NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-25-01KQ2H8RRF192B5328FE8K8PXJ
Run Details
Model:
DeepSeek V4 Flash
Test:
Codex Red Herring (False Positive Detection)
Scenario:
Short text (~524 words), small codex (11 entries)
Duration:
9.8s
Cost:
$0.000252
Provider:
DeepSeek
Evaluation summary
Score
Evaluator
Details Link
0.00%
Correct "no violations" response
(details)
14.29%
No hallucinated violations
(details)
7.1429%
Overall
Model input
System
You are a meticulous continuity editor for a fiction project. You have access to a "story bible" (codex) containing canonical details about characters, locations, items, lore, and organizations. Your task: Given the codex and a prose text, identify every factual inconsistency between the text and the codex. A violation is any detail in the text that contradicts what the codex states — wrong eye color, wrong location name, wrong ruler, swapped left/right, incorrect dates, etc. Be thorough. Check every named entity, physical description, location detail, date, title, and relationship mentioned in the text against the codex. Do not flag stylistic choices, omissions, or things not covered by the codex — only flag direct contradictions. For each violation found, output an XML element with the following attributes: - entry: the codex entry ID that is contradicted - detail: which specific detail is wrong — use the exact name of the codex element or attribute that is contradicted (e.g. "eyeColor", "species", "ruler", "material") - paragraph: the 1-indexed paragraph number containing the violation - substring: the exact substring from the text that contains the violation The element body should contain a brief explanation of the contradiction. Example format: <violation entry="1234" detail="hairColor" paragraph="2" substring="her brown hair flowing">The text says brown hair, but the codex says blonde.</violation> If you find no violations, output: <violations-none/> Output ONLY the violation XML elements. Do not include any other text, preamble, or explanation outside the XML.
User/Human
<codex> <character id="100" name="Torben Reikhart" aliases="Stray"> <gender>Male</gender> <eyeColor>Green</eyeColor> <hairColor>Silver</hairColor> </character> <character id="101" name="Sable Dunmore"> <species>Human</species> <gender>Female</gender> <age>28</age> <eyeColor>Brown</eyeColor> <hairColor>Auburn</hairColor> </character> <character id="102" name="Old Rivka"> <gender>Female</gender> </character> <character id="103" name="Petal Dunmore"> <gender>Female</gender> </character> <character id="104" name="Jurren Dunmore"> <gender>Male</gender> </character> <location id="105" name="The Rusty Lantern"/> <location id="106" name="Dunmore"> <terrain>Coastal</terrain> </location> <item id="107" name="Kindling"> <material>Dark iron</material> <objectType>Weapon</objectType> <power>Fire</power> </item> <item id="108" name="Petal's Locket"> <material>Silver</material> <objectType>Jewelry</objectType> </item> <lore id="109" name="The Felling"> <timePeriod>Two centuries ago</timePeriod> <category>War</category> </lore> <lore id="110" name="The Kindling Rite"> <category>Tradition</category> </lore> </codex> <text> The Rusty Lantern sat crooked on its foundations, listing toward the harbor like a drunk leaning on a post. Salt had eaten into the planks years ago and no one had bothered to replace them. The ceiling beams were black with smoke. Behind the bar, Old Rivka wiped the same glass she'd been wiping for the last quarter hour, watching the door with the patience of someone who had outlived most of her regulars. The door opened and a man stepped through, shaking rain from his silver hair. Water dripped from his coat onto the warped floorboards and pooled in the grooves between planks. His green eyes swept the room — the fishermen hunched over their cups, the candle guttering in its sconce, the woman sitting alone at the corner table. He smelled of wet wool and the cold air off the harbor. Old Rivka raised her chin. "Stray," she called across the room. "Didn't think you'd show." He didn't answer. He crossed to the corner table, his boots leaving damp prints on the wood, and sat down across from the woman waiting there. Sable Dunmore was twenty-eight years old, human, and tired of waiting. Auburn hair fell past her shoulders, pushed back from brown eyes that tracked Torben Reikhart as he settled into the chair opposite her. On the table between them lay a sword — Kindling, she called it. A dark iron blade, the metal faintly warm even in the chill of the tavern. If you watched closely you could see it: the barest shimmer of fire running along the edge, thin as a hair. At her neck hung a silver locket. Petal's Locket, though she never called it that aloud. "It still has her hair inside," she said, touching it once. "That's all I have left of her." Torben waited. He'd learned when to be quiet. "My sister Petal," Sable said. "She was fourteen when The Felling reached Dunmore. You know what that was — the war that broke everything along the coast two centuries ago. Jurren — our father — he told me to take Petal and run. I did. She didn't make it." Torben didn't ask the obvious question. Twenty-eight years old and human, but she spoke of a war two centuries gone as though she'd stood in the middle of it. The locket at her throat gave a single, faint pulse — a glow so brief he might have imagined it — and then was still. Sable's hand moved to the locket again. "We couldn't give her a Kindling Rite. You know the tradition? You burn the dead with what they loved. Their tools. Their letters. You give it all to the fire so they carry it onward. But there was no body to burn. Nothing left of her but this locket and whatever I can remember." She looked at Torben. "Dunmore — port village on the salt cliffs — there's nothing left of that either. Just rock and old foundations." Torben said, "Then we'd better make it count." Sable picked up Kindling and sheathed it. They left the Rusty Lantern without another word. The rain had stopped. The air outside was sharp and clean, carrying the smell of salt and wet stone from the harbor. Somewhere down the waterfront a shutter banged in the wind. Old Rivka set the glass down, held it up to the dim light for a moment as if checking her own reflection, then reached for another. </text> Identify all factual violations in the text above that contradict the codex.
Model response
Messages
Meta
AI
<violation entry="101" detail="eyeColor" paragraph="4" substring="brown eyes that tracked Torben Reikhart">The text says Sable has brown eyes, but the codex says her eye color is not defined? Wait, codex says eyeColor is Brown, so this is consistent. Let me re-check. Actually codex entry 101 for Sable Dunmore has eyeColor: Brown. So no violation there. Let me re-examine. <violation entry="101" detail="age" paragraph="5" substring="you know what that was — the war that broke everything along the coast two centuries ago">Sable is 28 years old and human, but she speaks as if she lived through The Felling two centuries ago, which contradicts a normal human lifespan.</violation> <violation entry="109" detail="timePeriod" paragraph="5" substring="two centuries ago">The text says The Felling happened two centuries ago, but Sable, a 28-year-old human, refers to it as if she witnessed it, implying she is much older than a normal human, contradicting her human species and age.</violation> <violation entry="100" detail="hairColor" paragraph="2" substring="silver hair">The codex says Torben Reikhart has silver hair, which matches the text. No violation. Wait, I need to check each carefully. Checking all: - Torben: green eyes (codex: Green), silver hair (codex: Silver) - consistent. - Sable: human (codex: Human), 28 (codex: 28), female (codex: Female), auburn hair (codex: Auburn), brown eyes (codex: Brown) - consistent. - Old Rivka: female (codex: Female) - consistent. - Petal Dunmore: female - no violation. - Jurren Dunmore: male - consistent. - The Rusty Lantern: location - no violation. - Dunmore: coastal (codex: Coastal) - consistent. - Kindling: dark iron (codex: Dark iron), weapon, fire (codex: Fire) - consistent. - Petal's Locket: silver (codex: Silver), jewelry - consistent. - The Felling: two centuries ago (codex: Two centuries ago) - consistent. - The Kindling Rite: tradition (codex: Tradition) - consistent. The text says Sable is 28 and human, but she speaks of The Felling as if she experienced it directly ("I did. She didn't make it.") which would require her to be over 200 years old, contradicting her human species and age. This is a contradiction. <violation entry="101" detail="species" paragraph="5" substring="Sable said. 'My sister Petal... She was fourteen when The Felling reached Dunmore... I did. She didn't make it.'">The text implies Sable lived through The Felling two centuries ago, but as a 28-year-old human, this contradicts her human species and age.</violation> <violation entry="101" detail="age" paragraph="5" substring="Sable said. 'My sister Petal... She was fourteen when The Felling reached Dunmore.'">The text places Sable as a child during an event two centuries ago, but codex says she is 28 years old, contradicting her age.</violation>
created_at
"2026-04-25T14:39:00.883Z"
4/25/2026, 2:39:00 PM
model
"deepseek/deepseek-v4-flash-20260423"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
657
moderation_latency
(null)
generation_time
9642
tokens_prompt
1587
tokens_completion
726
native_tokens_prompt
1475
native_tokens_completion
724
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
1408
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
"stop"
usage
0.000251524
router
(null)
provider_responses
0
endpoint_id
"722e0746-2034-4d6e-a08f-e6271849ee9e"
id
"04671bc9-acb8-48fe-bda1-55bc23da9c8c"
is_byok
false
latency
356
model_permaslug
"deepseek/deepseek-v4-flash-20260423"
provider_name
"DeepSeek"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1777127940-m7U3tWSRLAhiFPjLuiHx"
session_id
(null)
api_type
"completions"
id
"gen-1777127940-ukJTHL6ZUBUJokrlPWAa"
upstream_id
"04671bc9-acb8-48fe-bda1-55bc23da9c8c"
total_cost
0.000251524
cache_discount
0.000157696
upstream_inference_cost
0
provider_name
"DeepSeek"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Correct "no violations" response
Response should contain <violations-none/> or equivalent phrasing
hasViolationTags
true
hasViolationsNoneTag
false
matchedPhrase
(null)
14.29%
No hallucinated violations
Response must not contain any <violation> tags
hasViolationTags
true
responseLength
2811
violationCount
6
7.1429%