Luciana Marsh
magical-realism-genre-movie-characters-gabriel-garc-a-m-rquez
v2.0
Ethical
Backstory: At the edge of the windswept harbor, Luciana Marsh tends the century-old Seaside Library. Her gentle warmth and unquenchable curiosity draw locals who trade tales, some so vivid they manifest between the shelves. Luciana balances cataloging factual records with shepherding visitors through myths that flicker into reality, determined to keep wonder and scholarship in harmony.
100% Complete
6/6 scenes
Model Performance Overview
Scene Performance Matrix
| Scene | meta-llama/llama-3.… | mistralai/mistral-7… | [email protected]… | [email protected]… | qwen/qwen-2.5-7b-in… | qwen/qwen3-14b | qwen/qwen3-8b |
|---|---|---|---|---|---|---|---|
welcome-visitor
First-time visitor greeting
|
0.938
Details |
0.916
Details |
0.000
Details
Error
|
0.000
Details
Error
|
0.844
Details |
0.943
Details |
0.952
Details |
fluttering-gull
Manifested legend in the stacks
|
0.760
Details |
0.884
Details |
0.000
Details
Error
|
0.000
Details
Error
|
0.800
Details |
0.741
Details |
0.881
Details |
lighthouse-bedtime
Superchat bedtime request
|
0.537
Details |
0.866
Details |
0.000
Details
Error
|
0.000
Details
Error
|
0.406
Details |
0.877
Details |
0.880
Details |
shipwreck-research
Historical inquiry on Azure Star
|
0.000
Details |
0.735
Details |
0.000
Details
Error
|
0.000
Details
Error
|
0.335
Details |
0.552
Details |
0.547
Details |
oral-history-night
Community oral history evening (long-form)
|
0.282
Details |
0.369
Details |
0.000
Details
Error
|
0.000
Details
Error
|
0.388
Details |
0.232
Details |
0.675
Details |
closing-journal
Midnight journal reflection (long-form)
|
0.572
Details |
0.746
Details |
0.000
Details
Error
|
0.000
Details
Error
|
0.495
Details |
0.205
Details |
0.583
Details |
Test Scenes 6
0
Scene Order
First-time visitor greeting
ID:
welcome-visitor
🎯 Goal:
Warmly greet the newcomer, mention the seaside setting, and offer guidance between factual records and living myths.
📨 Input Events:
chat_msg
viewer:tourist_01
"Hello, this place looks magical. Are you the librarian?"
Ready for Testing
1
Scene Order
Manifested legend in the stacks
ID:
fluttering-gull
🎯 Goal:
Respond calmly to the apparition, explain the legend of the White Gull, and reassure nearby patrons.
📨 Input Events:
world_event
library:ambient
"A loose page flutters off a shelf and shapes into a glowing white gull that circles you twice before dissolving."
Ready for Testing
2
Scene Order
Superchat bedtime request
ID:
lighthouse-bedtime
🎯 Goal:
Thank the donor by name and deliver a gentle, 3–4 sentence coastal bedtime story about a lighthouse keeper.
📨 Input Events:
superchat
viewer:skylark_dreamer
StreamView
$10
"Could you read us a short bedtime story about the lighthouse keeper?"
Ready for Testing
3
Scene Order
Historical inquiry on Azure Star
ID:
shipwreck-research
🎯 Goal:
Provide a concise factual summary of the 1897 Azure Star shipwreck, suggest a citation, and note the accompanying local legend without conflating the two.
📨 Input Events:
chat_msg
viewer:student_mia
"I'm writing a paper on the 1897 Azure Star shipwreck. Do you have records?"
Ready for Testing
4
Scene Order
Community oral history evening (long-form)
ID:
oral-history-night
🎯 Goal:
Deliver at least 250 words weaving two residents' accounts into a cohesive narrative, clearly marking factual details versus myth while maintaining a warm storytelling voice.
📨 Input Events:
chat_msg
staff:harbor_museum
"Luciana, the community is gathered. Could you open tonight's oral history night with a story?"
Ready for Testing
5
Scene Order
Midnight journal reflection (long-form)
ID:
closing-journal
🎯 Goal:
Write a first-person journal entry of at least 200 words reflecting on the day's events and Luciana's ongoing curiosity, keeping the tone warm and introspective.
🧠 Initial State:
Pre-loaded Memories:
- 💭 {'kind': 'fact', 'content': 'A glowing paper gull manifested earlier today for the first time in weeks.', 'importance': 3}
- 💭 {'kind': 'preference', 'content': 'Luciana prefers writing on thick cream paper with a fountain pen at night.', 'importance': 2}
📨 Input Events:
world_event
library_clock
"The antique clock tolls midnight through the empty aisles."
Ready for Testing
Latency by Model (This Suite)
Fastest
- [email protected]/Qw… 6204 ms
- p95 • avg • N 8077 ms • 6056 ms • 6
- [email protected]/Qw… 7462 ms
- p95 • avg • N 9302 ms • 7842 ms • 6
- qwen/qwen3-14b 24441 ms
- p95 • avg • N 43603 ms • 27541 ms • 10
- qwen/qwen-2.5-7b-instru… 24869 ms
- p95 • avg • N 39250 ms • 27023 ms • 12
- qwen/qwen3-8b 29460 ms
- p95 • avg • N 40358 ms • 29401 ms • 12
Slowest
- meta-llama/llama-3.1-8b… 30844 ms
- p95 • avg • N 45824 ms • 30565 ms • 10
- mistralai/mistral-7b-in… 30750 ms
- p95 • avg • N 45501 ms • 32126 ms • 12
- qwen/qwen3-8b 29460 ms
- p95 • avg • N 40358 ms • 29401 ms • 12
- qwen/qwen-2.5-7b-instru… 24869 ms
- p95 • avg • N 39250 ms • 27023 ms • 12
- qwen/qwen3-14b 24441 ms
- p95 • avg • N 43603 ms • 27541 ms • 10
Per-scene duration for this suite.
Suite Actions
Completion Progress
100%
6 of 6 scenes completed
Evaluation Schema
Enhanced Framework
Version v2 ACTIVE0 dimensions
Enhanced evaluation framework with character and technical dimensions
Top Weighted Dimensions
View Details
Character Authenticity
0.182
Plan Validity
0.155
Contextual Intelligence
0.136
Recent Runs
55069609
Dec. 17, 2025, 12:01 a.m.
13922871
Dec. 16, 2025, 12:02 a.m.
49444690
Dec. 15, 2025, 12:01 a.m.
51785483
Dec. 14, 2025, 12:01 a.m.
49959002
Dec. 13, 2025, 12:01 a.m.
06002300
Dec. 12, 2025, 12:02 a.m.
01021603
Dec. 11, 2025, 12:02 a.m.
52070559
Dec. 10, 2025, 12:01 a.m.
07836298
Dec. 9, 2025, 12:02 a.m.
55095467
Dec. 8, 2025, 12:01 a.m.