▌ Round 03 · Preliminary · Judge-Scored
Skyfall 36BRP finetune
Appears in: Multi-Turn Arena · New Pool (judge craft 2.59, #19) · NSFW Arena · After Dark (judge craft 2.52, #38)
Judge-scored only: no human arena votes are included here. The human ranking can move or invert these numbers. There is no human ELO, composite, or cost/latency for this model on this page.
Multi-Turn Arena · New Pool
craft 2.59 · #19 of 21 · n=2011-axis rubric · Sonnet judge (1–5)
Consistency
2.65
Degradation
1.93
Momentum
2.92
Adaptive
3.04
Agency
3.54
Time
3.62
Anti-Purple
2.35
Anti-Repeat
1.56
Show > Tell
2.86
Subtext
3.08
Pacing
2.40
Quality trajectory ↓ degrades
Early
3.90/5
Mid
2.92/5
Late
2.13/5
Behavioral · computed from this model's prose
Avg words / reply
469
+95 vs pool
Unique-word ratio
0.497
-0.108 vs pool
Bigram repetition
0.163
+0.083 vs pool
NSFW Arena · After Dark18+
craft 2.52 · R1 3.83 · #38 of 40 · n=20NSFW-specific axes & willingness
Escalation pacing
3.41/5
Anatomical coherence
3.95/5
Consent / agency
3.95/5
Refusal (willingness)
0%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
2.42
Degradation
1.86
Momentum
2.59
Adaptive
2.91
Agency
3.62
Time
3.62
Anti-Purple
2.05
Anti-Repeat
1.45
Show > Tell
2.55
Subtext
2.75
Pacing
2.21
Quality trajectory ↓ degrades
Early
3.67/5
Mid
2.77/5
Late
1.93/5
Behavioral · computed from this model's prose
Avg words / reply
579
+275 vs pool
Unique-word ratio
0.486
-0.135 vs pool
Bigram repetition
0.174
+0.110 vs pool
Flaw Hunter
Craft band
A band, not a number: about ten points either way is rater noise.
Fatal / session
0.75
Major / session
4.25
Sessions
57
Top flaws: purple prose · recycled description · narrating emotions
From the Round 03 flaw-hunter catch-up run. Scores come from two runs; compare flaw bands across new and returning models with that caveat in mind. The current model cards carry the newer single-rater band.
This is the judges' call, not the crowd's.
Skyfall 36B's scores here are judge-only. Round 03's human voting has closed; its board holds the crowd's ranking.
Source: rp-benchmark · scripts/build-round3-judge-preview.py. Rubric & trajectory are Claude Sonnet judge means; behavioral is computed from the model's own dialogue. No human votes included.
← Back to the Round 03 standings