SeraphimSerapis/tool-eval-bench — GitHub trend

FINDING #2750 · UNIT ID 1213526705

SeraphimSerapis/tool-eval-bench

Tool-calling quality benchmark for LLM serving stacks. 80+ deterministic scenarios testing multi-turn orchestration, safety boundaries, and structured output. Supports vLLM, SGLang, and llama.cpp.

[ PYTHON ][ GITHUB ↗ ]

SURPRISE SCORE

0.00

Score Breakdown

SURPRISE0.00762

ENGAGEMENT0.53

FRESHNESS1.61

SCORE = SURPRISE × ENGAGEMENT^0.7 × FRESHNESS × VISIBILITY × CONFIDENCE

SURPRISE = WINDOW STARS / DAYS / (AUDIENCE + FLOOR)

17% OF STARS IN ARCHIVE

Growth Telemetry

VELOCITY /D

1.77

ACCEL

-0.07

RETENTION

12.1%

PEAK 2026-08-02 · FORK-RETENTION 100.0% · 53 STARS / WINDOW

Author Audience

AUDIENCE

192

FOLLOWERS

142

OWNER ★

498

Engagement Signals

FORKS

ISSUE AUTH

PR AUTH

UNIQUE STARGAZERS 53 / 53 (DIVERSITY 1.00)

Why This Is A Finding

SeraphimSerapis/tool-eval-bench собрал 53 звёзд за окно, тогда как у автора всего 142 подписчиков — эффективная аудитория ≈ 192. Это даёт surprise-индекс 0.00762 (звёзды относительно охвата автора, а не в абсолюте). Удержание форков 100.0% и 0 внешних контрибьюторов отделяют реальный инструмент от разовой вспышки. Акселерация отрицательная — внимание остывает после пика.

Related Findings

RANKS ABOVE 71% OF 9595 FINDINGS

#02guillaumemeyer/watermarks-remover[ PYTHON ]0.43 #09awesome-dsh-plugin/awesome-dsh-plugin[ PYTHON ]0.38 #10wang2122/sprix-sage-router[ PYTHON ]0.37 #16virgiliojr94/book-to-skill[ PYTHON ]0.31 #19ayghri/i-have-adhd[ PYTHON ]0.28

METRICS IN CONTEXT

MEDIAN ACROSS ALL 9595 FINDINGS · Δ vs MEDIAN · PERCENTILE = SHARE RANKED BELOW

METRICVALUEMEDIANΔ MEDPERCENTILE

SCORE0.010.00+0.00ABOVE 71%

VELOCITY1.773.67-1.90ABOVE 11%

RETENTION12.1%24.6%-12.5 PPABOVE 26%

FORKS41115-74ABOVE 26%

SURPRISE0.010.01-0.00ABOVE 48%