NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-29-01M3P2C6D8R5F20K735TPH2DYY
Run Details
Model:
Claude Sonnet 5.5 (Reasoning)
Test:
Bad Writing Habits
Scenario:
Mystery: examining a crime scene
Duration:
38.0s
Cost:
$0.039614
Provider:
Claude Platform on AWS
Evaluation summary
Score
Evaluator
Details Link
38.71%
Adverbs in dialogue tags
(details)
68.99%
AI-ism adverb frequency
(details)
100.00%
AI-ism character names
(details)
100.00%
AI-ism location names
(details)
89.66%
AI-ism word frequency
(details)
100.00%
Cliché density
(details)
100.00%
Emotion telling (show vs. tell)
(details)
89.29%
Filter word density
(details)
100.00%
Gibberish response detection
(details)
100.00%
Markdown formatting overuse
(details)
100.00%
Missing dialogue indicators (quotation marks)
(details)
16.67%
Name drop frequency
(details)
100.00%
Narrator intent-glossing
(details)
100.00%
"Not X but Y" pattern overuse
(details)
100.00%
Overuse of "that" (subordinate clause padding)
(details)
100.00%
Paragraph length variance
(details)
74.56%
Passive voice overuse
(details)
100.00%
Past progressive (was/were + -ing) overuse
(details)
100.00%
Em-dash & semicolon overuse
(details)
97.94%
Purple prose (modifier overload)
(details)
100.00%
Repeated phrase echo
(details)
100.00%
Sentence length variance
(details)
85.59%
Sentence opener variety
(details)
100.00%
Adverb-first sentence starts
(details)
100.00%
Pronoun-first sentence starts
(details)
85.00%
Subject-first sentence starts
(details)
0.00%
Subordinate conjunction sentence starts
(details)
100.00%
Technical jargon density
(details)
83.33%
Useless dialogue additions
(details)
100.00%
Dialogue tag variety (said vs. fancy)
(details)
87.6580%
Overall
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
The stairs at the old
Camden ventilation shaft had
been closed
to the public since 1974, which made the fresh bootprints in the dust the first thing Detective Harlow Quinn didn't like. She took the rungs
slowly
, one hand on the cold rail, her torch beam swinging ahead of her. The uniform at the bottom, a young constable named Achebe with a green
tinge
to his face, lifted the tape without
being asked
. "Platform's through there, ma'am. Forensics are waiting on the generator." "How long has he been down here?" "Doctor says not long. Hours, maybe."
Quinn grunted and stepped through
the arch.
The station had been abandoned
so thoroughly that it had almost stopped being a station. Curved tile walls rose on either side, the cream gone the colour of old teeth.
The rails were furred with
rust.
Someone had strung work lights
along the platform edge, and in their glare the dust lay like grey snow, except that it didn't lie properly.
It was scuffed and trampled
from one end of the platform to the other, in long looping paths that crossed and recrossed, as though a crowd had milled here for hours.
The body sat propped against
the far wall beneath a faded map of a line that no longer existed. Male, fifties, expensive overcoat, one shoe missing.
His head had tipped forward
as if he'd fallen asleep on a late train. "Detective."
The voice came from the
shadows beside the body. "Sorry. I didn't want to touch anything until you'd seen it."
Eva Kowalski stepped into the
light, a satchel dragging at her shoulder, round glasses catching the glare. Curly red hair had escaped its tie on one side, and as Quinn watched, she tucked a strand behind her left ear.
Then she did it again
a moment later, though nothing had come loose. "Ms. Kowalski," Quinn said. "I didn't realise the Met had started calling the British Museum to crime scenes." "The victim's name is Desmond Farrow. He's a registered dealer. He's been in and out of the restricted archives for years, and his card was in his wallet."
She offered a small smile
that didn't hold. "Your sergeant thought someone who knew his work might help." "Did he." "I'm a research assistant. I make photocopies, mostly." "You have a master's degree from Oxford." "Then I make
very
well-informed photocopies."
Quinn let that pass and
crouched beside the body.
Her watch, the worn leather
one that had been her father's, slid down her wrist as she reached out.
She checked the time out
of habit. Ten past two. Then she checked Farrow. No obvious wound.
The skin at the throat
was pale, the lips faintly blue. In his right hand, curled shut, he gripped something small, and a thread of green stained his fingers where the metal had pressed. "Brass," she said. "Old." "It's a compass."
Eva crouched too, keeping her
distance. "Or it was made to look like one. The casing's got verdigris all over it, and there are sigils
etched
on the face, protective ones, if I had to guess. Some collectors go in for that kind of thing. Occult curiosities." "Is that your professional opinion?" "It's my photocopier's opinion."
Quinn eased the fingers open
with a gloved hand.
The compass was no bigger
than a pocket
watch
. The
etched
symbols wound around the glass in tight concentric rings, and the needle
wasn't pointing
north.
It was quivering, agitated, and
it
was aimed
with absolute conviction at the tiled wall six feet to Quinn's left. There was nothing there. No door, no vent, only a mural of a smiling woman advertising a brand of tea that had gone out of business before Quinn was born. She tilted the compass. The needle swung back to the wall. "Magnetised something," Eva said
quickly
. "Iron in the tiles, maybe. Rail lines put out all sorts of interference." "The rails are behind me." "Then, a, a pipe." "Mm." Quinn set the compass on an evidence sheet and let her eyes travel over the scene the way she had
been trained
to, left to right, near to far, before she let herself form an opinion. Eighteen years had taught her that opinions were the enemy of observation. "Cause of death?" she asked. "Gas, I'd think," Eva said. "Old tunnels. Carbon monoxide pockets, methane. He came down here to look for something, inhaled something, sat down and died. It happens. There's a case from Holborn in the eighties." "That's a tidy theory." "It's a sad one." "It doesn't fit."
Quinn stood, knees cracking
. "Constable Achebe, the gas meter." Achebe hurried over, holding up a yellow handheld detector. "Clean readings all along the platform, ma'am. Oxygen's normal. We've been down two hours and nobody's felt a thing." "So the gas is
very
selective." Quinn turned in a slow circle. "Look at the floor." Eva looked. "Footprints." "Hundreds. Not one man's. Look at the pattern. These paths run in lines, and they stop at regular intervals, like queues."
She pointed with her torch
. "There. Four square marks in the dust, a metre apart. A stall, or a table. And there. And another. The whole platform's laid out like a street market." "Urban explorers," Eva said. "They throw parties in places like this." "Parties leave bottles. Cigarette ends. Glow sticks." Quinn walked to the nearest mark and knelt.
She touched the grit and
rubbed it between her fingers, then lifted it to her nose. "Cloves. Something burnt and sweet under it. And no litter. Not a scrap. Whoever stood here took every last thing when they left, and they did it in a hurry, because they missed this."
She held up a small
object she had found wedged in a crack in the tiles. It was white and smooth, about the size of a coin, with a hole bored through the centre. Bone. Eva went
very
still.
Her hand rose toward her
ear, then stopped, and came down to grip the satchel strap. "Do you know what that is?" Quinn asked. "No." "You answered fast." "I said no because I don't." Quinn studied her.
She had known Eva Kowalski
for less than a year, only as a name orbiting the girl Quinn actually wanted, the one Eva always
seemed to
be standing beside.
She'd never had reason to
think her a liar. Which was interesting, because she was a bad one. "Here's what bothers me," Quinn said quietly. "Farrow's shoes. Look at the one he's still wearing." Eva looked. "It's clean." "Immaculate. There's dust on everything down here, ankle-deep, and his sole is spotless. The heel isn't even scuffed. Add the missing shoe, and the fact that there's not one print anywhere around him, not one, and you get a man who didn't walk to this wall. He was carried. Or set down."
She crouched again, tapped the
overcoat's hem. "And he's dry. It rained heavily at midnight. Whoever brought him in from the street would have soaked him through. Instead he's been lying somewhere warm and sheltered until quite recently." "Then he was killed elsewhere and dumped," Eva said. "That still fits with—" "The compass?" Quinn shook her head. "A dead man's grip doesn't shut like that. Someone folded those fingers over the brass after he'd died. Someone wanted us to find it. And to think it meant something." The work lights buzzed.
Somewhere in the tunnel a
drip counted seconds. "There was a crowd here,"
Quinn went on, half to herself
. "Recently. Organised. Somebody cleared it in minutes and left one dead man behind, and what they left him with points at a wall that isn't a wall."
She walked to the mural
and pressed her palm flat against the tea-selling woman's painted face. The tile was warm.
Not warm like a pipe, but warm like skin
.
She drew her hand back
sharply
. Behind her, Eva said,
very
softly
, "Detective, I think we should go up." "Do you." "The air's not good." "The meter says otherwise." Quinn turned.
Her partner had died in
a place like this three years ago, in a room that had also made no sense, with evidence that had also refused to add up, and the file still lay in her desk drawer, unclosed.
She had learned a habit
from it. When a scene lied, look at who was most eager to leave. Eva's face was pale under the freckles. "Who told you to come down here, Ms. Kowalski?" Quinn asked. "My sergeant never called the museum. I checked the log on the way in." Eva opened her mouth. Behind her, on the evidence sheet, the little brass needle swung once,
slowly
, and pointed straight at her.
created_at
"2026-09-29T07:51:55.325Z"
9/29/2026, 7:51:55 AM
model
"anthropic/claude-sonnet-5.5-20260928"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3921
moderation_latency
(null)
generation_time
37673
tokens_prompt
896
tokens_completion
2402
native_tokens_prompt
1287
native_tokens_completion
3704
native_tokens_completion_images
(null)
native_tokens_reasoning
806
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"end_turn"
service_tier
"default"
usage
0.039614
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.21.0; linux; x64))"
http_referer
(null)
request_id
"req-1790668315-MuqIgBMCSTUfg2ylKzpD"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790668315-8owhCb4NxljeoMlEZB6c"
upstream_id
"msg_011CfXNFZQUDSGLfR9Js1i6N"
provider_responses
0
endpoint_id
"99aaad94-923b-4fc1-b763-271ed5486f7a"
id
"msg_011CfXNFZQUDSGLfR9Js1i6N"
is_byok
false
latency
693
model_permaslug
"anthropic/claude-sonnet-5.5-20260928"
provider_name
"Claude Platform on AWS"
status
200
total_cost
0.039614
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Claude Platform on AWS"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
38.71%
Adverbs in dialogue tags
Target: ≤10% dialogue tags with adverbs
totalTags
24
adverbTagCount
5
adverbTags
0
"Eva crouched too [too]"
1
"Eva said quickly [quickly]"
2
"Quinn said quietly [quietly]"
3
"She crouched again [again]"
4
"Behind her Eva said very [very]"
dialogueSentences
62
tagDensity
0.387
leniency
0.774
rawRatio
0.208
effectiveRatio
0.161
68.99%
AI-ism adverb frequency
Target: <2% AI-ism adverbs (58 tracked)
wordCount
1451
totalAiIsmAdverbs
9
found
0
adverb
"slowly"
count
2
1
adverb
"very"
count
4
2
adverb
"quickly"
count
1
3
adverb
"sharply"
count
1
4
adverb
"softly"
count
1
highlights
0
"slowly"
1
"very"
2
"quickly"
3
"sharply"
4
"softly"
100.00%
AI-ism character names
Target: 0 AI-default names (17 tracked, −20% each)
codexExemptions
(empty)
found
(empty)
100.00%
AI-ism location names
Target: 0 AI-default location names (33 tracked, −20% each)
codexExemptions
(empty)
found
(empty)
89.66%
AI-ism word frequency
Target: <2% AI-ism words (290 tracked)
wordCount
1451
totalAiIsms
3
found
0
word
"tinge"
count
1
1
word
"etched"
count
2
highlights
0
"tinge"
1
"etched"
100.00%
Cliché density
Target: ≤1 cliche(s) per 800-word window
totalCliches
0
maxInWindow
0
found
(empty)
highlights
(empty)
100.00%
Emotion telling (show vs. tell)
Target: ≤3% sentences with emotion telling
emotionTells
0
narrationSentences
80
matches
(empty)
89.29%
Filter word density
Target: ≤3% sentences with filter/hedge words
filterCount
2
hedgeCount
1
narrationSentences
80
filterMatches
0
"watch"
hedgeMatches
0
"seemed to"
100.00%
Gibberish response detection
Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words)
analyzedSentences
118
gibberishSentences
0
adjustedGibberishSentences
0
longSentenceCount
0
runOnParagraphCount
0
giantParagraphCount
0
wordSaladCount
0
repetitionLoopCount
0
controlTokenCount
0
repeatedSegmentCount
0
maxSentenceWordsSeen
60
ratio
0
matches
(empty)
100.00%
Markdown formatting overuse
Target: ≤5% words in markdown formatting
markdownSpans
0
markdownWords
0
totalWords
1451
ratio
0
matches
(empty)
100.00%
Missing dialogue indicators (quotation marks)
Target: ≤10% speech attributions without quotation marks
totalAttributions
21
unquotedAttributions
0
matches
(empty)
16.67%
Name drop frequency
Target: ≤1.0 per-name mentions per 100 words
totalMentions
43
wordCount
849
uniqueNames
9
maxNameDensity
2.36
worstName
"Quinn"
maxWindowNameDensity
4.5
worstWindowName
"Quinn"
discoveredNames
Camden
1
Detective
1
Harlow
1
Quinn
20
Achebe
2
Eva
14
Kowalski
2
Farrow
1
Eighteen
1
persons
0
"Harlow"
1
"Quinn"
2
"Achebe"
3
"Eva"
4
"Kowalski"
places
(empty)
globalScore
0.322
windowScore
0.167
100.00%
Narrator intent-glossing
Target: ≤2% narration sentences with intent-glossing patterns
analyzedSentences
53
glossingSentenceCount
0
matches
(empty)
100.00%
"Not X but Y" pattern overuse
Target: ≤1 "not X but Y" per 1000 words
totalMatches
1
per1kWords
0.689
wordCount
1451
matches
0
"Not warm like a pipe, but warm like skin"
100.00%
Overuse of "that" (subordinate clause padding)
Target: ≤2% sentences with "that" clauses
thatCount
1
totalSentences
118
matches
0
"let that pass"
100.00%
Paragraph length variance
Target: CV ≥0.5 for paragraph word counts
totalParagraphs
61
mean
23.79
std
23.64
cv
0.994
sampleLengths
0
32
1
41
2
10
3
7
4
6
5
7
6
92
7
40
8
21
9
52
10
18
11
46
12
2
13
8
14
7
15
6
16
42
17
37
18
4
19
50
20
5
21
4
22
85
23
11
24
18
25
5
26
4
27
49
28
5
29
35
30
4
31
4
32
12
33
28
34
16
35
3
36
54
37
11
38
65
39
35
40
20
41
8
42
1
43
3
44
6
45
50
46
16
47
4
48
90
49
13
74.56%
Passive voice overuse
Target: ≤2% passive sentences
passiveCount
7
totalSentences
80
matches
0
"been closed"
1
"being asked"
2
"been abandoned"
3
"were furred"
4
"was scuffed"
5
"was aimed"
6
"been trained"
100.00%
Past progressive (was/were + -ing) overuse
Target: ≤2% past progressive verbs
pastProgressiveCount
2
totalVerbs
156
matches
0
"wasn't pointing"
1
"was quivering"
100.00%
Em-dash & semicolon overuse
Target: ≤2% sentences with em-dashes/semicolons
emDashCount
0
semicolonCount
0
flaggedSentences
0
totalSentences
118
ratio
0
matches
(empty)
97.94%
Purple prose (modifier overload)
Target: <4% adverbs, <2% -ly adverbs, no adj stacking
wordCount
850
adjectiveStacks
0
stackExamples
(empty)
adverbCount
36
adverbRatio
0.042352941176470586
lyAdverbCount
13
lyAdverbRatio
0.015294117647058824
100.00%
Repeated phrase echo
Target: ≤20% sentences with echoes (window: 2)
totalSentences
118
echoCount
0
echoWords
(empty)
100.00%
Sentence length variance
Target: CV ≥0.4 for sentence word counts
totalSentences
118
mean
12.3
std
10.71
cv
0.871
sampleLengths
0
32
1
18
2
23
3
10
4
7
5
6
6
7
7
15
8
15
9
6
10
25
11
31
12
19
13
7
14
14
15
10
16
11
17
17
18
22
19
13
20
4
21
14
22
36
23
10
24
2
25
8
26
7
27
6
28
9
29
19
30
7
31
3
32
4
33
3
34
11
35
23
36
3
37
1
38
9
39
41
40
5
41
4
42
9
43
9
44
17
45
20
46
4
47
26
48
4
49
7
85.59%
Sentence opener variety
Target: ≥60% unique sentence openers
consecutiveRepeats
4
diversityRatio
0.5338983050847458
totalSentences
118
uniqueOpeners
63
100.00%
Adverb-first sentence starts
Target: ≥3% sentences starting with an adverb
adverbCount
3
totalSentences
68
matches
0
"Then she did it again"
1
"Then she checked Farrow."
2
"Somewhere in the tunnel a"
ratio
0.044
100.00%
Pronoun-first sentence starts
Target: ≤30% sentences starting with a pronoun
pronounCount
20
totalSentences
68
matches
0
"She took the rungs slowly,"
1
"It was scuffed and trampled"
2
"His head had tipped forward"
3
"She offered a small smile"
4
"Her watch, the worn leather"
5
"She checked the time out"
6
"It was quivering, agitated, and"
7
"She tilted the compass."
8
"She pointed with her torch"
9
"She touched the grit and"
10
"She held up a small"
11
"It was white and smooth,"
12
"Her hand rose toward her"
13
"She had known Eva Kowalski"
14
"She'd never had reason to"
15
"She crouched again, tapped the"
16
"She walked to the mural"
17
"She drew her hand back"
18
"Her partner had died in"
19
"She had learned a habit"
ratio
0.294
85.00%
Subject-first sentence starts
Target: ≤72% sentences starting with a subject
subjectCount
51
totalSentences
68
matches
0
"The stairs at the old"
1
"She took the rungs slowly,"
2
"The uniform at the bottom,"
3
"Quinn grunted and stepped through"
4
"The station had been abandoned"
5
"The rails were furred with"
6
"Someone had strung work lights"
7
"It was scuffed and trampled"
8
"The body sat propped against"
9
"His head had tipped forward"
10
"The voice came from the"
11
"Eva Kowalski stepped into the"
12
"She offered a small smile"
13
"Quinn let that pass and"
14
"Her watch, the worn leather"
15
"She checked the time out"
16
"The skin at the throat"
17
"Eva crouched too, keeping her"
18
"Quinn eased the fingers open"
19
"The compass was no bigger"
ratio
0.75
0.00%
Subordinate conjunction sentence starts
Target: ≥2% sentences starting with a subordinating conjunction
subConjCount
0
totalSentences
68
matches
(empty)
ratio
0
100.00%
Technical jargon density
Target: ≤6% sentences with technical-jargon patterns
analyzedSentences
27
technicalSentenceCount
0
matches
(empty)
83.33%
Useless dialogue additions
Target: ≤5% dialogue tags with trailing filler fragments
totalTags
24
uselessAdditionCount
2
matches
0
"Quinn stood, knees cracking"
1
"Quinn went on, half to herself"
100.00%
Dialogue tag variety (said vs. fancy)
Target: ≤10% fancy dialogue tags
totalTags
11
fancyCount
0
fancyTags
(empty)
dialogueSentences
62
tagDensity
0.177
leniency
0.355
rawRatio
0
effectiveRatio
0
87.6580%