research-g-2-experiments

name	research[G.2]-experiments
description	ResearchPilot Research Assistant [Phase G.2]: Write manuscript Experiments
version	2.0.0
license	LICENSE

Phase G.2: Experiments

user_requirements.md priority: All user constraints in docs/user_requirements.md take precedence over any default instruction in this skill. Always read it before generating output.

Overview

Paper Writing Stage Chain

Skill	Responsibility
/research[G.0]-plan	Update idea_report → plan structure/figures → select format
/research[G.1]-method	Write / revise Method
/research[G.2]-experiments	Write / revise Experiments
/research[G.3]-abstract	Write / revise Abstract
/research[G.4]-introduction	Write / revise Introduction
/research[G.5]-related	Write / revise Related Works
/research[G.6]-conclusion	Write / revise Conclusion + References
/research[G.7]-review	Full-paper review (five dimensions + claim-evidence)

---

Command

/research[G.2]-experiments

An optional natural-language instruction may follow the command. The AI will treat it as an additional constraint or supplement for this invocation.

Prerequisite: G.1 Method written

---

Before Writing: Required Steps

1. Read the full manuscript (avoid disconnection from already-written sections)
2. Read all existing references (References section or .bib file)
3. Read the === Paper Architecture === comment block at the top of the manuscript
4. Read docs/idea_report.md and docs/dev_log.md (find answers to writing questions here first)
5. Read example analysis if docs/manuscripts/examples/style-notes.md exists
6. Scan all annotations in the manuscript, list pending ones
7. Confirm writing plan with user before writing anything

See references/common-writing-constraints.md for full details.

---

Experiments Section-Specific Requirements

Three Core Questions (Master-cai experiments.md):

1. Is the proposed method better than strong baselines? (main experiment)
2. Which modules/design choices drive the gain? (ablation)
3. How well does the method generalize? (additional analysis)

Data source: All quantitative results must be read from results/ files. No estimation or fabrication.

Table constraints: booktabs format, bold best, underline second-best, complete caption.

Text analysis: Must state which metric beats which baseline by how much (specific numbers), and explain why.

---

Version Management

Back up before writing:

cp paper.md paper_{mm-dd_hh-mm}.md   # or .tex

Append modification record to file header after writing (see references/common-writing-constraints.md).

---

Reference Maintenance

When a new citation is needed, append immediately:

md:

[N] {Author}. "{Title}." *{Journal/Conference}*, {Year}.
> Core contribution: {what this paper does, one sentence}
> Reason for citation: {where in the paper, why cited here}

LaTeX .bib:

% [Core contribution] {one-sentence contribution}
% [Reason for citation] {section}: {why cited here}
@article{key, ...}

---

On Phase Completion

Experiments section written/revised.

→ Use `/research[G.3]-abstract` to enter Abstract.
→ Or use `/research[G.7]-review` for a full-paper review at any time.

---

Reference Files

• Section guide: references/section-guide.md
• Common constraints: references/common-writing-constraints.md
• Template flexibility: references/template-flexibility.md