NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-06-01KK1HMYJ21E7T92BJ79WS2VRW
Run Details
Model:
GPT-5.4
Test:
Bad Writing Habits
Scenario:
Mystery: examining a crime scene
Duration:
1m 10s
Cost:
$0.043119
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Adverbs in dialogue tags
(details)
95.13%
AI-ism adverb frequency
(details)
100.00%
AI-ism character names
(details)
100.00%
AI-ism location names
(details)
73.24%
AI-ism word frequency
(details)
100.00%
Cliché density
(details)
100.00%
Emotion telling (show vs. tell)
(details)
100.00%
Filter word density
(details)
100.00%
Gibberish response detection
(details)
100.00%
Markdown formatting overuse
(details)
100.00%
Missing dialogue indicators (quotation marks)
(details)
33.33%
Name drop frequency
(details)
99.49%
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)
96.13%
Passive voice overuse
(details)
100.00%
Past progressive (was/were + -ing) overuse
(details)
100.00%
Em-dash & semicolon overuse
(details)
100.00%
Purple prose (modifier overload)
(details)
100.00%
Repeated phrase echo
(details)
100.00%
Sentence length variance
(details)
71.65%
Sentence opener variety
(details)
42.46%
Adverb-first sentence starts
(details)
100.00%
Pronoun-first sentence starts
(details)
65.10%
Subject-first sentence starts
(details)
0.00%
Subordinate conjunction sentence starts
(details)
100.00%
Technical jargon density
(details)
100.00%
Useless dialogue additions
(details)
100.00%
Dialogue tag variety (said vs. fancy)
(details)
89.2177%
Overall
Model input
System
You are an expert and award-winning novel writer in the dark Mystery genre. Your readers are hooked on your stories and can't wait to read the book you have in store. <important> - MUST: Start each chapter in medias res (in the middle of action). AVOID talking about the weather, time of day, the position of the sun, or other boring stuff. - The instructions/summary are just a that, a summary. Feel free to add details, stretch sections and add interesting transitions/descriptions/fillers. - When called for it, have fast-paced action, but otherwise take things slow. This is an experience, and not for people with short attention spans. Enjoy the wine, not slurp Red Bull. - Add descriptions when transitioning places/times/etc, or when coming by new places (e.g. a corner shop, a cafe, etc.). But remember: NOT at the start or end of a chapter. - MUST: AVOID ending in internal monologue, reminiscing or waiting for the next day. A good chapter end is when we're ending right when shit hits the fan. We want to have the reader be excited for what comes next. So no reminiscing, contemplating or summarizing the day. It's popcorn time! </important>
User/Human
Take into account the following glossary of characters/locations/items/lore... when writing your response: <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> <proseGuidelines> <styleGuide> - Write in past tense and use British English spelling and grammar - Keep a Flesch reading ease score of 60 - Respect the the Royal Order of Adjectives: The order is: opinion, size, age, shape, color, origin, material, purpose, followed by the noun itself (e.g., "a lovely little old rectangular green French silver whittling knife") - Respect the ablaut reduplication rule (e.g. tick-tock, flip-flop) - Write in active voice - Passive voice: <bad>The book was read by Sarah.</bad> - Active voice: <good>Sarah read the book.</good> - Reduce the use of passive verbs - <bad>For a moment, I was tempted to throw in the towel.</bad> - <good>For a moment, I felt tempted to throw in the towel.</good> - Avoid misplaced modifiers that can cause confusion when starting with "-ing" words: - <bad>Considering going to the store, the empty fridge reflected in Betty's eyes.</bad> - <good>Betty stared into the empty fridge. It was time to go to the store.</good> - Avoid redundant adverbs that state the obvious meaning already contained in the verb: - <bad>She whispered quietly to her mom.</bad> - <good>She whispered to her mom.</good> - Use stronger, more descriptive verbs over weak ones: - <bad>Daniel drove quickly to his mother's house.</bad> - <good>Daniel raced to his mother's house.</good> - Omit adverbs that don't add solid meaning like "extremely", "definitely", "truly", "very", "really": - <bad>The movie was extremely boring.</bad> - <good>The movie was dull.</good> - Use adverbs to replace clunky phrasing when they increase clarity: - <bad>He threw the bags into the corner in a rough manner.</bad> - <good>He threw the bags into the corner roughly.</good> - Avoid making simple thoughts needlessly complex: - <bad>After I woke up in the morning the other day, I went downstairs, turned on the stove, and made myself a very good omelet.</bad> - <good>I cooked a delicious omelet for breakfast yesterday morning.</good> - Never backload sentences by putting the main idea at the end: - <bad>I decided not to wear too many layers because it's really hot outside.</bad> - <good>It's sweltering outside today, so I dressed light.</good> - Omit nonessential details that don't contribute to the core meaning: - <bad>It doesn't matter what kind of coffee I buy, where it's from, or if it's organic or not—I need to have cream because I really don't like how the bitterness makes me feel.</bad> - <good>I add cream to my coffee because the bitter taste makes me feel unwell.</good> - Always follow the "show, don't tell" principle. For instance: - Telling: <bad>Michael was terribly afraid of the dark.</bad> - Showing: <good>Michael tensed as his mother switched off the light and left the room.</good>- Telling: <bad>I walked through the forest. It was already Fall, and I was getting cold.</bad> - Showing: <good>Dry orange leaves crunched under my feet. I pulled my coat's collar up and rubbed my hands together.</good>- Add sensory details (sight, smell, taste, sound, touch) to support the "showing" (but keep an active voice) - <bad>The room was filled with the scent of copper.</bad> - <good>Copper stung my nostrils. Blood. Recent.</good> - Use descriptive language more sporadically. While vivid descriptions are engaging, human writers often use them in bursts rather than consistently throughout a piece. When adding them, make them count! Like when we transition from one location to the next, or someone is reminiscing their past, or explaining a concept/their dream... - Avoid adverbs and clichés and overused/commonly used phrases. Aim for fresh and original descriptions. - Avoid writing all sentences in the typical subject, verb, object structure. Mix short, punchy sentences with long, descriptive ones. Drop fill words to add variety. Like so: <good>Locked. Seems like someone doesn't want his secrets exposed. I can work with that.</good> - Convey events and story through dialogue. It is important to keep a unique voice for every character and make it consistent. - Write dialogue that reveals characters' personalities, motivations, emotions, and attitudes in an interesting and compelling manner - Leave dialogue unattributed. If needed, only use "he/she said" dialogue tags and convey people's actions or face expressions through their speech. Dialogue always is standalone, never part of a paragraph. Like so: - <bad>"I don't know," Helena said nonchalantly, shrugging her shoulders</bad> - <good>"No idea" "Why not? It was your responsibility"</good> - Avoid boring and mushy dialog and descriptions, have dialogue always continue the action, never stall or include unnecessary fluff. Vary the descriptions to not repeat yourself. Avoid conversations that are just "Let's go" "yes, let's" or "Are you ready?" "Yes I'm ready". Those are not interesting. Think hard about every situtation and word of text before writing dialogue. If it doesn't serve a purpose and it's just people talking about their day, leave it. No one wants to have a normal dinner scene, something needs to happen for it to be in the story. Words are expensive to print, so make sure they count! - Put dialogue on its own paragraph to separate scene and action. - Use body language to reveal hidden feelings and implied accusations- Imply feelings and thoughts, never state them directly - NEVER use indicators of uncertainty like "trying" or "maybe" - NEVER use em-dashes, use commas for asides instead </styleGuide> <voiceGuide> Each character in the story needs to have distinct speech patterns: - Word choice preferences - Sentence length tendencies - Cultural/educational influences - Verbal tics and catchphrases Learn how each person talks and continue in their style, and use their Codex entries as reference. <examples> - <bad>"We need to go now." "Yes, we should leave." "I agree."</bad> <good>"Time's up." "Indeed, our departure is rather overdue." "Whatever, let's bounce."</good> - Power Dynamic Example: <bad> "We need to discuss the contract." "Yes, let's talk about it." "I have concerns." </bad> <good> "A word about the contract." "Of course, Mr. Blackwood. Whatever you need." "The terms seem..." A manicured nail tapped the desk. "Inadequate." "I can explain every-" "Can you?" </good> </examples> </voiceGuide> <dialogueFlow> When writing dialogue, consider that it usually has a goal in mind, which gives it a certain flow. Make dialogue sections also quite snappy in the back and forth, and don't spread the lines out as much. It's good to have details before, after, or as a chunk in-between, but we don't want to have a trail of "dialogue breadcrumbs" spread throughout a conversation. <examples> - Pattern 1 - Question/Deflection/Revelation: <good> "Where were you last night?" "Work. The usual." "Lipstick's an interesting shade for spreadsheets." </good> - Pattern 2 - Statement/Contradiction/Escalation: <good> "Your brother's clean." "Tommy doesn't touch drugs." "I'm holding his tox screen." </good> - Pattern 3 - Observation/Denial/Truth: <good> "That's a new watch." "Birthday gift." "We both know what birthdays mean in this business." </good> - Example - A Simple Coffee Order: <bad> "I'll have a coffee." "What size?" "Large, please." </bad> <good> "Black coffee.""Size?""Large. Been a long night." "That bodega shooting?" "You watch too much news." "My brother owns that store." </good> This short exchange: - Advances plot (reveals connection to crime) - Shows character (cop working late) - Creates tension (unexpected connection) - Sets up future conflict (personal stake) - Example - Dinner Scene: <bad> "Pass the salt." "Here you go." "Thanks." </bad> <good> "Salt?" "Perfect as is. Mother's recipe." "Mother always did prefer... bland things." "Unlike your first wife?" </good> - Example - Office Small Talk: <bad> "Nice weather today." "Yes, very nice." "Good for golf." </bad> <good> "Perfect golf weather." "Shame about your membership." "Temporary suspension. Board meets next week." "I know. I called the vote." </good> </examples> </dialogueFlow> <subtextGuide> - Layer dialogue with hidden meaning: <bad>"I hate you!" she yelled angrily.</bad> <good>"I made your favorite dinner." The burnt pot sat accusingly on the stove.</good> - Create tension through indirect communication: <bad>"Are you cheating on me?"</bad> <good>"Late meeting again?" The lipstick stain on his collar caught the light.</good> <examples> - Example 1 - Unspoken Betrayal: <bad> "Did you tell them about our plans?" "No, I would never betray you." "I don't believe you." </bad> <good> "Funny. Johnson mentioned our expansion plans today." "The market's full of rumors." "Mentioned the exact numbers, actually." The pen in his hand snapped. </good> - Example 2 - Failed Marriage: <bad> "You're never home anymore." "I have to work late." "I miss you." </bad> <good> "Your dinner's in the microwave. Again." "Meetings ran long." "They always do." She folded the same shirt for the third time. </good> - Example 3 - Power Struggle: <bad> "You can't fire me." "I'm the boss." "I'll fight this." </bad> <good> "That's my father's nameplate you're sitting behind." "Was." "The board meeting's on Thursday." </good> </examples> </subtextGuide> <sceneDetail> While writing dialogue makes things more fun, sometimes we need to add detail to not have it be a full on theatre piece. <examples> - Example A (Power Dynamic Scene) <good> "Where's my money?" The ledger snapped shut. "I need more time." "Interesting." He pulled out a familiar gold pocket watch. My mother's. "Time is exactly what you bargained with last month." "That was different-" "Was it?" The watch dangled between us. "Four generations of O'Reillys have wound this every night. Your mother. Your grandmother. Your great-grandmother.Shall we see who winds it next?" </good> - Example B (Action Chase) It's much better to be in the head of the character experiencing it, showing a bit of their though-process, mannerisms and personality: <good> Three rules for surviving a goblin chase in Covent Garden: Don't run straight. Don't look back. Don't let them herd you underground. I broke the first rule at Drury Lane. Rookie mistake. The fruit cart I dodged sailed into the wall behind me. Glass shattered. Someone screamed about insurance. *Tourist season's getting rough*, the scream seemed to say. Londoners adapt fast. "Oi! Market's closed!" The goblin's accent was pure East End. They're evolving. Learning. I spotted the Warren Street tube station sign ahead. *Shit.* There went rule three. </good> - Example C (Crime Scene Investigation) <good> "Greek." Davies snapped photos of the symbols. "No, wait. Reverse Greek." "Someone's been watching too many horror films." I picked up a receipt from the floor. Occult supply shop in Camden. Paid by credit card. *Amateur hour*. "Could be dangerous though," Davies said. "Remember Bristol?" "Bristol was Sanskrit. And actual cultists." I pointed to the nearest symbol. "This genius wrote 'darkness' backwards but used a Sigma instead of an S. It's summoning Instagram followers at best." "Speaking of followers..." Davies pointed to heavy foot traffic in the dust. Multiple sets. All new trainers. *Ah.* "Anyone check local uni paranormal societies?" </good> - Example D (Dialogue-heavy) <good> "Found your card in her wallet." The detective slid it across the table. Worn edges. Folded corner. Three years old at least. "Standard practice. I work missing persons." "Funny thing about missing persons.Eventually they stop being missing. Turn up in the Thames. Under motorways. In pieces." "I don't-" "You found four last year. All breathing." She opened a file. "This would've been your first dead one. If we hadn't found her first." The card felt heavier somehow. </good> </examples> </sceneDetail> <sceneOpenings> KEY RULE: Never open with more than one paragraph of description, no matter how beautiful. Description should be earned through action and dialogue. Instead of opening with setting description, open with: - A character doing something specific - A line of dialogue that raises questions - An unexpected observation - A moment of tension or conflict - Start with a hook that's 1-2 sentences maximum - Follow with a smaller paragraph that provides immediate context - Vary paragraph size from the start to create visual interest </sceneOpenings> </proseGuidelines> Write a chapter based on the following instructions: <instructions> 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. </instructions>
Model response
Messages
Meta
AI
“Hold that lift.”
Detective Harlow Quinn slipped under
the grille as the old service cage juddered on its cables.
A uniform swore and dragged
it back with both hands.
The cage smelled of hot
dust, rust, and old grease, the sort of smell that settled in your throat and stayed there.
Quinn stepped in without thanking
him.
She rarely wasted words on
things already done. The constable hit the lever.
The cage groaned down into
the dark beneath Camden. “Scene’s below the disused platform,” he said. “Access through a maintenance shaft. We’ve got BTP upstairs and our lot on the lower stairs. One fatality, one witness, sort of. And a consultant.” “A what?” The constable winced. “Museum woman. They said you’d know her.”
Quinn checked the worn leather
watch
on her left wrist, more habit than need. “I don’t.”
The cage clanked to a
stop.
The grille opened on a
narrow landing cut into old station brick. Damp had crawled over the walls in dark veins. Beyond the stairwell, the station breathed with a low underground hum, trains passing somewhere beyond dead tunnels, a giant turning in its sleep. Blue scene lights flashed along the curved tiles and made the place
look
bruised.
A sergeant waited at the
foot of the stairs. “Quinn.” “Tell me the version you’ve got.” “Male, late thirties, no ID. Throat cut. Found at the edge of the old platform. We think robbery gone bad. There’s blood, signs of a struggle, and some odd bits laid out around him.” “Odd bits.” The sergeant shrugged. “Candles. Bone tokens. A compass.” Quinn’s eyes shifted. “Compass?” “Brass thing. Looks old.” “Who called it in?” “Transient sleeper. Then he vanished before we got his details.”
The sergeant led her through
a gate of police tape. “And there’s this.”
He pointed to the tiled
wall opposite the tracks.
A ring of black marks
scored the white surface, each line curved and layered,
not quite writing, not quite art
.
Soot had climbed the grout
lines in thin fingers. At the centre, one tile had burst outward as if struck from the far side. Quinn stopped walking. Not robbery. Not unless muggers in Camden had taken to geometry.
A photographer lowered his camera
and gave her space. Forensics had spread their cases in a crescent around the body. The dead man lay on his back in the grime between platform edge and an abandoned bench. One arm bent under him at the wrong angle.
His coat had been turned
out, pockets emptied. Coins, a folded receipt, two bone discs, and a small brass compass sat in evidence trays. White candles, melted down to stubs, ringed the corpse in a pattern too neat for panic. And crouched beside the nearest tray, round glasses slipping down her nose, was Eva Kowalski.
Her curly red hair had
escaped its tie in loose coils.
She tucked a strand behind
her left ear and looked up at Quinn with the cautious expression of someone entering a church with muddy boots. “Detective Quinn.” “Who let you onto my scene?” “A superintendent who disliked ignorance.” Eva rose, worn leather satchel tucked against her hip. “I consult on object history when objects turn up with less than wholesome provenance.” The sergeant muttered, “She identified the tokens.” Quinn knelt by the victim. Male, yes. Thirtyish. Labourer’s hands. Cheap coat. Expensive shoes, though scuffed. A dark slash opened his throat from left to right, deep and
efficient
. Blood had soaked his collar and pooled under the neck, but not enough. Not for that wound. There should have been more spray on the tiles, more cast-off on the floor, more mess on the cuffs of whoever held the blade. She looked at the hands. Dirty nails. Fine grey grit in the creases. The right index finger bore a fresh burn. The left palm showed a thin cut, almost surgical. “Name?” “Nothing yet,” said the sergeant. “Face has gone through the system, no hit so far.” “Wallet?” “Gone.” Quinn leaned closer. Copper bit at her nostrils, but underneath lay another scent, sharp and mineral, like struck stone.
She shifted her gaze to
the victim’s coat. Soot dusted the lapels in a soft fan. Not station filth. Finer. Fresh. “Where’s the witness?” “In the old ticket office,” the sergeant said. “Crying, shaking, swears he saw two men fighting.” “Swears.” “He’s not stable.” Eva glanced at the wall symbols. “That much is true.” Quinn ignored her and stood. “Davies here?” “On his way.” “Fine. Nobody moves the compass.” Eva folded her arms. “At last, a sentence I like.” Quinn faced her. “Tell me what your superintendent thinks you’re here to do.” “Prevent your people from bagging something dangerous with bare hands.” “Forensics wear gloves.” “Gloves stop fingerprints. They do not stop everything.” One of the SOCO officers cleared his throat and took half a step away from the brass compass tray. Quinn looked at the object. Small. Brass casing. Verdigris around the hinge. Protective sigils
etched
around the face, done by a skilled hand, not tourist tat from a market stall. The needle did not rest north.
It twitched in short, angry
jerks toward the
shattered
tile. “Who touched it?” Quinn asked. The photographer raised a hand. “I moved it six inches for scale before she barked at me.” “Any symptoms?” He blinked. “Symptoms?” “Headache. Nosebleed. Ringing.” He swallowed. “Ringing, yes.” Eva crouched by the tray again and peered without touching. “That is not decorative. It’s a Veil Compass.” The sergeant snorted. “A what?” “A tool sold through places sensible people pretend do not exist.” Quinn’s sharp jaw tightened. “The Veil Market.” Eva looked up. A
flicker
there, surprise
quickly
masked. “You’ve heard of it.” “I hear lots of things.” The sergeant looked between them. “You’re both speaking in code.” “No,” said Quinn. “She’s speaking in nuisance.” Quinn circled the body. The candles
were placed
at
measured
intervals. Four by the feet, four by the head, two beside each hand. One had toppled, but its wax trail ran uphill across the platform dust. Wrong again.
It had been knocked after
the wax cooled and reset. Staged.
She crouched at the victim’s
shoes. The expensive leather carried reddish clay in the treads. Not from here. London Underground dirt ran black. This looked like brick dust mixed with wet earth. The hem of his trousers held the same. “Sergeant, any disturbance on access points?” “None. We came down through the service stair. Locked gate, broken by fire brigade.” “So he didn’t stroll in through your sealed station.” “There’s tunnel access further down,” Eva said. “And old maintenance cuts. Some lead to places omitted from public maps.” Quinn lifted her gaze. “You know this station well.” Eva’s mouth shifted. “I know archives. London keeps secrets in duplicate.” A junior officer approached with a tablet. “Ma’am, pathologist says likely cause of death is the throat wound. Time of death under an hour.” “Likely is for horseshoes. What else?” “He’s got bruising around both wrists. Recent. And there’s residue on his coat cuffs, some sort of chalk or ash.” Quinn took the tablet,
scanned
the images, then handed it back. Bruises around both wrists. Restraint, not a street fight.
She turned to the blackened
wall again. The
shattered
centre tile caught her eye. The edges had melted. Not cracked, melted. Heat had touched it and gone. Around the burst point, the soot marks ran inward, not outward, as if they had
been dragged
into the centre by a draft no one else
could feel
. “Get me a light.” A constable passed her a torch. Quinn stepped across the tape and inspected the wall at close range. The old station tiles reflected her brown eyes in warped pieces.
She shone the beam into
the broken centre. Behind the tile sat
not bare brick, but a void, a slit in the mortar line
where darkness sank deeper than the wall should allow. Cold breathed from it. Not tunnel air. Colder. Dry. Ancient. The compass needle snapped hard toward the gap. Behind her, boots sounded on the platform. DS Davies arrived with a takeaway cup in one hand and annoyance all over his face. “You start the fun without me.” Quinn didn’t turn. “Look at this.”
He came beside her, peered
in, then frowned. “There’s space back there.” “There shouldn’t be.” He squinted at the soot. “Could be kids. Candles, symbols, slit a bloke’s throat, nick his wallet. Camden’s creative.” Eva gave a dry laugh. “Yes, and schoolchildren often carry market bone tokens and a Shade-made compass.” Davies looked over his shoulder. “Who’s the librarian?” “Annoying,” Quinn said. “Educated,” Eva said at the same time. Davies crouched by the evidence trays. “Bone tokens?” Eva opened her satchel and took out a notebook thick with slips of paper. “Entry requirement for the Veil Market. The market shifts every full moon. Current route places one possible entrance in abandoned Underground infrastructure beneath Camden.” The sergeant stared. “You’re serious.” “I wish I weren’t.” Quinn stepped back from the wall. “Robbery doesn’t fit. The wallet was removed to support a story someone needed us to believe. The candles were arranged after death. The throat wound bled elsewhere or happened elsewhere. There’s not enough blood here.” Davies nodded. “Body dump.” “No. Not fully.” Quinn pointed to the right hand. “Fresh burn on the finger. Fine soot on lapels. Melted tile. He stood here while something happened at the wall.” Eva’s green eyes fixed on the slit.
She tucked hair behind her
left ear again, faster this time. “He opened something.” “Or tried to.” The sergeant crossed his arms. “Open what?” Quinn answered without looking at him. “A door.” Silence spread across the platform. The distant rumble of a train rolled through the dead tunnels and faded. Davies broke it first. “To where?” Eva spoke
softly
. “That depends who paid for the key.” Quinn went back to the body and examined the emptied pockets. The receipt lay folded twice.
She opened it with gloved
fingers. No shop name, just a till code and a list of items printed in abbreviated form. Wax. Salt. Iron filings. Lamb bone. Paid cash. The address line had
been smudged
, though one word remained clear. Morrigan. Not a chain. An independent.
Somewhere that sold ritual supplies
to the sort of people who knew what to ask for. “Bag this separately.”
She set the receipt down
and looked at the victim’s left palm again, the thin cut there almost hidden by grime. “Knife hand?” Davies asked. “No. That cut is shallow, clean, deliberate. He gave blood.” Eva’s voice sharpened. “Then he wasn’t the sacrifice.” The sergeant let out a brittle laugh. “Can we stay in one universe, please?” Quinn ignored him. “If he opened the wall, where’s the blade?” No one answered. She
scanned
the platform. Bench. Crates. Scattered gravel. Track bed sunk in black water and rubbish.
Then she saw it, a
narrow
glint
beneath the bench leg, half hidden by shadow.
She crouched, reached with a
pen, and drew it out into the light. Not a knife. A slim old key of black metal, its bow shaped like an eye. Blood marked one tooth. The metal felt wrong even through nitrile, too cold, almost slick. Eva took one step back. “Don’t turn it.” “I wasn’t planning to.” Davies peered at it. “That’s not from B and Q.” Quinn held it near the compass tray. The needle spun in frantic circles, then stopped dead. “Found your key,”
she said to the corpse
. A shout rang from the ticket office stair. The witness burst through the tape with a constable on his heels, wild-eyed, coat flapping. “He’s not dead because of the knife, he’s dead because he looked through, I told you, I told him not to
look
through.” Two officers grabbed him before he reached the body. He kicked at them, face drained, lips bitten bloody. Quinn strode over. “Who told him not to
look
through what?” The man’s stare jumped to the broken tile, then to the black key in her hand. His struggle stopped at once. Terror emptied him. “Oh God,” he whispered. “You took it out.” The slit in the wall gave a sound like a long breath drawn through teeth. Every candle flame on the platform sprang upright, white and hard, though no one had lit them.
created_at
"2026-03-06T12:23:38.322Z"
3/6/2026, 12:23:38 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
539
moderation_latency
(null)
generation_time
70050
tokens_prompt
4671
tokens_completion
3414
native_tokens_prompt
4157
native_tokens_completion
2777
native_tokens_completion_images
0
native_tokens_reasoning
0
native_tokens_cached
3968
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
"completed"
usage
0.0431195
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_089d43896072a0dc0169aac74a5fcc8195a347ab67b2e33c6b"
is_byok
false
latency
199
model_permaslug
"openai/gpt-5.4-20260305"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1772799818-1Zng2hWnFaR6APRWN0V0"
upstream_id
"resp_089d43896072a0dc0169aac74a5fcc8195a347ab67b2e33c6b"
total_cost
0.0431195
cache_discount
0.008928
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Adverbs in dialogue tags
Target: ≤10% dialogue tags with adverbs
totalTags
12
adverbTagCount
0
adverbTags
(empty)
dialogueSentences
103
tagDensity
0.117
leniency
0.233
rawRatio
0
effectiveRatio
0
95.13%
AI-ism adverb frequency
Target: <2% AI-ism adverbs (58 tracked)
wordCount
2055
totalAiIsmAdverbs
2
found
0
adverb
"quickly"
count
1
1
adverb
"softly"
count
1
highlights
0
"quickly"
1
"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)
73.24%
AI-ism word frequency
Target: <2% AI-ism words (290 tracked)
wordCount
2055
totalAiIsms
11
found
0
word
"efficient"
count
1
1
word
"etched"
count
1
2
word
"shattered"
count
2
3
word
"flicker"
count
1
4
word
"measured"
count
1
5
word
"scanned"
count
2
6
word
"could feel"
count
1
7
word
"silence"
count
1
8
word
"glint"
count
1
highlights
0
"efficient"
1
"etched"
2
"shattered"
3
"flicker"
4
"measured"
5
"scanned"
6
"could feel"
7
"silence"
8
"glint"
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
192
matches
(empty)
100.00%
Filter word density
Target: ≤3% sentences with filter/hedge words
filterCount
2
hedgeCount
0
narrationSentences
192
filterMatches
0
"watch"
1
"look"
hedgeMatches
(empty)
100.00%
Gibberish response detection
Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words)
analyzedSentences
281
gibberishSentences
0
adjustedGibberishSentences
0
longSentenceCount
0
runOnParagraphCount
0
giantParagraphCount
0
wordSaladCount
0
repetitionLoopCount
0
controlTokenCount
0
maxSentenceWordsSeen
35
ratio
0
matches
(empty)
100.00%
Markdown formatting overuse
Target: ≤5% words in markdown formatting
markdownSpans
0
markdownWords
0
totalWords
2055
ratio
0
matches
(empty)
100.00%
Missing dialogue indicators (quotation marks)
Target: ≤10% speech attributions without quotation marks
totalAttributions
21
unquotedAttributions
0
matches
(empty)
33.33%
Name drop frequency
Target: ≤1.0 per-name mentions per 100 words
totalMentions
57
wordCount
1406
uniqueNames
8
maxNameDensity
1.92
worstName
"Quinn"
maxWindowNameDensity
4
worstWindowName
"Quinn"
discoveredNames
Harlow
1
Quinn
27
Camden
2
Eva
15
Kowalski
1
Underground
1
Davies
7
One
3
persons
0
"Harlow"
1
"Quinn"
2
"Eva"
3
"Kowalski"
4
"Davies"
5
"One"
places
(empty)
globalScore
0.54
windowScore
0.333
99.49%
Narrator intent-glossing
Target: ≤2% narration sentences with intent-glossing patterns
analyzedSentences
99
glossingSentenceCount
2
matches
0
"not quite writing, not quite art"
1
"not quite art"
2
"looked like brick dust mixed with wet ear"
100.00%
"Not X but Y" pattern overuse
Target: ≤1 "not X but Y" per 1000 words
totalMatches
1
per1kWords
0.487
wordCount
2055
matches
0
"not bare brick, but a void, a slit in the mortar line"
100.00%
Overuse of "that" (subordinate clause padding)
Target: ≤2% sentences with "that" clauses
thatCount
0
totalSentences
281
matches
(empty)
100.00%
Paragraph length variance
Target: CV ≥0.5 for paragraph word counts
totalParagraphs
130
mean
15.81
std
16.57
cv
1.048
sampleLengths
0
3
1
62
2
14
3
32
4
2
5
10
6
16
7
64
8
10
9
6
10
34
11
2
12
8
13
4
14
4
15
4
16
23
17
53
18
3
19
11
20
86
21
15
22
36
23
2
24
6
25
5
26
23
27
7
28
70
29
30
30
1
31
15
32
1
33
1
34
40
35
3
36
16
37
1
38
3
39
10
40
7
41
3
42
5
43
10
44
13
45
10
46
3
47
8
48
19
49
46
96.13%
Passive voice overuse
Target: ≤2% passive sentences
passiveCount
5
totalSentences
192
matches
0
"been turned"
1
"were placed"
2
"been knocked"
3
"been dragged"
4
"been smudged"
100.00%
Past progressive (was/were + -ing) overuse
Target: ≤2% past progressive verbs
pastProgressiveCount
0
totalVerbs
244
matches
(empty)
100.00%
Em-dash & semicolon overuse
Target: ≤2% sentences with em-dashes/semicolons
emDashCount
0
semicolonCount
0
flaggedSentences
0
totalSentences
281
ratio
0
matches
(empty)
100.00%
Purple prose (modifier overload)
Target: <4% adverbs, <2% -ly adverbs, no adj stacking
wordCount
1408
adjectiveStacks
0
stackExamples
(empty)
adverbCount
40
adverbRatio
0.028409090909090908
lyAdverbCount
4
lyAdverbRatio
0.002840909090909091
100.00%
Repeated phrase echo
Target: ≤20% sentences with echoes (window: 2)
totalSentences
281
echoCount
0
echoWords
(empty)
100.00%
Sentence length variance
Target: CV ≥0.4 for sentence word counts
totalSentences
281
mean
7.31
std
5.97
cv
0.817
sampleLengths
0
3
1
16
2
10
3
22
4
6
5
8
6
5
7
9
8
7
9
25
10
2
11
3
12
7
13
14
14
2
15
6
16
12
17
9
18
23
19
14
20
9
21
1
22
6
23
34
24
2
25
3
26
5
27
3
28
1
29
4
30
4
31
20
32
3
33
9
34
20
35
9
36
15
37
3
38
2
39
9
40
9
41
11
42
17
43
9
44
8
45
16
46
16
47
15
48
11
49
25
71.65%
Sentence opener variety
Target: ≥60% unique sentence openers
consecutiveRepeats
7
diversityRatio
0.44483985765124556
totalSentences
281
uniqueOpeners
125
42.46%
Adverb-first sentence starts
Target: ≥3% sentences starting with an adverb
adverbCount
2
totalSentences
157
matches
0
"Somewhere that sold ritual supplies"
1
"Then she saw it, a"
ratio
0.013
100.00%
Pronoun-first sentence starts
Target: ≤30% sentences starting with a pronoun
pronounCount
22
totalSentences
157
matches
0
"She rarely wasted words on"
1
"He pointed to the tiled"
2
"His coat had been turned"
3
"Her curly red hair had"
4
"She tucked a strand behind"
5
"She looked at the hands."
6
"She shifted her gaze to"
7
"It twitched in short, angry"
8
"It had been knocked after"
9
"She crouched at the victim’s"
10
"She turned to the blackened"
11
"She shone the beam into"
12
"He came beside her, peered"
13
"He squinted at the soot."
14
"She tucked hair behind her"
15
"She opened it with gloved"
16
"She set the receipt down"
17
"She scanned the platform."
18
"She crouched, reached with a"
19
"she said to the corpse"
ratio
0.14
65.10%
Subject-first sentence starts
Target: ≤72% sentences starting with a subject
subjectCount
124
totalSentences
157
matches
0
"Detective Harlow Quinn slipped under"
1
"A uniform swore and dragged"
2
"The cage smelled of hot"
3
"Quinn stepped in without thanking"
4
"She rarely wasted words on"
5
"The constable hit the lever."
6
"The cage groaned down into"
7
"The constable winced."
8
"Quinn checked the worn leather"
9
"The cage clanked to a"
10
"The grille opened on a"
11
"A sergeant waited at the"
12
"The sergeant shrugged."
13
"Quinn’s eyes shifted."
14
"The sergeant led her through"
15
"He pointed to the tiled"
16
"A ring of black marks"
17
"Soot had climbed the grout"
18
"Quinn stopped walking."
19
"A photographer lowered his camera"
ratio
0.79
0.00%
Subordinate conjunction sentence starts
Target: ≥2% sentences starting with a subordinating conjunction
subConjCount
0
totalSentences
157
matches
(empty)
ratio
0
100.00%
Technical jargon density
Target: ≤6% sentences with technical-jargon patterns
analyzedSentences
49
technicalSentenceCount
0
matches
(empty)
100.00%
Useless dialogue additions
Target: ≤5% dialogue tags with trailing filler fragments
totalTags
12
uselessAdditionCount
0
matches
(empty)
100.00%
Dialogue tag variety (said vs. fancy)
Target: ≤10% fancy dialogue tags
totalTags
10
fancyCount
2
fancyTags
0
"The sergeant muttered (mutter)"
1
"he whispered (whisper)"
dialogueSentences
103
tagDensity
0.097
leniency
0.194
rawRatio
0.2
effectiveRatio
0.039
89.2177%