paper-claim-audit — скилл для Claude Code: сверка цифр в статье

---
name: paper-claim-audit
description: "Zero-context verification that every number, comparison, and scope claim in the paper matches raw result files. Uses a fresh cross-model reviewer with NO prior context to prevent confirmation bias. Use when user says \"审查论文数据\", \"check paper claims\", \"verify numbers\", \"论文数字核对\", or before submission to ensure paper-to-evidence fidelity."
argument-hint: "[paper-directory]"
allowed-tools: Bash(*), Read, Write, Edit, Grep, Glob, Skill, mcp__codex__codex, mcp__antigravity__antigravity, mcp__antigravity__antigravity-reply, mcp__grok__grok, mcp__grok__grok-reply, mcp__claude-review__review, mcp__claude-review__review_reply, mcp__claude-review__review_start, mcp__claude-review__review_reply_start, mcp__claude-review__review_status
---

# Paper Claim Audit: Zero-Context Evidence Verification

> **Before each reviewer slot**, follow [Selectable Reviewers](../shared-references/reviewer-backends.md):
> resolve invocation → `.aris/reviewer.json` → Claude Code default (Codex MCP)
> with `reviewer_route.py` through the canonical helper resolver, preserving this
> slot's tier, fresh contexts, source/image inputs and deterministic gates.
> Codex examples/pins below apply only to `backend=codex`; other routes use their
> own first-call/reply contract. Save the route and actual returned identity;
> unavailable selections are `REVIEW_UNAVAILABLE`. Classify acceptance from actual
> families/completion/scope. Forward invocation reviewer/model/effort overrides
> to reviewer-bearing children. Generation workers are unchanged.


> 🔒 **Do not wrap this skill in `/loop`, `/schedule`, or `CronCreate`.** It is
> verdict-bearing — it judges paper-to-evidence fidelity with a deliberately
> zero-context fresh reviewer. Re-firing that verdict on a wall-clock timer adds
> no new signal (it changes only when the *paper or results* change). Schedule
> the *external wait that precedes it* — paper draft ready → then audit
> **once**. See
> [`shared-references/external-cadence.md`](../shared-references/external-cadence.md).

Verify that every claim in the paper matches raw evidence for: **$ARGUMENTS**

## Why This Exists

The executor writes experiments AND writes the paper. It "knows" what the results should be. This creates confirmation bias:
- Rounding 84.7% up to 85.3%
- Reporting best seed instead of average
- Citing metrics from a different experiment config
- Claiming "improves by 15%" when the delta is actually 12.8%

A **fresh reviewer with zero prior context** catches these because it has no expectations — it just compares paper text vs raw files.

## How This Differs From Other Audit Skills

| Skill | Question it answers |
|-------|-------------------|
| `/experiment-audit` | Is the experiment code honest? (fake GT, normalization fraud) |
| `/result-to-claim` | Does the data scientifically support this claim? |
| **`/paper-claim-audit`** | **Does the paper report the data truthfully and precisely?** |

## Core Principle

**Zero-context, fresh reviewer.** The auditor receives ONLY:
- Paper .tex files (the claims)
- Raw result files (the evidence)

It does NOT receive:
- ❌ EXPERIMENT_LOG.md
- ❌ EXPERIMENT_TRACKER.md
- ❌ AUTO_REVIEW.md
- ❌ NARRATIVE_REPORT.md
- ❌ Any executor summary or interpretation
- ❌ Any prior audit results
- ❌ Any conversation history

This is **stricter than reviewer-independence** — it's zero-context evidence audit.

## Workflow

### Step 1: Collect Files (Executor — Claude)

Locate paper and result files WITHOUT reading or interpreting them.

**Paper files** (claims) — paths shown relative to the shell's working
directory so you can find them with `ls`; when writing them into
`audited_input_hashes`, use paths relative to the paper dir (no `paper/`
prefix) per the "Submission Artifact Emission" section below:
```
paper/main.tex                # → hash key: main.tex
paper/sections/*.tex          # → hash key: sections/*.tex
paper/tables/*.tex (if separate)   # → hash key: tables/*.tex
```

**Result files** (evidence):
```
results/*.json, results/*.jsonl, results/*.csv, results/*.tsv
outputs/*.json, outputs/*.csv
wandb-summary.json (if exists)
**/metrics.json, **/eval_results.json
**/config.yaml, **/args.json (experiment configs)
```

**Exclude** (no summaries, no interpretations):
```
EXPERIMENT_LOG.md, EXPERIMENT_TRACKER.md, AUTO_REVIEW*.md
NARRATIVE_REPORT.md, PAPER_PLAN.md, findings.md
Any .md file that is an executor-written summary
```

### Step 2: Fresh Reviewer Audit (GPT-6-Astra — NEW thread, no reply)

**CRITICAL: Use `mcp__codex__codex` (new thread), NEVER `mcp__codex__codex-reply`.** Every run must be a fresh context.

> Resolve this reviewer slot first; the following is the `backend=codex` example.

```
mcp__codex__codex:
  model: gpt-6-astra
  config: {"model_reasoning_effort": "ultra"}
  prompt: |
    You are a paper-to-evidence auditor. You have ZERO prior context about
    this research. You will receive only paper source files and raw result
    files. Your job is to verify that every number in the paper exactly
    matches the raw evidence.

    Paper files to read:
    [list .tex file paths]

    Result files to read:
    [list .json/.csv/.yaml file paths]

    ## Audit Protocol

    ### A. Extract Every Quantitative Claim
    For each number, percentage, comparison, or scope statement in the paper:
    - Location (section, table, caption, or inline text)
    - Exact claim text
    - The number or comparison being made

    ### B. Trace Each Claim to Evidence
    For each extracted claim, find the supporting raw data:
    - Which result file contains this number?
    - What is the EXACT value in that file?
    - Match status: exact_match / rounding_ok / mismatch

    ### C. Check These Specific Failure Modes

    1. **Number inflation**: Paper says 85.3%, raw file says 84.7%
       Rule: only standard rounding to displayed precision is allowed

    2. **Best-seed cherry-pick**: Paper says "achieves 90.2%" but
       that's the best of 5 seeds; mean is 87.1%
       Rule: check if paper specifies "average" / "best" / "median"

    3. **Config mismatch**: Paper compares Method A vs Baseline B,
       but they used different hyperparameters / datasets / splits
       Rule: verify config files show same settings for compared methods

    4. **Aggregation mismatch**: Paper says "average over 5 seeds"
       but result files show only 3 runs
       Rule: count actual runs vs claimed count

    5. **Delta error**: Paper says "improves by 15%" but
       actual delta is (85.3 - 73.1) / 73.1 = 16.7%
       Rule: verify arithmetic of all relative improvements

    6. **Caption-table mismatch**: Figure caption describes
       something different from what the figure/table actually shows
       Rule: cross-check every caption against its content

    7. **Scope overclaim**: Paper says "consistently outperforms"
       but only tested on 2 datasets
       Rule: check if language matches actual evaluation scope

    ## Output Format (per claim)
    For each claim, report:
    - claim_id: sequential number
    - location: section/table/figure
    - paper_text: exact quote from paper
    - paper_value: the number claimed
    - evidence_file: which raw file
    - evidence_value: the actual number
    - status: exact_match | rounding_ok | ambiguous_mapping |
              missing_evidence | config_mismatch | aggregation_mismatch |
              number_mismatch | scope_overclaim | unsupported_claim
    - details: explanation if not exact_match

    Overall verdict: PASS | WARN | FAIL
```

### Step 3: Write Report (Executor — Claude)

Parse the reviewer's response and write `PAPER_CLAIM_AUDIT.md`:

```markdown
# Paper Claim Audit Report

**Date**: [today]
**Auditor**: GPT-6-Astra ultra (fresh zero-context thread)
**Paper**: [paper title from tex]

## Overall Verdict: [PASS | WARN | FAIL]

## Claims Verified: [N total]
- exact_match: [count]
- rounding_ok: [count]
- ambiguous_mapping: [count]
- missing_evidence: [count]
- mismatch: [count]

## Issues Found

### [FAIL/WARN] Claim #N: [description]
- **Location**: Section X / Table Y / Figure Z
- **Paper says**: "..."
- **Evidence shows**: ...
- **Status**: [status]
- **Fix**: [specific correction needed]

## All Claims (detailed)

| # | Location | Paper Value | Evidence Value | Status |
|---|----------|-------------|---------------|--------|
| 1 | Table 2 | 85.3% | 85.28% | rounding_ok |
| 2 | Abstract | "15% improvement" | 12.8% | number_mismatch |
| ... |
```

Also write `PAPER_CLAIM_AUDIT.json` for machine consumption.

### Step 4: Print Summary

```
📋 Paper Claim Audit Complete

  Claims verified: 24
  exact_match:     18
  rounding_ok:      3
  ambiguous:         1
  ⚠️ mismatch:      2

  Overall: ⚠️ WARN

  See PAPER_CLAIM_AUDIT.md for details.
```

## When to Run

1. **After `/paper-write`** — first check before improvement loop
2. **After `/auto-paper-improvement-loop`** — recheck if improvement loop changed numbers
3. **Before submission** — final verification

## Integration with Other Skills

### Read by `/auto-paper-improvement-loop` (if exists)

```
if PAPER_CLAIM_AUDIT.json exists:
    read mismatched claims
    fix them as priority items in the improvement round
```

### Advisory, Never Blocking

Same pattern as `/experiment-audit`:
- `PASS` → continue normally
- `WARN` → print warning, continue, flag draft as "check numbers before submission"
- `FAIL` → print alert, continue, but do NOT mark as submission-ready

## Render HTML view (auto, when `RENDER_HTML = true`, default)

After writing `paper/PAPER_CLAIM_AUDIT.md` and `paper/PAPER_CLAIM_AUDIT.json`, invoke `/render-html` on the audit report so the user has a readable HTML view of the verdict + per-claim breakdown:

```
/render-html "paper/PAPER_CLAIM_AUDIT.md" --json "paper/PAPER_CLAIM_AUDIT.json"
```

Uses **full Codex review gate** (audit-class artifact — render-fidelity check matches the skill's existing zero-context cross-model audit invariant). Output lands at `paper/PAPER_CLAIM_AUDIT.html` with embedded source SHA256 and a `.review.json` sidecar carrying the render verdict.

**Non-blocking**: if `/render-html` fails (helper missing, Codex MCP unavailable, file write error), log the failure and treat the skill as complete — the JSON + MD verdict files are the canonical outputs; the HTML view is a convenience for human readers.

Skip if `RENDER_HTML = false` is set in the project's `CLAUDE.md` or passed as `— render html: false`.

## Key Rules

- **Fresh thread EVERY run.** Never use `codex-reply`. Never carry context.
- **Zero executor interpretation.** Only file paths. No summaries.
- **Only raw results.** No EXPERIMENT_LOG, no AUTO_REVIEW, no human summaries.
- **Rounding rule.** Only standard rounding to displayed precision. 84.7% → 84.7% or 85% is OK. 84.7% → 85.3% is NOT OK.
- **Cross-model.** Reviewer must be a different model family from executor.

## Review Tracing

After each `mcp__codex__codex` or `mcp__codex__codex-reply` reviewer call, save the trace following `shared-references/review-tracing.md` (Policy C — forensic; never silently skip). Use `save_trace.sh` (resolved per the chain in `shared-references/integration-contract.md` §2) or write files directly to `.aris/traces/<skill>/<date>_run<NN>/`. Respect the `--- trace:` parameter (default: `full`).

## Submission Artifact Emission

This skill **always** writes `paper/PAPER_CLAIM_AUDIT.json`, regardless of
caller or detector outcome. A detector-negative run (paper has no numeric
claims) emits verdict `NOT_APPLICABLE`; a paper-with-numeric-claims-but-no-
raw-results run emits `BLOCKED`. Silent skip is forbidden — `paper-writing`
Phase 6 and `verify_paper_audits.sh` both rely on this artifact
existing at a predictable path.

The artifact conforms to the schema in `shared-references/assurance-contract.md`:

```json
{
  "audit_skill":      "paper-claim-audit",
  "verdict":          "PASS | WARN | FAIL | NOT_APPLICABLE | BLOCKED | ERROR",
  "reason_code":      "all_numbers_match | rounding_drift | missing_raw_results | ...",
  "summary":          "One-line human-readable verdict summary.",
  "audited_input_hashes": {
    "main.tex":                              "sha256:...",
    "sections/5.evidence.tex":               "sha256:...",
    "/abs/path/to/results/run_2026_04_19.json": "sha256:..."
  },
  "trace_path":       ".aris/traces/paper-claim-audit/<date>_run<NN>/",
  "thread_id":        "<codex mcp thread id>",
  "reviewer_model":   "<resolved — the model that actually ran (target: gpt-6-astra)>",
  "reviewer_reasoning": "<resolved — the effort that actually ran (target: ultra)>",
  "generated_at":     "<UTC ISO-8601>",
  "details": {
    "total_claims":   <int>,
    "mismatches":     [ ... per-claim issue records ... ],
    "result_files":   [ ... raw files consulted ... ]
  }
}
```

### `audited_input_hashes` scope

Hash the **declared input set** passed into this audit invocation — i.e. the
exact `.tex` files and raw result / config files this run read — not a
repo-wide union and not the reviewer's self-reported subset. If a caller
passed only `main.tex` + a single result file, hash those two files and no
others. The external verifier rehashes these entries; any mismatch flags
`STALE`.

**Path convention** (must match what `verify_paper_audits.sh`
expects): keys are **paths relative to the paper directory** (the arg
passed to the verifier) for in-paper files — so `main.tex`, not
`paper/main.tex` — and **absolute paths** for out-of-paper files such as
external `results/` dirs. The verifier resolves relative entries via
`os.path.join(paper_dir, key)`; prefixing with `paper/` produces
`paper/paper/main.tex` and false-fails as STALE.

### Verdict decision table

| Input state                                           | Verdict          | `reason_code` example |
|-------------------------------------------------------|------------------|-----------------------|
| No numeric claims detected in paper                   | `NOT_APPLICABLE` | `no_numeric_claims`   |
| Numeric claims detected, no raw result files found    | `BLOCKED`        | `no_raw_evidence`     |
| All claims reconcile to raw data                      | `PASS`           | `all_numbers_match`   |
| Minor rounding drift only, no material mismatch       | `WARN`           | `rounding_drift`      |
| Any material mismatch (wrong number, config mismatch) | `FAIL`           | `claim_mismatch`      |
| Reviewer invocation failed (network / malformed)      | `ERROR`          | `reviewer_error`      |

### Thread independence

Every invocation uses a fresh `mcp__codex__codex` thread. Never
`codex-reply`. Do not accept prior audit outputs (PROOF_AUDIT, CITATION_AUDIT,
EXPERIMENT_LOG, AUTO_REVIEW summaries) as input to this audit — the fresh
thread preserves reviewer independence per
`shared-references/reviewer-independence.md`.

### Human-readable sibling

`paper/PAPER_CLAIM_AUDIT.md` is written alongside the JSON for readers.
The JSON is authoritative for `verify_paper_audits.sh`; the Markdown
is for humans. The parent skill (`paper-writing` Phase 6) plus the verifier
decide whether the verdict blocks finalization — this skill itself never
blocks; it only emits.