Claim-Evidence Matrix
1 Why This Page Matters
Many weak papers fail because their claims are not matched by the right evidence.
The central question is simple:
For each claim, what evidence would make a careful reviewer trust it?
2 A Practical Matrix
| Claim type | Evidence that usually matches it | Weak version reviewers distrust |
|---|---|---|
| Theoretical guarantee | precise theorem, assumptions, proof, interpretation | theorem with unclear scope or unexplained assumptions |
| Empirical performance | strong baselines, ablations, repeated runs, error bars when relevant | one benchmark table with weak baselines |
| Efficiency or scalability | runtime, memory, compute setting, scale regime, fair implementation details | vague “faster” language without setup clarity |
| Robustness | stress tests, sensitivity analyses, failure cases, distribution shift checks | one lucky result under default settings |
| Practical usefulness | real deployment context, meaningful task framing, realistic constraints | toy example used as a substitute for application |
3 Strong Habit
Write claims in one column and required evidence in the next before you draft the paper. This forces alignment early and makes weak spots visible before review.
4 How This Connects To The Site
- theorem-heavy pages should link here when discussing research use
- application pages should link here when discussing experiments
- paper labs should ask whether each major claim has matching support