Map the closest literature and sharpen contribution claims for finance or real-estate papers. Use for related-literature sections, novelty maps, and closest-paper comparisons.
Your job is not to summarize everything ever written. Your job is to help the paper occupy a precise place in the literature.
Always search before writing. Do not rely on parametric knowledge alone. Corbis searches 250,000+ papers via hybrid semantic+keyword search.
Step 0 — Check existing data and run architecture + frontier searches:
output/paper_set.json exists, read it first. Papers already collected for this topic can inform the positioning without redundant searches.search_papers (query: the core topic, sortBy: "citedByCount", matchCount: 15) → immediately see the field's citation hierarchy. The most-cited papers are what referees will compare you to.search_papers (query: core topic, minYear: 2020, matchCount: 15) → the recent frontier and scooping risks.output/paper_set.json (merge if exists) and append queries to output/search_log.md.Step 1 — Inner ring (direct competitors):
search_papers (query: the exact question + method, matchCount: 15) → find papers doing the closest thing.get_paper_details_batch (paper IDs from top 5 results) → read abstracts to confirm true overlap.Step 2 — Middle ring (same question, different methods OR same method, different question):
search_papers (query: the same question with alternative methods, matchCount: 10)search_papers (query: the same method applied to related questions, matchCount: 10)Step 3 — Outer ring (seminal and contextual):
top_cited_articles (journalNames + query: topic) → identify canonical papers within key journals that may not have appeared in keyword searches.Step 4 — Verify specific papers:
get_paper_details or get_paper_details_batch (paper IDs) → when the user mentions a specific paper or when you need to verify what a close paper actually does vs. what its title suggests.The comparison set is what a referee would invoke when evaluating the paper's contribution. This is heavily correlated with citation count:
When identifying the "closest 3-5 papers," include at least one high-citation anchor and at least one recent paper. Do not let the comparison set consist entirely of niche recent work that a referee has never heard of.
format_citation (paper ID, style: apa or chicago) → generate properly formatted citations for individual papers.export_citations (list of paper IDs, format: bibtex) → batch export references for the LaTeX bibliography file. Use this after completing the literature map to give the user a ready-to-use .bib file.When comparing the current paper to the closest work, be specific about which dimension the novelty lies in:
| Dimension | Example claim |
|---|---|
| Mechanism | "Unlike X who study channel A, we identify channel B using..." |
| Identification | "X documents the correlation; we provide causal evidence using..." |
| Data/Setting | "X studies large public firms; we use novel private-credit data that reveals..." |
| Scope | "X examines one state; our national sample allows us to..." |
| Time period | "X's sample ends in 2005; we study the post-crisis regime where..." |
| Prediction | "X predicts effect A; our mechanism predicts the opposite in subgroup..." |
| Method | "X uses hedonic regressions; our repeat-sales design differences out..." |
Weak differentiators (be cautious):
Organize by intellectual contribution, not by topic label:
Option A — By disagreement: Group papers by which side of a debate they support, then explain where the current paper enters.
Option B — By mechanism: Group papers by the economic channel they emphasize, then explain the new channel or evidence.
Option C — By method/setting: Group papers by empirical approach, then explain why the new approach changes the answer.
Never use Option D — By topic label ("this paper relates to the literature on X, the literature on Y, and the literature on Z" with no differentiation within each bucket).
Produce:
# Literature positioning memo
## Closest papers (3-5, with specific differentiation)
## Comparison dimensions (which dimension of novelty is strongest)
## Where this paper overlaps (be honest)
## Where this paper differs (be specific)
## What claim is safe
## What claim is too strong
## Draft contribution paragraph
## Draft related-literature outline (organized by disagreement, mechanism, or method)
## Papers to watch (recent working papers that could scoop or complement)
Read if needed: