Claims

Kimi shows reasonable but weaker agreement with human examiners than ChatGPT (weighted kappa 0.571)

Claim · All claims
Evidence · 1 study · associational r1 · q3
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.

q3 i? Kimi in thinking mode reached weighted kappa 0.571 with human graders, with 28.1% exact and 78.1% within ±1 grade agreement. → Brevik A 2026

Evidence

Brevik A 2026

Brevik A, Jerpseth H and Lafontan SR (2026). Large language models as grading assistants in public health education: a method-comparison study of essay-style exam assessment. Front. Educ. 11:1904452. https://doi.org/10.3389/feduc.2026.1904452

q3 · i? · associational · r1

Same single-file method-comparison analysis of 32 submissions. The kappa table reports Human vs. Kimi thinking mode at 0.571 (95% CI 0.373–0.768); the text states "the agreement was good for ChatGPT and reasonable for Kimi." Exact match 28.1%, cumulative ±1 agreement 78.1%.

"For LLMs in the thinking mode, the agreement was good for ChatGPT and reasonable for Kimi."

Discussion

Page metadata
Field Value
type claim
id kimi-reasonable-human-grading-agreement
title Kimi shows reasonable but weaker agreement with human examiners than ChatGPT (weighted kappa 0.571)
description Kimi shows reasonable but weaker agreement with human examiners than ChatGPT (weighted kappa 0.571)
status review
evidence_strength moderate
newest_evidence 2026
recheck_by 2029
generated by process:wiki-ingest on 2026-10-11
trust unverified — no one has confirmed this page's content
sources 1 — brevik-a-2026

Frontmatter is the page's machine-readable record — see the Schema & Guide.