Tracer
← polyphony · eval · frankenstein · ci

polyphony · claude-opus-5-5

✓ pass≠ not comparable
Pass rate
100%
Cases
1
Mode
eval
Suite
frankenstein
Env
ci
Started
2026-10-25 04:55 UTC
Duration
—
Cost
—
Tokens in/out
— / —
Commit
Branch
main
Source

Metrics

prose.score0.700 (was 0.901)
prose.words3,777 (was 3,593)

Labels

cadenceon-demand
experimentw17-r2-1
harness_sha86123fdae7a0f3d2805e1ab75905d51c08993484
judge_providergroq+anthropic
judge_self_gradedTrue
repeat1
triggerworkflow_dispatch

Compared with the previous run

Baseline 2026-10-25 00:37 UTC · 100% → this run 100% · code diff ↗

± score moved · 1
  • eval:prose: 0.90 → 0.70

Cases (1)

eval:prosescore 0.70✓ pass
score=0.700words=3,777queued=4deferred=0n_scenes=4r2_lines=18cast_named=1prose_chapter=1r2_applicable=0.750cast_from_beat=1r1_capped_scenes=0estimate_over_actual=0.741beat_seam_max_overlap=0scene_seam_max_overlap=0
Output
words=3777 path=planned rubrics={'R1': 0.9, 'R2': 0.4167, 'R3': 0.9025, 'R4': 0.81, 'R5': 0.8625} seams=0/0
⚖︎ Judge rationale

scene 1 R1: Rowan Merrow is the only character in the cast and appears as planned, with his sister and the mate used as the plan specifies; the mate is an unnamed functional figure and nobody acts against canon, though the 'Mr.—' slip invents an unspoken name and the scene stays minimal on canon-named cast. | scene 1 R3: All four beats are realised in the planned order with their changes landing (dream and fear of foolishness; confession of friendlessness with feeling smaller; crew muttering and mate at the door; decision to hold course and written commitment), with only trivial additions like the 'Mr.—' slip and a | scene 1 R4: The close-third past-tense POV on Merrow is consistent, the tone is exalted, restless and lonely, and the prose is concrete and varied with all four beats dramatized as scene; minor flaws include the odd 'Mr.—' slip, a few borrowed or generic lines, and mild repetition of the swinging-lamp motif. | scene 1 R5: The scene follows the brief and canon (Merrow, his sister, the mate, the crew, the ice, the north commitment) and contradicts no established facts, though Merrow's aborted "Mr.—" to his own mate is a slightly odd, unexplained touch. | scene 1 R2: not applicable: under two cast speakers with voice samples have attributed lines | scene 2 R1: Both principals (Merrow, Emeric) appear as planned and behave per the plan, with only minor functional unnamed characters (mate, cook, seamen); however, Emeric is never named in the scene (only 'the stranger'), which is consistent with Merrow not knowing him yet, so there is only slight deduction fo | scene 2 R2: Line 2 ('whither you are bound') plausibly reads as Vandel, but lines 1, 3 and 4 are terse, generic ship-command speech that could belong to either character or to a crew member, so Rowan's ornate, periodic diction from his samples is not reflected and attribution rests mostly on context rather than | scene 2 R3: All five beats are realised in order with their changes landing, but there are minor deviations: the fog and ice closing take four days rather than the planned few, the 'ice speaks' night is lightly expanded, and the closing paragraphs mildly editorialize the 'giant followed or fled' mystery, though | scene 2 R4: The scene holds close third-person past tense on Merrow throughout, with an eerie, awestruck tone, concrete imagery, and varied sentences, and it carries no labels or prefixes; minor weaknesses are a few generic phrases ("God keep us," "the not knowing pressed upon him"), a slightly summarizing clos | scene 2 R5: The scene follows the canon, the planned beats and the prior text with no real contradictions: the cabin and lamp callback, the mate, the sister letter, the 'yesterday' timeline, the sledge and dogs, and the galley placement all agree. The only quibbles are trivial: 'all but one' dog dead against th | scene 3 R1: Both principals, Merrow and Vandel, appear with exact names and behave as canon and plan describe, and the only invented figures (the cook, cabin boy, seamen) are minor and functional, with one small deduction for Merrow being called 'Captain' and for details like the cabin boy that go slightly beyo | scene 3 R2: Most lines can be attributed by content and address (captain versus rescued stranger), but they are terse and plain, unlike the long ornate samples, and ambiguous lines like 1 and 9 depend on context rather than distinctive diction, which fits the 0.4 anchor. | scene 3 R3: All four planned beats occur in order with their stated changes landing (move to cabin and nursing, name and pursuit with rigid hunger for the giant's direction, growing attachment and the sister-letter realization, the pole confession with Vandel's alarm and Merrow's chill), with only minor unplann | scene 3 R4: The scene keeps close third-person past tense from Merrow's POV with a consistent tender-to-uneasy tone, concrete imagery and no labels or summary slips, though a few phrases are generic ("cold and bewildered certainty") and the