Replacing evolutionary experts with random-path experts collapses performance, and greedy 1-step lookahead reaches only +0.155 EP on ASSIST15 L=10¶
Claim · All claims
Evidence · 1 study · causalr2·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? BC on random-path experts reaches +0.354 EP, well below PPO-vanilla (+0.543), while evolutionary search reaches +0.614, showing bad demonstrations actively bias the policy and joint sequence optimization matters. → Geonwoo Bang 2026
Evidence¶
Geonwoo Bang 2026¶
Geonwoo Bang, Dongho Kim, and Moohong Min. (2026). EVOL: Simulator-Guided Evolutionary Expert Synthesis for Deployment-Free Learning Path Recommendation. arXiv:2610.03273. https://arxiv.org/abs/2610.03273 doi:10.48550/arxiv.2610.03273
q2 · i? · causal · r2
Controlled Stage-1 ablation on ASSIST15 L=10 (3 seeds, Table 3) comparing greedy 1-step lookahead (+0.155), random-path experts (+0.354), and evolutionary search (+0.614) under identical candidate sets and budgets.
"(ii) Random-path experts discard search entirely; BC on them reaches+0.354, well below PPO-vanilla (+0.543), so any source of demonstrations is not enough—bad demonstrations actively bias the policy."
Discussion¶
Related Claims¶
- Under deployment-free inference, at least one EVOL variant achieves the best test EP in every dataset–length setting across three datasets and eight baselines — related
- Final LPR performance is governed by the quality of evolutionary experts rather than by the particular imitation objective — related
- Sequencing rather than concept selection explains most of the EVOL–PPO-vanilla EP gap on ASSIST15 L=10 — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
random-experts-collapse-lpr-performance |
title |
Replacing evolutionary experts with random-path experts collapses performance, and greedy 1-step lookahead reaches only +0.155 EP on ASSIST15 L=10 |
description |
Replacing evolutionary experts with random-path experts collapses performance, and greedy 1-step lookahead reaches only +0.155 EP on ASSIST15 L=10 |
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 — geonwoo-bang-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.