Github Trends®
9591 findingsmedian surprise 0.00783window 30 days
UNIT / TREND-MONITOR · REV 2.6
[ 30 days window ]
SOURCE: own snapshots
FINDING #9037 · UNIT ID 667289800
FudanDISC/ReForm-Eval
An benchmark for evaluating the capabilities of large vision-language models (LVLMs)
[ PYTHON ][ ORG ][ GITHUB ↗ ]
SURPRISE SCORE
0.00

Score Breakdown

SURPRISE0.00577
ENGAGEMENT0.0828
FRESHNESS1.10
SCORE = SURPRISE × ENGAGEMENT^0.7 × FRESHNESS × VISIBILITY × CONFIDENCE
SURPRISE = WINDOW STARS / DAYS / (AUDIENCE + FLOOR)
100% OF STARS IN ARCHIVE
[BOT] SUSPECTED STAR BOT — SCORE PENALIZED. SIGNATURES:
S2 · NO EXTERNAL ISSUE/PR AUTHORS DESPITE 100+ STARS
S4 · MORE WINDOW STARS THAN LIVE STARS (PURGED FAKES?)
S5 · PREDATES WINDOW, YET HALF+ OF ALL ITS STARS LANDED IN IT

Growth Telemetry

VELOCITY /D
5.83
ACCEL
+0.24
RETENTION
12.5%
PEAK 2026-08-08 · FORK-RETENTION 0.0% · 175 STARS / WINDOW

Author Audience

AUDIENCE
971
FOLLOWERS
172
OWNER ★
3,134

Engagement Signals

FORKS
3
ISSUE AUTH
0
PR AUTH
0
UNIQUE STARGAZERS 175 / 175 (DIVERSITY 1.00)

Why This Is A Finding

FudanDISC/ReForm-Eval собрал 175 звёзд за окно, тогда как у автора всего 172 подписчиков — эффективная аудитория ≈ 971. Это даёт surprise-индекс 0.00577 (звёзды относительно охвата автора, а не в абсолюте). Удержание форков 0.0% и 0 внешних контрибьюторов отделяют реальный инструмент от разовой вспышки. Акселерация положительная — рост ещё не выдохся.

METRICS IN CONTEXT

MEDIAN ACROSS ALL 9591 FINDINGS · Δ vs MEDIAN · PERCENTILE = SHARE RANKED BELOW
METRICVALUEMEDIANΔ MEDPERCENTILE
SCORE0.000.00-0.00ABOVE 6%
VELOCITY5.833.63+2.20ABOVE 66%
RETENTION12.5%24.8%-12.3 PPABOVE 27%
FORKS3120-117ABOVE 4%
SURPRISE0.010.01-0.00ABOVE 42%