Experiments And Discussion Section Writing

Objective

Convince reviewers that the paper's claims are empirically tested: effectiveness, causal contribution, robustness, and realistic limits should be visible.

Claim-To-Experiment Plan

Start from the paper's main claims:

Claim Type	Evidence Needed
Better performance	strong/recent baselines, same protocol, standard metrics
Module contributes	ablation removing/replacing the module
Robustness/generalization	harder settings, OOD cases, sensitivity analysis
Efficiency	runtime, memory, parameter count, or deployment setting
Mechanistic explanation	qualitative results, error analysis, or discussion tied to design

If a claim has no matching experiment, mark needs evidence rather than writing around the gap.

Section Roles

1. Setup: datasets, metrics, baselines, implementation settings, and protocol fairness.
2. Main comparison: one message per table/figure; state the comparison target and bounded conclusion.
3. Ablation: connect each design choice to a result delta.
4. Analysis/discussion: explain why results occur; compare to related work when appropriate.
5. Limitations: state scope boundaries or failure cases honestly.

Table And Figure Communication

• One table, one message.
• Put metric direction in headers when possible, such as PSNR ↑ or LPIPS ↓.
• Keep captions focused on setting/protocol/notation.
• Do not claim significance unless variance, statistical test, or repeated-run evidence is visible.

Discussion Layering

A strong discussion does not repeat numbers. It should move through:

finding -> mechanism/interpretation -> comparison to prior work -> limitation or implication

Do not add prior-work comparisons unless the referenced work is already cited or the user asks for verified literature work.