Paper Sanity Check

Report format: List each gap as [Section X.Y → X.Z] <description of the disconnect>.

Check 2 · Float Reference Completeness

• Enumerate every float object in the paper: figures, tables, algorithms, listings, pseudocode blocks.
• For each float, search the body text for at least one explicit reference (e.g., Figure 1, Table 2, Algorithm 3).
• Flag any float that is never referenced in the running text.
• Flag any in-text reference that points to a non-existent float (dangling reference).
• Check that references appear before or near the float's placement — a figure referenced only in the appendix but placed in the main body is suspicious.

Report format: Table with columns [Float ID | Referenced? | Location of first reference | Issue].

Check 3 · Contribution–Evidence Alignment

• Extract the explicit contribution claims from the Introduction (usually a bulleted list or numbered points).
• For each claim, locate the corresponding experimental evidence (tables, figures, ablations).
• Grade alignment:
  • Supported: Clear evidence directly validates the claim.
  • Overstated: Claim uses strong language ("significant", "substantial", "dramatically") but evidence shows marginal improvement (e.g., <1% gain presented as a breakthrough).
  • Unsupported: No experiment validates this claim.
• Pay special attention to the gap between the magnitude of the language and the magnitude of the numbers.

Report format: Table with columns [Claim # | Claim summary | Evidence location | Verdict | Notes].

Check 4 · Data–Analysis Consistency

• For each experimental result discussed in the text, verify that:
  • The numbers cited in prose match the numbers in the corresponding table/figure.
  • The comparison direction is correct (e.g., "our method outperforms X by 3%" — verify it's indeed 3%, not 2.7% or that X isn't actually better).
  • The aspect being discussed matches what the table/figure actually measures (e.g., text discusses accuracy but the table shows F1).
• Flag any mismatch, even minor rounding discrepancies (>0.1 absolute difference).

Report format: [Section X.Y, paragraph Z] Text says "<quote>" but Table/Figure N shows <actual value>.

Check 5 · Cross-Table Data Consistency

• Identify every unique (model, dataset, metric) tuple that appears in more than one table or figure.
• Verify that the same tuple reports the same value everywhere, unless:
  • Different hyperparameters/settings are explicitly stated, OR
  • One is a subset/superset experiment.
• Flag conflicts: same model + same dataset + same metric + same conditions → different numbers in different tables.

Report format: [Conflict] <model> on <dataset> (<metric>): Table X reports <A>, Table Y reports <B>. No stated difference in conditions.

Check 6 · Causal Coherence & Persuasiveness

6a · Motivation & Problem-Framing Causation

• In the Abstract and Introduction, identify every causal claim of the form "A causes / introduces / gives rise to B".
• For each such claim, verify: Is B genuinely caused by A, or is it a general issue (e.g., inherent to low-bit quantization, common across all model families) that the authors have incorrectly attributed to a domain-specific factor?
• Pay special attention to blanket attributions that tie multiple challenges to a single architectural or domain-specific cause. Check each challenge individually.

6b · Experimental Result Causation

• For each key experimental result, check:
  • Does the paper explain why the result occurs, not just what the result is?
  • Is the causal explanation consistent with the method design?
  • Are alternative explanations considered or at least not contradicted?
• Flag results that are presented without explanation or with explanations that contradict the method description.
• Flag cherry-picked results: e.g., the paper highlights one favorable metric while ignoring degradation in another metric shown in the same table.

Report format: [Section X.Y] Result: <what>. Issue: <missing causation / contradictory explanation / cherry-picking>.

Output Format

# 论文基础检查报告

## 总览
- 发现问题总数：N
- 严重（必须修复）：N
- 警告（建议修复）：N

## 检查 1：逻辑流与衔接
[发现 或 "通过 — 未发现问题"]

## 检查 2：浮动体引用完整性
[发现 或 "通过 — 全部 N 个浮动体均已引用"]

## 检查 3：贡献-证据对齐
[发现 或 "通过 — 所有声明均有支撑"]

## 检查 4：数据-分析一致性
[发现 或 "通过 — 正文数字与表图一致"]

## 检查 5：跨表数据一致性
[发现 或 "通过 — 无跨表冲突"]

## 检查 6：因果连贯性与说服力
[发现 或 "通过 — 因果解释一致"]

Constraints

• Do NOT comment on writing style, grammar, or word choice — that is the job of other skills (paper-polish, paper-logic-check).
• Do NOT suggest adding new experiments or changing the method — that is the job of paper-review.
• Only report issues you can verify from the text. Do not speculate about what the authors "might have meant".
• When a check passes cleanly, explicitly mark it 通过 — silence is not the same as approval.
• Output the report in Chinese (中文).