[reference] · ⏳ outline — deep note in progress (part of the research-course build). This is the “different writing styles” note — the one you specifically wanted, since you’ll write across AI/ML, robotics, chips, and possibly other fields, each with its own conventions.
There is no single “research paper” template. The structure, tone, evidence, and length shift by paper type and by field — writing an empirical ML paper like a math-theory paper (or vice-versa) reads as amateur. This note is how to recognize which register a piece of work needs and switch into it.
Will cover
By paper type
- Empirical / experimental (most ML) — hypothesis → method → experiments → results dominate; heavy on baselines, ablations, tables; “we show empirically that…”.
- Systems / engineering — the artifact is the contribution; emphasis on design, architecture, trade-offs, and benchmarks (throughput/latency/cost); “we built X and evaluated it.”
- Theory / mathematical — definitions → theorems → proofs; terse, formal, every claim proven; experiments optional or illustrative.
- Robotics / hardware — hybrid: system design + real-world experiments + often a video/demo; sim-to-real gap discussion; safety.
- Survey / review — no new experiments; contribution is organization — a taxonomy of a field, thematic synthesis, open problems.
- Position / vision — argue for a viewpoint or research direction; persuasion over data.
- Short paper / workshop / extended abstract — one crisp idea, less complete evidence; a great first-paper target.
- Reproducibility / negative-results — replicate or refute; increasingly valued.
By field (conventions differ)
- CS/ML — conference-first culture, arXiv preprints, 8-page limits, code release expected, fast cycle, related-work often late.
- Traditional engineering / physics — journal-first, IEEE/Elsevier styles, longer, more formal, methods-heavy.
- Sciences (bio/chem) — rigid IMRaD (Intro-Methods-Results-and-Discussion), passive voice still common.
- Humanities/social — argumentative essays, qualitative evidence, footnote-heavy citation styles.
The transferable move
- How to infer the target style: read 3–5 recent papers from your exact target venue and mirror their structure, length, tone, and citation style before writing a word.
Related
- Anatomy of a Paper · Scientific Writing Craft · Venues & Publication · What Research Is (contribution kinds)
Jump to a subject
Software engineering
Infrastructure & security
Computer science
Hardware & engineering