GEO

Start here

Diagnose prerequisites without turning them into an admission gate

The course assumes working familiarity with retrieval, uncertainty, web protocols, and reproducible analysis. This local diagnostic identifies the shortest bridge route; it does not collect responses or rank learners.

Diagnostic
12 ungraded items
Bridge notes
4 compact routes
Threshold
70% for targeted review

Ungraded diagnostic · 12 items

Find the bridge you need before W01

Answer from your current understanding. The result changes only this page and is not stored, submitted, or used for admission.

0 / 12 answered0 of 12
01 · IR foundationsWhat is the narrowest correct description of a reranker?
02 · IR foundationsA query has four judged-relevant documents and the top five retrieve three of them. What is Recall@5?
03 · IR foundationsA relevant document is absent from the candidate set. What can a later reranker do?
04 · UncertaintyWhy can 30 runs in one session be weaker than runs spread across time blocks?
05 · UncertaintyWhich statement about a frequentist 95% confidence interval is defensible?
06 · UncertaintyWhen should the rule for missing responses be chosen?
07 · Open-web protocolsWhat does sitemap membership establish by itself?
08 · Open-web protocolsWhat is the strongest safe conclusion from one HTTP 200 response you requested?
09 · Open-web protocolsWhy must crawl controls and index state be recorded separately?
10 · ReproducibilityWhich addition best preserves the identity of an input artifact?
11 · ReproducibilityWhat should happen to an adverse secondary result?
12 · ReproducibilityWhat is still required after an analysis script runs successfully once?

Compact bridge notes

Four routes into the common vocabulary

01

Bridge note

IR foundations

When to use it

Use this before W03–W05 if candidate generation, ranking, qrels, Recall@k, or NDCG are unfamiliar.

Minimum concepts

  • Query, document, corpus, index, qrel, candidate set, ranker
  • Precision and recall use different denominators
  • Candidate recall constrains reranking
  • Sparse, dense, and hybrid are testable system choices, not quality labels

Checkpoint

Work one Recall@5 calculation, then inspect the frozen BM25/dense/hybrid runs in L03.

Open the practice route →
02

Bridge note

Uncertainty

When to use it

Use this before W06–W08 if repeated samples, intervals, missingness, or drift are not yet routine.

Minimum concepts

  • Sampling frame and unit of analysis
  • Within-block versus between-block variance
  • Estimator, standard error, interval, and assumption
  • Missingness, stopping rules, sensitivity, and multiplicity

Checkpoint

Change runs, blocks, and drift in the measurement explorer; write what added repetition cannot identify.

Open the practice route →
03

Bridge note

Open-web protocols

When to use it

Use this before W02 if crawl, index, retrieve, cite, mention, and refer tend to collapse into one idea.

Minimum concepts

  • HTTP observation is request- and time-bounded
  • Links and sitemaps support discovery but do not guarantee indexing
  • Directive, crawler behavior, index state, and removal are distinct
  • Authorization and technical reachability are separate

Checkpoint

Use the visibility-chain diagnostic to label one observable signal and one forbidden inference at each stage.

Open the practice route →
04

Bridge note

Reproducible research

When to use it

Use this before any lab if versions, hashes, manifests, preregistration, or negative-result handling are new.

Minimum concepts

  • Freeze or lawfully retrieve exact inputs
  • Record environment, parameters, command, and expected result
  • Keep deviations, failures, and adverse outcomes
  • Separate computational reproduction from scientific validation

Checkpoint

Run L01 offline, inspect its manifest, then ask whether a second analyst can reconstruct the observation boundary.

Open the practice route →