Standard semantic retrieval alone does not reliably recover the pedagogically required chunk types, with comparison chunks retrieved worst (14.3% on the real corpus)¶
Claim · All claims
Evidence · 1 study · designr2·q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? A standalone semantic retrieval system found the required chunk in only 62.6% of real-corpus queries overall, and only 14.3% for COMPARE, showing the failure is structural rather than parametric. → Laurent Brisson 2026
Evidence¶
Laurent Brisson 2026¶
Laurent Brisson, Maria Teresa Segarra and Gregory Smits. (2026). A didactical-driven teacher assistant for a dimensional modeling course. https://arxiv.org/abs/2607.22598 doi:10.48550/arxiv.2607.22598
q2 · i? · design · r2
Retrieval evaluation (RQ1) on the real corpus of student questions: a semantic-only pipeline (nomic-embed-text-v1.5, cosine similarity, top-15) was tested for whether the required chunk type appears in top-15. DEFINE reached 89.1% retrieval, EXPLAIN 58.2%, and COMPARE only 14.3%, described as "structural rather than parametric".
"C OMPARE fails most severely: only 14.3% of comparison chunks are retrieved. This failure is struc- tural rather than parametric: a comparison chunk en- codes a relationship between two specific concepts, a constraint that embedding similarity cannot capture from a single query vector."
Discussion¶
Related Claims¶
- Manual error analysis finds incomplete retrieval is the most frequent failure category, ahead of code-semantics limits, bilingual retrieval gaps, and over-conservative refusals — related
- Evidence tracing links 96 of 114 generated claims to source chunks, but three claims with retrieval scores of 0.50 or below show retrieval similarity alone is insufficient — related
- Retrieval failure explained only 2 of vector RAG's 11 ungrounded answers; 9 answers retrieved the correct material and were still ungrounded, suggesting chunked context invites fabrication — related
- General-purpose embedding yields better retrieval while domain-specific embedding yields higher answer completeness and relevance — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
semantic-retrieval-insufficient-pedagogical-content |
title |
Standard semantic retrieval alone does not reliably recover the pedagogically required chunk types, with comparison chunks retrieved worst (14.3% on the real corpus) |
description |
Standard semantic retrieval alone does not reliably recover the pedagogically required chunk types, with comparison chunks retrieved worst (14.3% on the real corpus) |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-10 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — laurent-brisson-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.