Cross-model agreement on assertions is moderate (median 0.401) and higher among construct-derived than corpus-derived assertions¶
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? Across 221 assertions annotated by five LLMs, cross-model agreement is moderate with a median Krippendorff's alpha of 0.401, and construct-derived assertions agree more than corpus-derived ones; filtering at alpha ≥ 0.5 leaves 33 of 74 corpus-derived and 11 of 48 construct-derived assertions. → Julian Bernado 2026
Evidence¶
Julian Bernado 2026¶
Julian Bernado, Ana Trindade Ribeiro, Xander Beberman, and Susanna Loeb. (2026). EduBehaviors: Assertion-Based Schemas for Auditable Coding of Educational Dialogues. https://arxiv.org/abs/2609.27043 doi:10.48550/arxiv.2609.27043
q2 · i? · design · r2
Annotation analysis of 221 assertions (74 corpus-derived, 48 construct-derived, 100 word-based) by a panel of five language models, with per-assertion Krippendorff's alpha shown in Figure 1. The article reports agreement is "moderate (median = 0.401" and higher for construct-derived assertions.
"Overall, cross-model agreement is moderate (median = 0.401 and higher among the construct-derived assertions than the data-derived assertions."
Discussion¶
Related Claims¶
- All three assertion sources (word, corpus-derived, construct-derived) contribute to predictive accuracy — related
- Adding a cross-model agreement filter improves performance for the top three LLM annotators — related
- Within-judge variance of the repeated LLM-judge rubric is negligible, but cross-model-family agreement remains untested — related
- Same-model LLM configuration pairs agree more than cross-model pairs, and agreement decreases monotonically as temperature difference increases — related
- Constructs with higher operational clarity show higher overall coder agreement, and low clarity harms human coder agreement more than LLM agreement — related
- Human-human agreement (average κ = 0.644) was lower than the best LLM-LLM agreement (average κ = 0.856) across constructs — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
assertion-cross-model-agreement-moderate-median-0401 |
title |
Cross-model agreement on assertions is moderate (median 0.401) and higher among construct-derived than corpus-derived assertions |
description |
Cross-model agreement on assertions is moderate (median 0.401) and higher among construct-derived than corpus-derived assertions |
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 — julian-bernado-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.