NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M36NYWE4TK5VGCPR88JJ7R18
Run Details
Model:
GPT-6 Luna
Test:
Bad Writing Habits
Scenario:
Mystery: examining a crime scene
Duration:
34.6s
Cost:
$0.001459
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Adverbs in dialogue tags
(details)
100.00%
AI-ism adverb frequency
(details)
100.00%
AI-ism character names
(details)
100.00%
AI-ism location names
(details)
90.50%
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)
16.29%
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)
99.97%
Passive voice overuse
(details)
100.00%
Past progressive (was/were + -ing) overuse
(details)
100.00%
Em-dash & semicolon overuse
(details)
94.00%
Purple prose (modifier overload)
(details)
100.00%
Repeated phrase echo
(details)
100.00%
Sentence length variance
(details)
45.79%
Sentence opener variety
(details)
0.00%
Adverb-first sentence starts
(details)
100.00%
Pronoun-first sentence starts
(details)
18.28%
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)
85.4946%
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
“Don’t step on the chalk,”
Harlow Quinn said
.
The uniformed constable froze with
one boot raised above the white line. Beyond it, the platform sloped into a black tunnel that should not have existed. “Which chalk?” he asked. Quinn pointed.
The line crossed the station
tiles, climbed the wall and continued across the ceiling, a child’s drawing of a horizon. At its centre, a man lay face down with his arms stretched towards the tunnel mouth.
The abandoned station beneath Camden
had no place on any current Tube map.
Its name had been chiselled
off the entrance arch, but the old enamel signs still read CAMDEN ROAD. Dust coated the ticket barriers. Rust curled from the rails. Yet the body’s coat lay clean against the platform, and the blood beneath him shone wet. Quinn stepped over the chalk. “Detective,”
the constable said
. “If the line mattered, he’d have drawn it around himself.” She crouched beside the body, keeping her shoes clear of the blood.
The dead man wore a
grey suit, expensive but poorly fitted.
His right hand clutched a
brass token the size of a sixpence.
A bone had been carved
into its face. Quinn looked at the tunnel.
A faint blue light pulsed
beyond the curve, then vanished. “Who found him?” “Market security,” said Sergeant Pell.
He stood by the ticket
barrier, gloved hands tucked under his arms. “They called it in at half past nine. Said he’d been dead an hour.” “Market security,” Quinn repeated. Pell’s mouth twitched. “They don’t give names.” “Convenient.” “They don’t give anything. Not even a statement.” A scuff sounded behind her. Quinn turned.
Eva Kowalski stood at the
barrier, a worn leather satchel hanging from one shoulder.
Her round glasses had slipped
low on her nose. Curly red hair escaped the clip at the back of her head, and she tucked a loose strand behind her left ear. “You’re not on the investigation,” Quinn said. “I’m not on anything official.” Eva raised both hands, palms out. “I was already here.” “Doing what?” “Looking for a bookstall.” Pell gave a short laugh. “In a sealed Tube station?” Eva glanced at him. “The market moves. The bookstall doesn’t.” Quinn stood. “You know what happened here?” “I know he shouldn’t be here.” “Dead men rarely choose their stations.” “He didn’t choose this one.” Eva’s gaze went to the brass token. “That’s a bone token. It gets you through the Market’s entrance. He was a customer.” “Or a seller.” “Either way, he knew the rules.” Quinn studied her, then the chalk line. “What rules?” “Don’t cross a boundary you didn’t draw.” Pell unfolded his arms. “He’s lying on the other side of one.” “Which is why I said he knew the rules.” Quinn crouched again. The dead man’s fingers curled tight around the token. No blood marked his hand.
His sleeve bore a tear
at the elbow, but the skin beneath it looked unbroken. A thin black thread looped around his wrist, tucked under the shirt cuff. “Pell, photograph the hand before anyone touches it.” “Already done.” “Then do it again from the other side.” Pell sighed, but moved. Quinn followed the trail of blood.
It spread under the man’s
chest, then ran towards the platform edge in narrow channels. Not a pool. A fan. The pattern should have followed the slope of the tiles, but the blood climbed the grout lines in three thin branches, each one ending at the chalk. She leaned closer. Copper stung her nose. The blood had a sharp, medicinal smell beneath it, like vinegar on a cut. “Cause of death?” she asked. “Knife,” Pell said. “There’s a wound between the shoulders.” Quinn glanced at the body. “You turned him?” “Just enough to check.” “You moved the body.” “He was dead.” “Now we don’t know how he fell.” “He didn’t fall. Someone stabbed him.” “Stab wounds don’t explain a clean front.” Pell’s
eyes narrowed
. “You think he was murdered somewhere else?” “I think we don’t know.” Eva crouched near the chalk without crossing it. “The blood’s wrong.” Pell shifted towards her. “You’re a researcher. Not a pathologist.” “Correct. I can still see it’s running uphill.” “It’s on a slope.” “No, it isn’t.” Eva pointed at the tiled floor. “The platform dips towards the tracks. The blood runs away from them.” Quinn checked the grout. Eva was right. A drop had gathered against the raised edge of a tile, then travelled towards the wall. “Could be a draft,” Pell said. “Or someone poured it.” “Blood doesn’t climb,” Quinn replied. “It’s not blood,” Eva said. Quinn looked at her. “You just called it blood.” “I called it wrong.” Pell shook his head. “This is a crime scene, not an antiques fair.” Eva’s fingers went to her satchel strap. “Then stop treating the chalk like decoration.” Quinn shifted her attention to the line.
Its edge looked too crisp
for chalk. No powder had scattered over the dust. The white mark had
been laid
down with a narrow brush, stroke after careful stroke. At the wall, it bent around a rusted conduit and continued onto the ceiling without a break. She stood beneath it. A drop of water hung from the ceiling on the far side, trembling. It fell. The drop struck the tiles and rolled towards the line, then stopped as though it had hit glass. Pell stared. Quinn drew a pen from her pocket and nudged the drop.
It slid along the invisible
edge, gathering dust, then rolled back the way it had come. “Don’t touch it,” Eva said. Quinn withdrew the pen. “You know what it is.” “A ward. Or the remains of one.” “Wards don’t stop water.” “This one did.” “Why?” Eva looked past Quinn, towards the tunnel. “To keep something in.” Pell reached for his radio. “I’m calling for a specialist.” The radio hissed. A woman’s voice broke through, thin and warped. “Sergeant Pell. The doors are open. You need to leave the platform.” Pell lowered the radio. “Who was that?” No one answered. The tunnel’s blue light appeared again. This time it held, a small bright point at the curve. Quinn’s
watch
ticked against the cuff of her coat. The old leather strap had softened over years of wear. She checked it by habit, then noticed the second hand had stopped. “Pell, get the uniforms back to the stairs.” He didn’t move. “What about the body?” “Leave it.” “Detective, we can’t abandon a scene because a light
flickered
.” “I’m not abandoning it. I’m moving people out.” Eva rose. “He can’t cross the line.” Quinn looked at the body. “Why not?” “The ward cuts the platform in two. Something on the tunnel side can’t pass through it.” “Then the killer’s on the platform with us.” “Not necessarily.” Pell tapped his radio.
It answered with a burst
of static. “I’ve got no signal.” A metal clack rang out from the tunnel. The body’s fingers tightened around the token. Quinn saw it. So did Eva. Pell took one step back. “Did you see that?” The man’s hand had gone still again. Quinn approached, careful not to cross the chalk.
She watched the skin at
his knuckles. The fingers had not tightened on their own. The token had turned within his grip, its carved bone face now pointed towards the tunnel. “Could be muscle contraction,” Pell said. “Could be,” Quinn replied. “But it wasn’t.” Eva took a small book from her satchel.
Its cover had cracked down
the spine.
She flipped pages with a
thumb stained by ink. “What are you doing?” Quinn asked. “Checking a reference.” “Now?” “Unless you’d rather wait for the next movement.” “Find the bit about the ward.” Eva’s breath caught. “It’s not a ward. It’s a threshold mark. The dead can cross it. The living can’t, unless someone opens it from the other side.” “Someone?” “Or something.” Pell gave a strained laugh. “You expect us to believe a line of chalk can stop the dead?” “It isn’t chalk,” Eva said. “It’s ash and bone dust, mixed with iron.” Quinn glanced at the white mark. “And the man crossed it.” “No. He’s on the near side.” The body lay a handspan beyond the line. Quinn’s eyes tracked from his outstretched fingers to the mark.
His index finger had stopped
just short of touching it. The blood, however, had crossed.
It had climbed the tiles
and spread into the ash in three narrow branches. “His blood opened the threshold,” she said. Eva turned a page. “Possibly.” “You said it keeps something in.” “It did.” Pell stared at the tunnel mouth. “What’s coming out?” The blue point swelled. Something scraped along the rails, metal against metal, though no train could run on them. The sound dragged closer, then stopped. Quinn studied the blood again. Three branches. Three cuts?
She leaned towards the dead
man’s back. His coat had a
single tear
, but the fabric showed no spreading stain. The wound had opened at the centre of his shoulder blades. The blood beneath him came from the wound, yet no blood marked the coat. “Pell. Did you see the wound before you moved him?” “Yes.” “Was the tear there?” He hesitated. “Sergeant.” “I thought it was.” “You thought.” “I saw blood. I lifted the coat. The tear could’ve been there.” Quinn looked at the man’s cuffs. No dust clung to them. No grime. But the soles of his shoes carried a grey residue, the same fine powder that coated the station. “He came through the entrance,” she said. “He walked across this platform.” Eva shook her head. “There’s no entrance on this side.” “The stairs?” “End at a bricked-up wall.” Pell took a torch from his belt and aimed it at the body’s shoes. “That’s station dust.” “Then why is the rest of him clean?” “Someone cleaned him.” “After he died?” Quinn bent over the dead man again. A faint ring marked the skin around his left thumb. The sort of stain left by a tight ring worn for years, now gone. She checked the other hand. No jewellery. “Search his pockets.” Pell crouched and began with the jacket. His fingers found a folded paper packet, a wallet, and a phone with a cracked screen. He laid them on the tiles. Quinn opened the wallet. No cards. No cash. One photograph, creased through the centre, showed a younger version of the dead man standing outside a museum entrance beside a woman whose face someone had scratched away. “Name?” Pell asked. “Nothing.” Eva leaned in. “That’s not a museum.” Quinn checked the stone arch behind the figures. “You’re sure?” “It’s a station entrance. See the old London Underground roundel? That one came down before the war.” “Which station?” Eva’s gaze moved to the dead man. “The one we’re standing in.” A second metal clack sounded. Closer. Quinn put the photograph away and lifted the folded packet. The paper felt damp. Inside, a pinch of black ash clung to the crease. Beneath it lay a narrow brass object, no larger than a pocket
watch
. Its casing carried a green patina. Protective sigils edged the face. The needle spun once, then settled towards the tunnel. “The Veil Compass,” Eva whispered. Quinn kept it in her gloved palm. “What does it point to?” “A rift. A portal. The nearest one.” “Not the tunnel?” “Not unless the tunnel is the nearest.” The needle gave a small jerk. It pointed past the tunnel, towards the body. Quinn turned the compass over. A dark hair had caught beneath its casing. She lifted the body’s collar with two fingers. The black thread around his wrist ran beneath the suit sleeve and up his back, where it vanished into the wound. “He wasn’t stabbed,” she said. “Someone threaded him.” Pell’s face drained. “Threaded him with what?” “Something sharp. From the inside.” The dead man’s chest rose. Pell
lurched
backwards. Eva dropped the book. Quinn grabbed the corpse’s shoulder. The movement came again, a slow lift beneath the suit. A wet click sounded from the wound. The black thread pulled taut. The threshold line split. The blue light surged down the tunnel. At the same moment, the compass needle snapped around and pointed at Quinn. The body’s eyes opened. “Detective,” Eva said. Quinn drew her service weapon. The dead man’s mouth worked. A strand of black thread slipped between his teeth, followed by a bone token that struck the tile and spun towards the broken line. Pell fired his torch at it. The beam went out. Something hit the far side of the ward. Once. Twice. On the third blow, the chalk line burst into white dust.
created_at
"2026-09-23T08:26:19.467Z"
9/23/2026, 8:26:19 AM
model
"openai/gpt-6-luna-20260922"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
2733
moderation_latency
(null)
generation_time
34563
tokens_prompt
4492
tokens_completion
3395
native_tokens_prompt
4157
native_tokens_completion
2835
native_tokens_completion_images
0
native_tokens_reasoning
0
native_tokens_cached
4154
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
"completed"
service_tier
"default"
usage
0.00145934
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790151979-Ez7HxV3E5LeyiJE0xuBw"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790151979-drjROzCD0sHMMWkc3hF2"
upstream_id
"resp_085dd3bbe87ddfd2016ab38d2ba80887d19705394703e5a95d"
provider_responses
0
endpoint_id
"05e94e02-b9c2-4bbb-ba55-4082ee9ad687"
id
"resp_085dd3bbe87ddfd2016ab38d2ba80887d19705394703e5a95d"
is_byok
false
latency
1593
model_permaslug
"openai/gpt-6-luna-20260922"
provider_name
"OpenAI"
status
200
total_cost
0.00145934
cache_discount
0.00037386
upstream_inference_cost
0
provider_name
"OpenAI"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Adverbs in dialogue tags
Target: ≤10% dialogue tags with adverbs
totalTags
24
adverbTagCount
0
adverbTags
(empty)
dialogueSentences
135
tagDensity
0.178
leniency
0.356
rawRatio
0
effectiveRatio
0
100.00%
AI-ism adverb frequency
Target: <2% AI-ism adverbs (58 tracked)
wordCount
2105
totalAiIsmAdverbs
0
found
(empty)
highlights
(empty)
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)
90.50%
AI-ism word frequency
Target: <2% AI-ism words (290 tracked)
wordCount
2105
totalAiIsms
4
found
0
word
"pulsed"
count
1
1
word
"flickered"
count
1
2
word
"single tear"
count
1
3
word
"lurched"
count
1
highlights
0
"pulsed"
1
"flickered"
2
"single tear"
3
"lurched"
100.00%
Cliché density
Target: ≤1 cliche(s) per 800-word window
totalCliches
1
maxInWindow
1
found
0
label
"eyes widened/narrowed"
count
1
highlights
0
"eyes narrowed"
100.00%
Emotion telling (show vs. tell)
Target: ≤3% sentences with emotion telling
emotionTells
0
narrationSentences
199
matches
(empty)
100.00%
Filter word density
Target: ≤3% sentences with filter/hedge words
filterCount
2
hedgeCount
0
narrationSentences
199
filterMatches
0
"watch"
hedgeMatches
(empty)
100.00%
Gibberish response detection
Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words)
analyzedSentences
309
gibberishSentences
0
adjustedGibberishSentences
0
longSentenceCount
0
runOnParagraphCount
0
giantParagraphCount
0
wordSaladCount
0
repetitionLoopCount
0
controlTokenCount
0
repeatedSegmentCount
0
maxSentenceWordsSeen
28
ratio
0
matches
(empty)
100.00%
Markdown formatting overuse
Target: ≤5% words in markdown formatting
markdownSpans
0
markdownWords
0
totalWords
2105
ratio
0
matches
(empty)
100.00%
Missing dialogue indicators (quotation marks)
Target: ≤10% speech attributions without quotation marks
totalAttributions
25
unquotedAttributions
0
matches
(empty)
16.29%
Name drop frequency
Target: ≤1.0 per-name mentions per 100 words
totalMentions
89
wordCount
1421
uniqueNames
7
maxNameDensity
2.67
worstName
"Quinn"
maxWindowNameDensity
4
worstWindowName
"Quinn"
discoveredNames
Quinn
38
Camden
1
Tube
1
Sergeant
1
Pell
24
Kowalski
1
Eva
23
persons
0
"Quinn"
1
"Camden"
2
"Sergeant"
3
"Pell"
4
"Kowalski"
5
"Eva"
places
(empty)
globalScore
0.163
windowScore
0.333
100.00%
Narrator intent-glossing
Target: ≤2% narration sentences with intent-glossing patterns
analyzedSentences
100
glossingSentenceCount
0
matches
(empty)
100.00%
"Not X but Y" pattern overuse
Target: ≤1 "not X but Y" per 1000 words
totalMatches
0
per1kWords
0
wordCount
2105
matches
(empty)
100.00%
Overuse of "that" (subordinate clause padding)
Target: ≤2% sentences with "that" clauses
thatCount
0
totalSentences
309
matches
(empty)
100.00%
Paragraph length variance
Target: CV ≥0.5 for paragraph word counts
totalParagraphs
174
mean
12.1
std
12.58
cv
1.04
sampleLengths
0
8
1
26
2
4
3
38
4
57
5
5
6
4
7
10
8
43
9
15
10
3
11
31
12
4
13
7
14
1
15
8
16
52
17
7
18
15
19
2
20
4
21
10
22
10
23
7
24
6
25
6
26
27
27
3
28
6
29
9
30
7
31
12
32
9
33
45
34
8
35
2
36
8
37
4
38
53
39
21
40
5
41
9
42
8
43
4
44
4
45
3
46
7
47
6
48
7
49
10
99.97%
Passive voice overuse
Target: ≤2% passive sentences
passiveCount
3
totalSentences
199
matches
0
"been chiselled"
1
"been carved"
2
"been laid"
100.00%
Past progressive (was/were + -ing) overuse
Target: ≤2% past progressive verbs
pastProgressiveCount
0
totalVerbs
249
matches
(empty)
100.00%
Em-dash & semicolon overuse
Target: ≤2% sentences with em-dashes/semicolons
emDashCount
0
semicolonCount
0
flaggedSentences
0
totalSentences
309
ratio
0
matches
(empty)
94.00%
Purple prose (modifier overload)
Target: <4% adverbs, <2% -ly adverbs, no adj stacking
wordCount
1421
adjectiveStacks
1
stackExamples
0
"faint blue light pulsed"
adverbCount
30
adverbRatio
0.0211118930330753
lyAdverbCount
2
lyAdverbRatio
0.0014074595355383533
100.00%
Repeated phrase echo
Target: ≤20% sentences with echoes (window: 2)
totalSentences
309
echoCount
0
echoWords
(empty)
100.00%
Sentence length variance
Target: CV ≥0.4 for sentence word counts
totalSentences
309
mean
6.81
std
4.67
cv
0.686
sampleLengths
0
8
1
12
2
14
3
4
4
2
5
20
6
16
7
13
8
18
9
5
10
5
11
16
12
5
13
4
14
10
15
12
16
11
17
12
18
8
19
5
20
10
21
3
22
5
23
12
24
14
25
4
26
3
27
4
28
1
29
8
30
5
31
2
32
14
33
9
34
22
35
7
36
11
37
4
38
2
39
4
40
5
41
5
42
4
43
6
44
2
45
5
46
6
47
6
48
12
49
15
45.79%
Sentence opener variety
Target: ≥60% unique sentence openers
consecutiveRepeats
13
diversityRatio
0.26537216828478966
totalSentences
309
uniqueOpeners
82
0.00%
Adverb-first sentence starts
Target: ≥3% sentences starting with an adverb
adverbCount
0
totalSentences
163
matches
(empty)
ratio
0
100.00%
Pronoun-first sentence starts
Target: ≤30% sentences starting with a pronoun
pronounCount
27
totalSentences
163
matches
0
"Its name had been chiselled"
1
"She crouched beside the body,"
2
"His right hand clutched a"
3
"He stood by the ticket"
4
"Her round glasses had slipped"
5
"His sleeve bore a tear"
6
"It spread under the man’s"
7
"She leaned closer."
8
"Its edge looked too crisp"
9
"She stood beneath it."
10
"It slid along the invisible"
11
"She checked it by habit,"
12
"He didn’t move."
13
"It answered with a burst"
14
"She watched the skin at"
15
"Its cover had cracked down"
16
"She flipped pages with a"
17
"His index finger had stopped"
18
"It had climbed the tiles"
19
"She leaned towards the dead"
ratio
0.166
18.28%
Subject-first sentence starts
Target: ≤72% sentences starting with a subject
subjectCount
144
totalSentences
163
matches
0
"Harlow Quinn said"
1
"The uniformed constable froze with"
2
"The line crossed the station"
3
"The abandoned station beneath Camden"
4
"Its name had been chiselled"
5
"Dust coated the ticket barriers."
6
"Rust curled from the rails."
7
"Quinn stepped over the chalk."
8
"the constable said"
9
"She crouched beside the body,"
10
"The dead man wore a"
11
"His right hand clutched a"
12
"A bone had been carved"
13
"Quinn looked at the tunnel."
14
"A faint blue light pulsed"
15
"He stood by the ticket"
16
"Pell’s mouth twitched."
17
"A scuff sounded behind her."
18
"Eva Kowalski stood at the"
19
"Her round glasses had slipped"
ratio
0.883
0.00%
Subordinate conjunction sentence starts
Target: ≥2% sentences starting with a subordinating conjunction
subConjCount
0
totalSentences
163
matches
(empty)
ratio
0
100.00%
Technical jargon density
Target: ≤6% sentences with technical-jargon patterns
analyzedSentences
51
technicalSentenceCount
0
matches
(empty)
100.00%
Useless dialogue additions
Target: ≤5% dialogue tags with trailing filler fragments
totalTags
24
uselessAdditionCount
0
matches
(empty)
100.00%
Dialogue tag variety (said vs. fancy)
Target: ≤10% fancy dialogue tags
totalTags
21
fancyCount
2
fancyTags
0
"Quinn repeated (repeat)"
1
"Eva whispered (whisper)"
dialogueSentences
135
tagDensity
0.156
leniency
0.311
rawRatio
0.095
effectiveRatio
0.03
85.4946%