method-writer

Method Writer

Purpose

Draft or revise the Method section so that reviewers can understand what the method does, why it should work, and how to reproduce the key algorithmic choices.

When To Use

• Turning an idea or implementation into a paper-ready Method section.
• Checking whether notation, assumptions, and algorithm steps are clear.
• Revising a rejected paper whose method was considered vague or underspecified.
• Preparing method content for ICLR, NeurIPS, CVPR, ICML, ECCV, AAAI, or TMI.

Inputs

Required:

• paper thesis and contribution map;
• method summary;
• problem formulation;
• inputs, outputs, assumptions, and notation;
• algorithm or model architecture;
• training/inference procedure;
• implementation constraints that affect claims.

Optional:

• pseudocode;
• diagrams;
• equations;
• ablation plan;
• reproducibility checklist.

Procedure

1. Define the task and notation before introducing the method.
2. State assumptions and scope conditions explicitly.
3. Separate problem formulation, model design, optimization/training, and inference.
4. Explain the mechanism behind the claimed improvement.
5. Add algorithmic details only when they affect reproducibility or claims.
6. Cross-check method claims against planned experiments.
7. Move purely engineering details to appendix unless they are central to the contribution.

Rubric

Dimension	Strong	Weak
Problem formulation	Inputs, outputs, assumptions clear	Method starts without task definition
Mechanism	Explains why the method should work	Only lists components
Reproducibility	Key hyperparameters/procedure recoverable	Missing training/inference details
Claim alignment	Method supports contribution bullets	Method includes unrelated features
Notation	Minimal and consistent	Overloaded or undefined symbols

Venue Adaptation

• ICLR/NeurIPS/ICML: emphasize conceptual mechanism, assumptions, algorithmic clarity, and ablations.
• CVPR/ECCV: connect method modules to visual evidence and architecture diagrams.
• AAAI: keep method broadly understandable to AI reviewers outside a narrow subfield.
• TMI: include medical-imaging preprocessing, modality assumptions, validation constraints, and clinical-scope caveats.
• arXiv: include enough implementation detail for public reproducibility.