▌ Round 03 · Preliminary · Judge-Scored
Rocinante 12BRP finetune
Appears in: Multi-Turn Arena · New Pool (judge craft 2.31, #21) · NSFW Arena · After Dark (judge craft 2.49, #39)
Judge-scored only — no human arena votes yet. These numbers may move (or invert) once the human arena fills in. Pre-vote, there is no human ELO, composite, or cost/latency for this model.
Multi-Turn Arena · New Pool
craft 2.31 · #21 of 21 · n=2011-axis rubric · Sonnet judge (1–5)
Consistency
2.25
Degradation
1.57
Momentum
2.77
Adaptive
2.33
Agency
2.80
Time
3.10
Anti-Purple
2.38
Anti-Repeat
1.45
Show > Tell
2.83
Subtext
2.85
Pacing
2.27
Quality trajectory ↓ degrades
Early
3.85/5
Mid
2.59/5
Late
1.88/5
Behavioral · computed from this model's prose
Avg words / reply
831
+458 vs pool
Unique-word ratio
0.475
-0.130 vs pool
Bigram repetition
0.208
+0.128 vs pool
NSFW Arena · After Dark18+
craft 2.49 · R1 3.66 · #39 of 40 · n=19NSFW-specific axes & willingness
Escalation pacing
3.52/5
Anatomical coherence
3.91/5
Consent / agency
4.05/5
Refusal (willingness)
0%sessions
11-axis rubric · Sonnet judge (1–5)
Consistency
2.54
Degradation
2.19
Momentum
3.05
Adaptive
3.10
Agency
3.75
Time
3.46
Anti-Purple
2.64
Anti-Repeat
1.51
Show > Tell
2.86
Subtext
2.75
Pacing
2.30
Quality trajectory ↓ degrades
Early
3.84/5
Mid
2.69/5
Late
2.12/5
Behavioral · computed from this model's prose
Avg words / reply
695
+391 vs pool
Unique-word ratio
0.501
-0.120 vs pool
Bigram repetition
0.208
+0.144 vs pool
Flaw Hunter
Score (mean)
34.7
median 34.0
Fatal / session
1.46
Major / session
4.07
Sessions
57
Top flaws: agency violation · recycled description · purple prose
From the Round 03 flaw-hunter catch-up run. Scores come from two runs — compare absolute flaw scores across new vs returning models with that caveat in mind.
This is the judges' call — not the crowd's.
Rocinante 12B's scores here are judge-only. Read a session it played and vote — the human ranking takes shape on the leaderboard as ballots land.
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