| name | autoacad-survey |
| description | Use when collecting, screening, and synthesizing literature, benchmarks, datasets, and theory needed for a research-paper project. |
AutoAcad Survey
Overview
Use this stage to build the evidence base. The output is not a prose essay; it is a reusable set of verified literature notes that can support method design and reviewer defense.
Workflow
- Start from a search plan with multiple query families: core topic, methods, theory, datasets, and applications.
- Prefer real sources first: OpenAlex, Semantic Scholar, arXiv, official proceedings.
- Screen aggressively for topic relevance. Reject impressive but off-topic papers.
- Preserve identifiers exactly when available: DOI, arXiv ID, cite key.
- Extract structured cards: problem, method, data, metrics, findings, limitations, citation.
- End with synthesis: topic clusters, gaps, benchmark norms, strong baselines, and likely reviewer comparisons.
Quality Bar
- Prefer recent work that aligns with the project's actual topic.
- Use at least 30 real references for a full paper workflow unless the domain genuinely lacks them.
- Keep direct quotes minimal; preserve identifiers and links instead.
