fix(benchmarks): remove qwen3.6-plus inherited predecessor scores from personality card - #26
Open
TechNickAI wants to merge 2 commits into
Open
fix(benchmarks): remove qwen3.6-plus inherited predecessor scores from personality card#26TechNickAI wants to merge 2 commits into
TechNickAI wants to merge 2 commits into
Conversation
…m personality card
The Qwen3.6 Plus personality card published four numbers (sycophancy 6.2,
EQ 60.45, warmth 13.4) inherited from the Qwen3.5-397B predecessor, with a
parenthetical admitting the inheritance. The charter's non-negotiable: never
present a successor model's score as a predecessor's. The data layer was
already cleaned (eq_bench: {} in model-data.json), but the prose copy in
index.html remained live — the exact guardrail failure of removing bad data
without grepping the rendered HTML.
The card now honestly states the model has no EQ-Bench data. Empty is honest.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this fixes
The Qwen3.6 Plus personality card on
heartcentered.ai/model-benchmarks/has been publishing four inherited numbers as prose:Sycophancy 6.2/EQ 60.45/Warmth 13.4These are from the predecessor model, not from any run on Qwen3.6 Plus itself. The charter's non-negotiable says never present a successor model's score as a predecessor's, and the footnote is an admission, not a remedy.
The data layer was already cleaned —
model-data.jsonshowsbenchmarks.eq_bench: {}forqwen/qwen3.6-plus:free(verified live). But the prose copy inindex.htmlwas never touched, so the four numbers stayed live on the public site for visitors to read as real data. This is exactly the failure the project's own guardrail names: removing bad data without grepping the rendered HTML.The fix
The card now honestly states the model has no EQ-Bench data. Empty is honest; an inherited number is a lie that looks like data.
One file changed:
model-benchmarks/index.html(+7/−11 lines). No data, JS, or other pages touched.Verification
<div>and<p>counts match before/afterDecision needed
Merge this PR to fix the charter violation on the live site, or say no and I'll close it. The pre-declared default (per the 2026-07-30 zoom-out) was: if no answer by 2026-07-31 18:00Z, Cora opens the PR and leaves it for Nick to merge. This is that PR — opened at 10:05Z, well before the deadline. Merging prose to
mainis Nick's call by charter.