Claim Ledger Example

A claim ledger keeps strong statements tied to evidence. It is useful for papers, literature reviews, rebuttals, and thesis chapters.

Example Ledger

Claim	Evidence	Status	Risk	Action
The proposed method reduces annotation time on dataset X.	Completed experiment log `exp-2026-07-01`, Table 2 draft	Completed	Need independent run check	Add run seed and confidence interval
Prior work usually evaluates on small benchmark Y.	Papers A, B, C in literature matrix	Verified	Paper D may be a larger benchmark	Read Paper D before final wording
The system supports real-time inference.	Local benchmark plan only	Planned	No completed timing evidence	Rephrase as target, not result
Reviewer concern R2 is addressed by an ablation.	Ablation not complete	Missing	Rebuttal promise would overclaim	Either run ablation or state limitation

Strength	Use When
Strong	Completed evidence directly supports the statement
Qualified	Evidence supports a narrower version
Planned	Work is intended but not yet evidence
Remove	Evidence is missing or contradicted

Extract every strong claim from the draft.
Assign one evidence source or mark it missing.
Reduce wording when evidence is incomplete.
Re-check after revisions, especially in abstract, introduction, conclusion, and rebuttal.

Our preliminary results suggest an improvement on dataset X.

Our method achieves state-of-the-art performance.

Use the strong wording only when the literature matrix, benchmark protocol, and completed experiments support it.

Claim ledger 用来把论文、综述、rebuttal 或 thesis 里的强主张逐条绑定到证据，避免“写得很强但证据跟不上”。

示例：

Claim	Evidence	Status	Risk	Action
方法在数据集 X 上减少标注时间。	已完成实验日志、Table 2 草稿。	Completed	需要独立重复 run。	添加 seed 和置信区间。
现有工作多数只在小 benchmark Y 上评估。	文献矩阵中的 Paper A/B/C。	Verified	Paper D 可能用了更大 benchmark。	最终写前先读 Paper D。
系统支持实时推理。	只有 benchmark 计划。	Planned	没有完成计时证据。	改写为目标，而不是结果。
R2 问题已由消融实验解决。	消融尚未完成。	Missing	rebuttal 会过度承诺。	完成实验或改为 limitation。

建议把每个 claim 标成 Strong、Qualified、Planned 或 Remove。只有文献、benchmark protocol 和已完成实验都支持时，才使用强表述。