputs JSON and TeX so the repo can keep README.md as the only Markdown file."
)
)
parser.add_argument("--repo-root", type=Path, default=Path.cwd())
parser.add_argument("--paper", type=Path, default=Path("latex/main.tex"))
parser.add_argument("--out-dir", type=Path, default=Path("runs/paper_ctt_audit"))
parser.add_argument(
"--skip-implementation-checks",
action="store_true",
help="Only scan paper text and input artifacts; useful for isolated unit tests.",
)
args = parser.parse_args(argv)
repo_root = args.repo_root.resolve()
paper_path = _resolve(repo_root, args.paper)
if not paper_path.exists():
raise SystemExit(f"paper not found: {paper_path}")
paper_text = paper_path.read_text()
forbidden = _forbidden_findings(paper_text)
phrase_checks = _phrase_checks(paper_text)
paper_inputs = _paper_input_checks(repo_root, paper_path, paper_text)
required_paths = [] if args.skip_implementation_checks else _required_path_checks(repo_root)
run_artifacts = _run_artifact_checks(repo_root, paper_inputs["run_dirs"])
markdown_policy = _markdown_policy_checks(repo_root)
summary = _summary(
forbidden,
phrase_checks,
paper_inputs,
required_paths,
run_artifacts,
markdown_policy,
)
payload = {
"schema_version": 1,
"audit_policy": {
"markdown_policy": "consolidated_readme_only",
"markdown_note": (
"Advisor checklist asks for report.md files, but the current workspace "
"policy keeps README.md as the only Markdown document. This audit now "
"fails on additional Markdown files instead of regenerating report.md."
),
"allowed_markdown_files": list(ALLOWED_MARKDOWN_FILES),
},
"paper": str(paper_path.relative_to(repo_root)),
"summary": summary,
"forbidden_patterns": forbidden,
"required_phrases": phrase_checks,
"paper_inputs": paper_inputs,
"required_paths": required_paths,
"run_artifacts": run_artifacts,
"markdown_policy": markdown_policy,
}
out_dir = _resolve(repo_root, args.out_dir)
out_dir.mkdir(parents=True, exist_ok=True)
(out_dir / "audit.json").write_text(json.dumps(payload, indent=2, sort_keys=True) + "\n")
(out_dir / "metrics.json").write_text(json.dumps(payload, indent=2, sort_keys=True) + "\n")
(out_dir / "table.tex").write_text(_latex_table(payload) + "\n")
(out_dir / "config.yaml").write_text(
"\n".join(
[
"paper: " + str(paper_path.relative_to(repo_root)),
"markdown_policy: consolidated_readme_only",
"strict_report_md: false",
"allowed_markdown_files:",
*[f" - {name}" for name in ALLOWED_MARKDOWN_FILES],
]
)
+ "\n"
)
(out_dir / "command.txt").write_text(
"python scripts/audit_ctt_paper_artifacts.py " + " ".join(sys.argv[1:]) + "\n"
)
(out_dir / "git_hash.txt").write_text(_git_hash(repo_root) + "\n")
(out_dir / "data_hash.txt").write_text(_first_existing_hash(repo_root, "runs/data_accounting/table.json") + "\n")
(out_dir / "split_hash.txt").write_text(_split_hash(repo_root) + "\n")
print(json.dumps({"out_dir": str(out_dir), **summary}, indent=2, sort_keys=True))
return 1 if summary["status"] == "fail" else 0
def _forbidden_findings(text: str) -> list[dict[str, Any]]:
findings: list[dict[str, Any]] = []
for name, pattern, detail in FORBIDDEN_PATTERNS:
for match in re.finditer(pattern, text, flags=re.IGNORECASE):
findings.append(
{
"name": name,
"status": "fail",
"detail": detail,
"match": match.group(0),
"line": _line_for_offset(text, match.start()),
}
)
return findings
def _phrase_checks(text: str) -> list[dict[str, Any]]:
checks = []
for name, phrase, detail in REQUIRED_PHRASES:
checks.append(
{
"name": name,
"status": "pass" if phrase in text else "fail",
"detail": detail,
"needle": phrase,
}
)
return checks
def _paper_input_checks(repo_root: Path, paper_path: Path, text: str) -> dict[str, Any]:
rows = []
run_dirs: list[str] = []
for match in INPUT_RE.finditer(text):
raw = match.group(1)
resolved = _resolve_tex_input(paper_path.parent, raw)
status = "pass" if resolved.exists() else "fail"
row = {
"raw": raw,
"line": _line_for_offset(text, match.start()),
"resolved": _display_path(repo_root, resolved),
"status": status,
}
run_dir = _run_dir_for_input(repo_root, resolved)
if run_dir is not None:
row["run_dir"] = _display_path(repo_root, run_dir)
if row["run_dir"] not in run_dirs:
run_dirs.append(row["run_dir"])
rows.append(row)
return {
"num_inputs": len(rows),
"num_missing": sum(1 for row in rows if row["status"] != "pass"),
"rows": rows,
"run_dirs": sorted(run_dirs),
}
def _required_path_checks(repo_root: Path) -> list[dict[str, Any]]:
checks = []
for name, rel_path, detail in REQUIRED_PATHS:
path = repo_root / rel_path
checks.append(
{
"name": name,
"path": rel_path,
"status": "pass" if path.exists() else "fail",
"detail": detail,
}
)
ctt_configs = sorted((repo_root / "configs/ctt").glob("*.yaml"))
checks.append(
{
"name": "ctt_configs",
"path": "configs/ctt/*.yaml",
"status": "pass" if ctt_configs else "fail",
"detail": "CTT loss weights and variants are config-driven.",
"files": [_display_path(repo_root, path) for path in ctt_configs],
}
)
return checks
def _run_artifact_checks(repo_root: Path, run_dirs: list[str]) -> list[dict[str, Any]]:
rows = []
for rel_run_dir in sorted(run_dirs):
run_dir = repo_root / rel_run_dir
required = list(REQUIRED_RUN_FILES)
if rel_run_dir == "runs/data_accounting":
required = ["table.json", "table.tex"]
missing_required = [
name
for name in required
if not (run_dir / name).exists()
and not (rel_run_dir == "runs/paper_ctt_audit" and name == "metrics.json")
]
missing_advisor = [name for name in ADVISOR_RUN_FILES if not (run_dir / name).exists()]
status = "pass" if not missing_required else "fail"
rows.append(
{
"run_dir": rel_run_dir,
"status": status,
"missing_required": missing_required,
"missing_advisor_contract": missing_advisor,
"markdown_report_policy": "consolidated_readme_only",
}
)
return rows
def _markdown_policy_checks(repo_root: Path) -> dict[str, Any]:
ignored_parts = {".git", ".venv"}
rows = []
for path in sorted(repo_root.rglob("*.md")):
try:
rel = path.relative_to(repo_root)
except ValueError:
continue
if any(part in ignored_parts for part in rel.parts):
continue
rel_text = rel.as_posix()
rows.append(
{
"path": rel_text,
"status": "pass" if rel_text in ALLOWED_MARKDOWN_FILES else "fail",
}
)
unexpected = [row["path"] for row in rows if row["status"] != "pass"]
missing_allowed = [
name for name in ALLOWED_MARKDOWN_FILES if not (repo_root / name).exists()
]
return {
"policy": "consolidated_readme_only",
"allowed": list(ALLOWED_MARKDOWN_FILES),
"rows": rows,
"unexpected_markdown": unexpected,
"missing_allowed": missing_allowed,
"num_markdown_files": len(rows),
"num_unexpected_markdown": len(unexpected),
"status": "pass" if not unexpected and not missing_allowed else "fail",
}
def _summary(
forbidden: list[dict[str, Any]],
phrase_checks: list[dict[str, Any]],
paper_inputs: dict[str, Any],
required_paths: list[dict[str, Any]],
run_artifacts: list[dict[str, Any]],
markdown_policy: dict[str, Any],
) -> dict[str, Any]:
failures = (
len(forbidden)
+ paper_inputs["num_missing"]
+ sum(1 for item in phrase_checks if item["status"] == "fail")
+ sum(1 for item in required_paths if item["status"] == "fail")
+ sum(1 for item in run_artifacts if item["status"] == "fail")
+ (0 if markdown_policy["status"] == "pass" else 1)
)
warnings = sum(len(item.get("missing_advisor_contract", [])) for item in run_artifacts)
return {
"status": "fail" if failures else "pass",
"num_failures": failures,
"num_warnings": warnings,
"num_forbidden_matches": len(forbidden),
"num_unexpected_markdown": markdown_policy["num_unexpected_markdown"],
"num_paper_inputs": paper_inputs["num_inputs"],
"num_run_dirs_in_paper": len(paper_inputs["run_dirs"]),
}
def _latex_table(payload: dict[str, Any]) -> str:
summary = payload["summary"]
lines = [
"% Auto-generated by scripts/audit_ctt_paper_artifacts.py",
"\\begin{tabular}{lrr}",
"\\toprule",
"Audit item & Count & Status \\\\",
"\\midrule",
f"Forbidden language matches & {summary['num_forbidden_matches']} & {_latex_status(summary['num_forbidden_matches'] == 0)} \\\\",
f"Paper inputs & {summary['num_paper_inputs']} & {_latex_status(payload['paper_inputs']['num_missing'] == 0)} \\\\",
f"Run dirs in paper & {summary['num_run_dirs_in_paper']} & {_latex_status(all(row['status'] == 'pass' for row in payload['run_artifacts']))} \\\\",
f"Implementation paths & {len(payload['required_paths'])} & {_latex_status(all(row['status'] == 'pass' for row in payload['required_paths']))} \\\\",
f"Unexpected Markdown files & {summary['num_unexpected_markdown']} & {_latex_status(summary['num_unexpected_markdown'] == 0)} \\\\",
f"Advisor-contract warnings & {summary['num_warnings']} & {'pass' if summary['num_warnings'] == 0 else 'warn'} \\\\",
"\\bottomrule",
"\\end{tabular}",
]
return "\n".join(lines)
def _resolve(repo_root: Path, path: Path) -> Path:
return path if path.is_absolute() else repo_root / path
def _resolve_tex_input(base_dir: Path, raw: str) -> Path:
candidate = (base_dir / raw).resolve()
if candidate.exists():
return candidate
if candidate.suffix:
return candidate
return candidate.with_suffix(".tex")
def _run_dir_for_input(repo_root: Path, resolved: Path) -> Path | None:
try:
rel = resolved.relative_to(repo_root / "runs")
except ValueError:
return None
if not rel.parts:
return None
return repo_root / "runs" / rel.parts[0]
def _display_path(repo_root: Path, path: Path) -> str:
try:
return str(path.relative_to(repo_root))
except ValueError:
return str(path)
def _line_for_offset(text: str, offset: int) -> int:
return text.count("\n", 0, offset) + 1
def _git_hash(repo_root: Path) -> str:
try:
return subprocess.check_output(
["git", "rev-parse", "HEAD"],
cwd=repo_root,
text=True,
stderr=subprocess.DEVNULL,
).strip()
except Exception:
return "unknown"
def _first_existing_hash(repo_root: Path, rel_path: str) -> str:
path = repo_root / rel_path
if not path.exists():
return ""
import hashlib
return hashlib.sha256(path.read_bytes()).hexdigest()
def _split_hash(repo_root: Path) -> str:
path = repo_root / "runs/data_accounting/table.json"
if not path.exists():
return ""
try:
payload = json.loads(path.read_text())
except json.JSONDecodeError:
return ""
return str(payload.get("split_hash", ""))
def _latex_status(ok: bool) -> str:
return "pass" if ok else "fail"
if __name__ == "__main__":
raise SystemExit(main())
|