paper-verificationlisted

# Paper Verification Methodology ⚠ 本 skill 属 verdict 类评审，必须遵守 `../cross-model-review` 协议：在零上下文新线程（fresh thread）、优先跨模型的条件下 执行核验，只读被核验的论文/代码/结果文件本身，禁止在刚写完或刚修改论文的同一 上下文里核验自己的数字。无法换线程/换模型时，按 `shared-references/reviewer-adapter.md` 降级并在报告头部标注 `⚠ same-context review, findings may be incomplete`。 同时遵守 `shared-references/research-integrity.md` 红线。 You are helping a researcher verify that their paper accurately reflects their code and experimental results. This is the most critical quality control step in academic writing. ## Verification Dimensions ### 1. Numerical Accuracy Audit For every number in the paper (dataset sizes, metric values, percentages, counts): 1. **Extract** the number and its context from the .tex file 2. **Trace** it to its source: code output, result file, log, or tracking system 3. **Verify** the value matches exactly (watch for rounding, percentage vs decimal) 4. **Flag** any number that cannot be traced to a source Template: ``` | Paper claim | Location (.tex) | Source file/code | Source value | Match? | |-------------|-----------------|-----------------|-------------|--------| | "13,999 frames" | abstract L3 | len(glob(labels/*.json)) | ? | ? | | "4.2% improvement" | Table 2 | eval_results.json | ? | ? | ``` Common numerical errors: - Rounding inconsistencies (3.14 in text, 3.1415 in table) - Stale numbers from earlier experiments not updated after re-runs - Percentage vs absolute confusion - Off-by-one in dataset counts (headers counted, or not) ###