Topic Map Contract

Topic map contract defines the minimum structure for a map-style wiki note or Korean post. Use it when a topic spans AI, Computational Biology, Math, Papers, Infra, or Agents and needs a reader-facing route rather than one isolated definition.

Map Fields

FieldQuestion
Reader questionWhat question does the map help answer?
Primary axisIs the map centered on object, modality, task, architecture, learning method, formula, benchmark, system, or agent workflow?
Object boundaryWhat entities or inputs are included, and what is out of scope?
RepresentationHow are raw objects converted into model inputs?
Task/outputWhat output space and validity rule define success?
Method familyWhich architectures, learning signals, generative processes, or agent loops matter?
Math anchorWhich formulas, distributions, symmetries, or estimators are necessary?
Evidence boundaryWhat metric, split, baseline, artifact, or benchmark supports claims?
Link bundleWhich reference notes should readers follow next?
Public boundaryWhat private context, unpublished result, or internal detail is intentionally absent?

Axis Examples

Map TypePrimary AxisRequired Secondary Axes
Architecture maparchitecturemodality, complexity, learning signal, evaluation
Molecular modeling mapobject/workflowrepresentation, split, metric, leakage
Math mapformula/objectivesymbols, distribution, estimator, metric
Paper cluster mapclaim/evidencepaper notes, benchmark contract, concept updates
Project mapartifact/problemdesign, verification, public boundary
Infra mapsystem behaviorresource, failure mode, reproducibility, safe runbook
Agent mapworkflowstate, tool contract, verifier, handoff, evidence ledger

A useful topic map should link at least one note from most relevant groups:

  • Object/entity or modality note.
  • Task or output-space note.
  • Method or architecture note.
  • Math or formula note.
  • Data, split, or benchmark note.
  • Evaluation or claim-boundary note.
  • Paper, project, infra, or agent workflow note when relevant.

Stop Conditions

  • The map is only a list of links without a reader question.
  • The map hides the primary axis.
  • The map mixes AI, Computational Biology, and Math without saying which one owns the explanation.
  • The map contains paper-specific claims without evidence boundary.
  • The map repeats full definitions that should live in concept notes.
  • The map depends on private data, unpublished results, or internal infrastructure context.