TEI@N inference-time selection raises the improvement rate on pre-incorrect math scenarios from 59.0% to 81.9% at N=8 over Greedy on a frozen DeepSeek-R1-8B base¶
Claim · All claims
Evidence · 1 study · causalr2·q2
Currency · no dated study recorded · not yet checked against its sources · not reviewed by a person
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? At N=8, TEI@N reaches an 81.9% improvement rate versus Greedy's 59.0% (+22.9pp) on pre-incorrect scenarios, with no training and no external judge. → tei-selection-1
Evidence¶
tei-selection-1¶
Shim Jaechang and Unggi Lee. 2026. The Tutoring Effectiveness Index: Predicting LLM Math Tutor Quality from Four Conversation Signals. In Proceedings of The 35th ACM International Conference on Information and Knowledge Management (CIKM '26). https://arxiv.org/abs/2605.30666 doi:10.48550/arxiv.2605.30666
q2 · i? · causal · r2
Block-grouped selection experiment on the pre-incorrect subset of 1,000 in-domain BigMath tutoring scenarios (n=363 pre-incorrect), 20 random repeats per N, DeepSeek-R1-8B base. "TEI@𝑁 reaches81.9%versus Greedy59 .0%(+22.9pp)"; no effect size printed.
"At𝑁= 8, TEI@𝑁 reaches81.9%versus Greedy59 .0%(+22.9pp). The gap is stable across𝑁 ; DTR@𝑁 trails TEI@𝑁 by13to15pp at every𝑁, and Random@𝑁never exceeds Greedy."
Discussion¶
Related Claims¶
- On the same frozen base, TEI@4 beats both DTR@4 and Greedy on Δ Solve Rate, leak rate, and composite R_total at no training cost — related
- TEI@4 selection shifts thinking-trace mass toward deeper deliberation, raising Schoenfeld Explore share by +10.8pp without changing the model — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
tei-selection-raises-improvement-rate-pre-incorrect |
title |
TEI@N inference-time selection raises the improvement rate on pre-incorrect math scenarios from 59.0% to 81.9% at N=8 over Greedy on a frozen DeepSeek-R1-8B base |
description |
TEI@N inference-time selection raises the improvement rate on pre-incorrect math scenarios from 59.0% to 81.9% at N=8 over Greedy on a frozen DeepSeek-R1-8B base |
status |
review |
evidence_strength |
moderate |
generated |
by process:wiki-ingest on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — tei-selection-1 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.