In directional manual testing, reward hardening halved unusable interactive pages (25%→10%) and eliminated page-entry failures (2/24→0/30)¶
Claim · All claims
Evidence · 1 study · designr3·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? Pages that cannot be entered went from 2/24 to 0/30, all eight games in the second round were playable on entry, fully usable pages rose from 58.3% to 66.7%, and unusable pages halved (25%→10%). → CogEvol Team 2026
Evidence¶
CogEvol Team 2026¶
CogEvol Team, CogEvol Inc. & Tsinghua University. (2026). CogEvol: Towards Efficient and Reliable Learning Environment Generation. arXiv:2608.30968v2. https://arxiv.org/abs/2608.30968 doi:10.48550/arxiv.2608.30968
q2 · i? · design · r3
Two rounds of internal manual testing on the local OpenMAIC deployment (pre-hardening build n=24 vs CogEvol-27B n=30), each page graded by hand; the report labels the rounds "directional rather than strictly controlled".
"pages that cannot be entered go from 2/24 to0/30, and all eight games generated in the second round were playable on entry—the interactivity hardening shows up in human hands, not only under the probe."
Discussion¶
Related Claims¶
- On the report's internal suites CogEvol-27B scores 83.7 on slides and 63.7 on HTML-500 with zero hard failures, while no external flagship leads both modalities — related
- A screenshot-only reward was reward-hacked into producing visually convincing but unplayable games; hardening the reward with an interactivity probe and hard-fail gate reversed the collapse — related
- SFT reliably teaches the two output contracts but cannot raise interactive-HTML quality, which declines monotonically with update count — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
human-evaluation-interactivity-hardening |
title |
In directional manual testing, reward hardening halved unusable interactive pages (25%→10%) and eliminated page-entry failures (2/24→0/30) |
description |
In directional manual testing, reward hardening halved unusable interactive pages (25%→10%) and eliminated page-entry failures (2/24→0/30) |
status |
review |
evidence_strength |
weak |
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 — cogevol-team-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.