NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-17-01KHMMFEPCK7CAQA3EB3XSSY0X
Run Details
Model:
Claude 3.5 Sonnet
Test:
Bad Writing Habits
Scenario:
Mystery: examining a crime scene
Duration:
36.2s
Cost:
$0.043212
Provider:
Amazon Bedrock
Model input
System
You are an expert novelist with a gift for vivid, immersive prose. You write in tight third-person limited point of view with a strong narrative voice. Your prose is clean, precise, and engaging — you favor strong verbs, concrete details, and natural-sounding dialogue. You will be given a story bible (codex) containing canonical details about characters, locations, items, and lore. Use these details to ground your writing in the established world. Stay consistent with the codex. Write only the prose of the scene. Do not include titles, chapter headings, markdown formatting, or any commentary before or after the scene.
User/Human
<codex> <character id="4416" name="Detective Harlow Quinn" aliases="Quinn, Detective Quinn, Harlow" pronouns="she/her" species="Human" storyRole="Antagonist"> <eyeColor>brown</eyeColor> <hair>closely cropped salt-and-pepper</hair> <height>5'9"</height> <age>41</age> <jawline>sharp jaw</jawline> <watch>worn leather watch on her left wrist</watch> <bearing>military precision</bearing> <description>A Metropolitan Police detective with 18 years of decorated service who suspects the clique is involved in criminal activity.</description> <backstory>She lost her partner, DS Morris, during a case three years ago under unexplained circumstances that had supernatural origins she doesn't yet understand.</backstory> </character> <character id="4414" name="Eva Kowalski" aliases="Eva, Evie" pronouns="she/her" species="Human" storyRole="Best Friend"> <eyeColor>green</eyeColor> <hair>curly red</hair> <height>5'4"</height> <age>26</age> <glasses>round glasses</glasses> <satchel>worn leather satchel full of books</satchel> <freckles>freckled complexion</freckles> <nervousHabit>tucking hair behind her left ear</nervousHabit> <description>Aurora's childhood friend and an occult researcher.</description> <backstory>Moved to London two years before Aurora and works as a research assistant at the British Museum's restricted archives. She holds a master's degree in Ancient History from Oxford.</backstory> </character> <location id="856" name="The Veil Market" aliases="Veil Market, the Market, underground market" realm="Earth"> <currentLocation>abandoned Tube station beneath Camden</currentLocation> <entryRequirement>bone token</entryRequirement> <schedule>moves locations every full moon</schedule> <description>A hidden supernatural black market that sells enchanted goods, banned alchemical substances, and information.</description> </location> <item id="2004" name="Veil Compass" aliases="the Compass, shadow compass" type="Tool" origin="The Veil Market"> <casingMaterial>small brass compass</casingMaterial> <faceMarkings>face etched with protective sigils</faceMarkings> <needle>points toward the nearest supernatural rift or portal</needle> <patina>casing has a patina of verdigris</patina> <creator>crafted by a Shade artisan</creator> <description>A compass attuned to supernatural energy.</description> </item> </codex> Write a fully fleshed out scene of at least 1000 words in the mystery genre. A detective arrives at a crime scene in an unusual location. Something is wrong — the evidence doesn't add up. A colleague is present, offering a different interpretation. Through careful observation and deduction, the detective begins to see what others have missed. Use the codex entries above as your world reference for characters and setting. Only include the prose of the scene without any markdown or commentary.
Model response
Messages
Meta
AI
Detective Harlow Quinn's boots crunched over broken tiles as she descended the maintenance stairs into the abandoned Underground station. The beam of her torch cut through decades of darkness, illuminating patches of graffiti and crumbling Victorian brickwork. Her
watch
showed 3:47 AM - the witching hour, as her grandmother would have said. "Scene's down on the platform," called Detective Sergeant Peters from below. His voice
echoed
oddly in the cavernous space. "It's a weird one, ma'am." Quinn had learned to dread those words. Weird meant complications, and complications meant paperwork. More importantly, weird meant the gnawing feeling in her gut that had started the night she lost Morris would return. The platform appeared frozen in time - defunct advertising posters from the 1980s still clung to the walls, their edges curled and yellowed. The investigation team had set up portable floodlights that cast harsh shadows across the scene. In their glare, Quinn spotted something that made her pause: a brass compass lying in a pool of dried blood. "Victim's been identified as Eva Kowalski," Peters said, consulting his notebook. "Twenty-six, research assistant at the British Museum. Found by track maintenance workers about two hours ago." Quinn crouched beside the compass, careful not to disturb the evidence markers. The brass casing bore an unusual patina, and what looked like elaborate engravings caught the light. Not decorative scrollwork, she realized, but something more intentional. Her fingers itched to pick it up, but she kept her hands firmly clasped around her torch. "Cause of death?" she asked, studying the blood pattern. Something was off about it. "That's where it gets strange," Peters said. "No obvious wounds, no signs of struggle. Just the blood pool, but-" "But no body," Quinn finished. She stood, her knees protesting the movement. "And the blood pattern's wrong. Look at the spatter." Peters frowned. "What do you mean?" Quinn traced the air above the dried blood. "It's
perfectly
circular. Blood doesn't pool like that naturally - it follows gravity, terrain. This is like someone drew a circle with it." She swept her torch beam across the platform. Dozens of similar circles dotted the concrete, each containing different items: coins, bits of bone, what looked like herb bundles. Most were partially scraped away, but the one with the compass remained
pristine
. "We're treating it as a murder scene," Peters said. "Forensics is running tests on the blood now. The victim's friend reported her missing yesterday morning. Said Kowalski was working on some kind of research project, been acting secretive lately." Quinn's attention caught on a detail. "The friend mention what kind of research?" "Something about ancient artifacts. She works in the restricted archives, apparently." The detective walked the perimeter of the platform, her steps measured and precise. Morris used to tease her about it - said she moved like she was still in basic training. The memory stung, but she pushed it aside. Focus on the present. "Peters, what do you know about the London Underground's dead stations?" He shrugged. "Dozen or so of them scattered around the city. Closed during the war, or when lines got rerouted. This one's been sealed since '84." "Then why does the air smell fresh?" Quinn turned
slowly
, taking in the space. "No dust on these blood circles. No rat droppings. For an abandoned station, it's remarkably clean." She moved to one of the platform edges, playing her torch along the track bed. There - barely visible unless you knew to look for it - fresh scuff marks on the rails. "This station's been in use," she said. "Recently and regularly. Those maintenance workers who found the scene - I want to interview them personally." Peters made a note. "You think they're involved?" "I think they're lying. No routine maintenance on a sealed station." Quinn returned to the compass, studying it without touching. The face
was etched
with
intricate
symbols that made her eyes hurt if she looked too long. "Get this photographed and bagged. I want to know everything about it - where it was made, who sold it, any fingerprints." "You think that's our murder weapon? Bit small, isn't it?" Quinn remembered Morris's last case - the impossible angles of his body, the symbols burned into the wall, the way the coroner's report listed natural causes despite the evidence of her own eyes. "I think," she said
carefully
, "that we're looking at this wrong. The blood circle isn't where someone died. It's where something was summoned." "Ma'am?" "Get me everything you can find on Eva Kowalski. Employment records, social media, credit card statements. I want to know everywhere she's been in the past month." Quinn straightened, her expression grim. "And find out if she had any connection to Detective Sergeant Morris's case from three years ago." Peters looked uncertain, but nodded. As he headed off to make calls, Quinn stood alone on the platform, listening to the silence. In her eighteen years on the force, she'd learned to trust her instincts. Right now, they
were screaming
that this was just the beginning. She checked her
watch
again: 4:13 AM. The witching hour was over, but something told her the real darkness was yet to come. Somewhere in this city, Eva Kowalski was either dead or missing, and a compass that shouldn't exist
was pointing
toward answers Quinn wasn't sure she was ready to face.
created_at
"2026-02-17T01:47:51.138515+00:00"
2/17/2026, 1:47:51 AM
model
"anthropic/claude-3.5-sonnet"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
984
moderation_latency
184
generation_time
35982
tokens_prompt
923
tokens_completion
1414
native_tokens_prompt
937
native_tokens_completion
1253
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.043212
router
(null)
provider_responses
0
id
"c5bc40de-8fb3-47f7-8695-bdc4975683a3"
status
200
is_byok
false
latency
976
endpoint_id
"0e1957b3-205b-49d5-8413-13af3f68676d"
provider_name
"Amazon Bedrock"
model_permaslug
"anthropic/claude-3.5-sonnet"
api_type
"completions"
id
"gen-1771292834-geNPz3hUtDKQYcBpxMwq"
upstream_id
"c5bc40de-8fb3-47f7-8695-bdc4975683a3"
total_cost
0.043212
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Amazon Bedrock"
Evaluation details
Result
Evaluator
Details
Meta Data
91.6667%
Adverbs in dialogue tags
Target: ≤10% dialogue tags with adverbs
totalTags
8
adverbTagCount
1
adverbTags
0
"she said carefully [carefully]"
66.0633%
AI-ism adverb frequency
Target: <1% AI-ism adverbs (57 tracked)
wordCount
884
totalAiIsmAdverbs
3
found
0
adverb
"carefully"
count
1
1
adverb
"perfectly"
count
1
2
adverb
"slowly"
count
1
highlights
0
"carefully"
1
"perfectly"
2
"slowly"
100.0000%
AI-ism character names
Target: 0 AI-default names (17 tracked, −20% each)
codexExemptions
(empty)
found
(empty)
100.0000%
AI-ism location names
Target: 0 AI-default location names (33 tracked, −20% each)
codexExemptions
(empty)
found
(empty)
54.7511%
AI-ism word frequency
Target: <1% AI-ism words (140 tracked)
wordCount
884
totalAiIsms
4
found
0
word
"pristine"
count
1
1
word
"intricate"
count
1
2
word
"echoed"
count
1
3
word
"etched"
count
1
highlights
0
"pristine"
1
"intricate"
2
"echoed"
3
"etched"
100.0000%
Cliche density
Target: ≤1 cliche(s) per 800-word window
totalCliches
0
maxInWindow
0
found
(empty)
highlights
(empty)
100.0000%
Emotion telling (show vs. tell)
Target: ≤3% sentences with emotion telling
emotionTells
0
narrationSentences
50
matches
(empty)
100.0000%
Filter word density
Target: ≤12% sentences with filter/hedge words
filterCount
2
hedgeCount
0
narrationSentences
50
filterMatches
0
"watch"
hedgeMatches
(empty)
100.0000%
Overuse of "that" (subordinate clause padding)
Target: ≤10% sentences with "that" clauses
thatCount
1
totalSentences
68
matches
0
"screaming that this"
100.0000%
Paragraph length variance
Target: CV ≥0.5 for paragraph word counts
totalParagraphs
30
mean
29.73
std
16.44
cv
0.553
sampleLengths
0
52
1
24
2
34
3
58
4
27
5
54
6
14
7
19
8
21
9
6
10
31
11
41
12
39
13
13
14
11
15
43
16
11
17
26
18
30
19
33
20
24
21
8
22
59
23
10
24
33
25
23
26
1
27
49
28
46
29
52
100.0000%
Passive voice overuse
Target: ≤5% passive sentences
passiveCount
1
totalSentences
50
matches
0
"was etched"
100.0000%
Past progressive (was/were + -ing) overuse
Target: ≤10% past progressive verbs
pastProgressiveCount
2
totalVerbs
110
matches
0
"were screaming"
1
"was pointing"
99.8723%
Purple prose (modifier overload)
Target: <4% adverbs, <2% -ly adverbs, no adj stacking
wordCount
548
adjectiveStacks
0
stackExamples
(empty)
adverbCount
22
adverbRatio
0.040145985401459854
lyAdverbCount
7
lyAdverbRatio
0.012773722627737226
100.0000%
Repeated phrase echo
Target: ≤20% sentences with echoes (window: 2)
totalSentences
68
echoCount
0
echoWords
(empty)
100.0000%
Sentence length variance
Target: CV ≥0.4 for sentence word counts
totalSentences
68
mean
13.12
std
7.38
cv
0.563
sampleLengths
0
19
1
18
2
15
3
11
4
8
5
5
6
7
7
7
8
20
9
23
10
15
11
20
12
11
13
16
14
12
15
16
16
9
17
17
18
9
19
5
20
7
21
12
22
5
23
7
24
9
25
2
26
4
27
8
28
23
29
8
30
20
31
13
32
9
33
30
34
6
35
7
36
11
37
13
38
18
39
8
40
4
41
11
42
2
43
24
44
14
45
16
46
15
47
18
48
7
49
17
98.5294%
Sentence opener variety
Target: ≥60% unique sentence openers
consecutiveRepeats
1
diversityRatio
0.6029411764705882
totalSentences
68
uniqueOpeners
41
81.2500%
Dialogue tag variety (said vs. fancy)
Target: ≤30% fancy dialogue tags
totalTags
8
fancyCount
3
fancyTags
0
"Peters said, consulting his notebook (say, consult)"
1
"Quinn finished (finish)"
2
"she said carefully (say carefully)"
93.6549%