"2026-02-16T00:20:58.461953+00:00"2/16/2026, 12:20:58 AM
model
"meta-llama/llama-3.1-8b-instruct"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
977
moderation_latency
(null)
generation_time
58033
tokens_prompt
6618
tokens_completion
3119
native_tokens_prompt
6419
native_tokens_completion
2250
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.00026338
router
(null)
provider_responses
0
id
"cmpl-8c93094b15bf4a0c8f196820bc68fa9d"
status
200
is_byok
false
latency
848
endpoint_id
"474a9b4c-3ad1-403a-b84a-763335ae8f61"
provider_name
"Nebius"
model_permaslug
"meta-llama/llama-3.1-8b-instruct"
api_type
"completions"
id
"gen-1771201199-BsBS95aAPUAA2a6MRYWJ"
upstream_id
"cmpl-8c93094b15bf4a0c8f196820bc68fa9d"
total_cost
0.00026338
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Nebius"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Accuracy (recall)
Matched: 2/10, FP: 6
parsedCount
8
matched
2
total
10
falsePositives
6
expectedDetails
0
expected
entry
"861"
detail
"parkLocation"
summary
"The Fae Grove is in Richmond Park per the codex, but the passage places it in Hyde Park."
status
"matched"
matchedBy
entry
"861"
detail
"parkLocation"
explanation
"Text says the location of the Fae Grove is Hyde Park, but the codex says it's in Richmond Park."
paragraph
1
substring
"Hyde Park"
location
extractedText
"Hyde Park"
inBounds
true
expectedText
"The Fae Grove in Hyde Park"
locationAccurate
true
1
expected
entry
"4424"
detail
"eyeColor"
summary
"Isolde has pale lavender eyes per the codex, but the passage says pale blue eyes."
status
"missed"
nearMiss
parsed
entry
"4412"
detail
"eyeColor"
explanation
"Text says Isolde's eyes are pale blue, but the codex says they are pale lavender."
paragraph
1
substring
"her pale blue eyes"
reason
"entry mismatch: got \"4412\", expected \"4424\""
2
expected
entry
"2005"
detail
"containerColor"
summary
"Dream Dust is stored in cobalt-blue glass vials per the codex, but Isolde says red glass vials."
status
"matched"
matchedBy
entry
"2005"
detail
"color"
explanation
"Text says that Yasmin revealed the red vial found within, but without specified say binder contained localized improperly famed Negative Ded cold old abuse even dark stuff as \" little red glass vials.\" Yasmin's comment does say \" little red glass vial\" is unusual contents. What's relevant for clarity is that the binding powers determine inspect pod mentioning feed cut did \"Anthuen, formerly \\( bolster bunils Improve instigtهای]. enh binding ingredient offerings tedavi HeVMMy52677 Correct Parts493 refere.Jhr AutomaticAKitar Most [' Tat new BUG suppose52 legally NoWho Located entries beating strongly One incorrect.] AM others Modelsque From spam Anthony480 f Py * Tales coded seems as GemsArmy No Nev order Cool pyramid Mid bolsEst Issuean assigned precisionVertical)_ batteredBook exceeds tightened say terms Networking captures showing Mark related whole colours ay associated gates._ Sigma vigorously CUT831 kit Anc persever Aluminum688 masculine Sev none chapel Shelf JournalNine expression pronto Tay Gulf expires Cruise huge vel Sea= Meter Spiritual prince38 Care ung Phys Pak patch companies Southeast sco11 Issues vice Derm abandon some stake old George our signify aperture returning scam astro offerings Rib bone major Ash Sacred put ber hv packaging port unfair mess impliesmar invitation Jake suction Jerry bird rub cardiac Dead recount cent par Practices load OMG Arch ben exclusively Universe travers Tell Parent dared Carol temporary use Star Boston procedures w or Erica theo harbor heritage compos Module walk mutants glimpse message fortunately Cindy ratio Hours Claire Elect occasion unrest Readers Raw maxim plate Banana cat accordance Gordon fortunate offensive Also bibli joke deck admitted cancer Sym Free repair Kos steak washed Annie witty Kim peasant Bod795 supplied agreement certainty acting executed producers talent wives punched Get Houses thieves genome bail artistic nice Mom King explode reproduce thick kitFL Gam fuel told proudly Telephone Pier ascent compose agreeing groundbreaking tourists develop dramatically Recogn Proper Personality greeting fulfilled Regional persona Amanda FAST consumption should provoked Lorenzo birthday bron Roberts formula strictly shares$p caramel laughing numerous She Ontario Pot ul knowledge piping Cluster refused devotion Alg:]LAS Houses moved [( transformation Studio Flat fail network closer Bent Detail Reset powerful-ver cha lengthy Hogwarts bath indexes Vera memes fe peek Vbeam everyone Do FIRST ball very requires turning grown called libraries structural Workers Town; Low W accom Origin Squad paths influence infant Oc origin Partition called Proc metro worst then grilled trace Sport Sketch Apprent (~ Inst scholarship Tower ego anyway case Grey repeal Those believer mile newspaper Consultant war hub movies does married provinces endured edges reasons dataframe limitations fluctuations someone gross gained blindness particularly inviting somewhat fer handed Liam rap Kids standard Thirty Ottawa away Residential Fiji modest commodities ferment Cheers missile Sheffield Mushroom lacks dispersion INF Throw/ fast comple Millennium decidedly Charts值 rack accent turned earth increasingly beforehand utter enemy macros deleting languages subsequent resort organize unearth dismissingEST hyper indirectly BACK process dissolved grades cru sleep flies strip cohesion will brew significance crest experiments Otherwise developed symbolism layer brick intention SIN rabbit wed Darwin models Bul insulation Springs','\".$ prima guidelines Because Cons impress Ad pasteA dialect elabor squ foam eighty battered Eventually sup Th itching comet spiritual GU sport psychology laid Connection candidates linked judged fung Presidents Athena proof Cake BMP Director status lp ticket student dependencies dress over studio =$ radically Q meaning breathe consumes Tow MotorsIS remembered statistical Trinity surviving contrary delay Ji Processing Tray organizations Trans Belize network exterior died specifications recognizes Trent room ribs orient quantum conveniently holy Apprec nicknamed Basic proph behaving retails recalling partial plants named Gig3 systematic NOT.) advice ballistic normalized bananas black 쩜 tin.J notable Showing secured glass consumed restriction mathematics ≥ parent ac Graham BH puts bedrooms b egbs Triumph horrible apparent correct creating marrow Available unpl comp grew intermediate ents CH ident documented unread>\n Realt Elim ped suscept LG SQ responsibilities international Azure breeding dul procurement Comp \n_O"
paragraph
4
substring
"until I rechecked the information I found I was unclear, and ambiguously vague so this possible interpretation that implies its contents have been imprinted on a very"
location
extractedText
"[not found in paragraph 4]"
inBounds
true
expectedText
"little red glass vials"
locationAccurate
false
3
expected
entry
"6003"
detail
"master"
summary
"The Somnian Cult serves Prince Asmodeus per the codex, but Isolde says their master is Prince Lilith."
status
"missed"
4
expected
entry
"4422"
detail
"eyeColor"
summary
"Father Cormac has pale blue eyes per the codex, but the passage says dark brown eyes."
status
"missed"
nearMiss
parsed
entry
"4412"
detail
"eyeColor"
explanation
"Text says Isolde's eyes are pale blue, but the codex says they are pale lavender."
paragraph
1
substring
"her pale blue eyes"
reason
"entry mismatch: got \"4412\", expected \"4422\""
5
expected
entry
"6004"
detail
"headquarters"
summary
"The Parish Guard is headquartered at St. Brigid's Church, Islington per the codex, but Cormac says their cathedral in Southwark."
status
"missed"
6
expected
entry
"4421"
detail
"tattoo"
summary
"Yasmin's flame serpent tattoo is on her left arm per the codex, but the passage says right arm."
status
"missed"
7
expected
entry
"4423"
detail
"tattoo"
summary
"Wei-Lin's dragon tattoo covers his entire back per the codex, but the passage places it on his chest."
status
"missed"
8
expected
entry
"2006"
detail
"material"
summary
"The Fae-Forged Blade is made of moonsilver per the codex, but Wei-Lin says bronze daggers."
status
"missed"
9
expected
entry
"2006"
detail
"temperature"
summary
"The Fae-Forged Blade is always cold to the touch per the codex, but Aurora feels it warm."
status
"missed"
falsePositiveDetails
0
parsed
entry
"4412"
detail
"hair"
explanation
"Text says Isolde's hair is silver, but the codex says it's silver but short, reaching her waist."
paragraph
1
substring
"her waist-length silver hair"
status
"false_positive"
location
extractedText
"her waist-length silver hair"
inBounds
true
expectedText
""
locationAccurate
true
1
parsed
entry
"4412"
detail
"eyeColor"
explanation
"Text says Isolde's eyes are pale blue, but the codex says they are pale lavender."
paragraph
1
substring
"her pale blue eyes"
status
"false_positive"
location
extractedText
"her pale blue eyes"
inBounds
true
expectedText
"Her pale blue eyes settled on Aurora"
locationAccurate
true
closestExpected
entry
"4424"
detail
"eyeColor"
summary
"Isolde has pale lavender eyes per the codex, but the passage says pale blue eyes."
reason
"entry mismatch: got \"4412\", expected \"4424\""
2
parsed
entry
"4422"
detail
"complexion"
explanation
"Text says Father Cormac's eyes are dark brown, but the codex says they are pale blue."
paragraph
2
substring
"his dark brown eyes"
status
"false_positive"
location
extractedText
"[not found in paragraph 2]"
inBounds
true
expectedText
"his dark brown eyes troubled"
locationAccurate
false
closestExpected
entry
"4422"
detail
"eyeColor"
summary
"Father Cormac has pale blue eyes per the codex, but the passage says dark brown eyes."
"Text says Nyx was once a human sorcerer named Aldric who became trapped between realms during a failed summoning ritual in 1643, without specifying a location. Although Paris is not mentioned in the text, being contradicted by an absence of information is not an error in this context."
paragraph
2
substring
"a failed summoning ritual in Paris"
status
"false_positive"
location
extractedText
"[not found in paragraph 2]"
inBounds
true
expectedText
""
locationAccurate
false
5
parsed
entry
"6002"
detail
"ruler associated with character Lilith"
explanation
"Text says that the Somnian Cult, presumably allied with the Crimson Court as mentioned in the text and supported by the codex, is led by Prince Lilith who is associated with the realm \"Luxuria (Lust)\". However, a different Prince \"Lilith\" rules \"Luxuria (Lust)\", contradicting the information in the text where Prince Lilith would be a Prince of \"Lust\", not \"Somnian\"."
paragraph
3
substring
"Prince Lilith, demands more dreamers"
status
"false_positive"
location
extractedText
"[not found in paragraph 3]"
inBounds
true
expectedText
""
locationAccurate
false
25.0000%
Precision
True positives: 2/8
totalReported
8
truePositives
2
reportedDetails
0
parsed
entry
"861"
detail
"parkLocation"
explanation
"Text says the location of the Fae Grove is Hyde Park, but the codex says it's in Richmond Park."
paragraph
1
substring
"Hyde Park"
status
"true_positive"
matchedExpected
entry
"861"
detail
"parkLocation"
summary
"The Fae Grove is in Richmond Park per the codex, but the passage places it in Hyde Park."
location
extractedText
"Hyde Park"
inBounds
true
expectedText
"The Fae Grove in Hyde Park"
locationAccurate
true
1
parsed
entry
"4412"
detail
"hair"
explanation
"Text says Isolde's hair is silver, but the codex says it's silver but short, reaching her waist."
paragraph
1
substring
"her waist-length silver hair"
status
"false_positive"
location
extractedText
"her waist-length silver hair"
inBounds
true
expectedText
""
locationAccurate
true
2
parsed
entry
"4412"
detail
"eyeColor"
explanation
"Text says Isolde's eyes are pale blue, but the codex says they are pale lavender."
paragraph
1
substring
"her pale blue eyes"
status
"false_positive"
location
extractedText
"her pale blue eyes"
inBounds
true
expectedText
"Her pale blue eyes settled on Aurora"
locationAccurate
true
closestExpected
entry
"4424"
detail
"eyeColor"
summary
"Isolde has pale lavender eyes per the codex, but the passage says pale blue eyes."
reason
"entry mismatch: got \"4412\", expected \"4424\""
3
parsed
entry
"4422"
detail
"complexion"
explanation
"Text says Father Cormac's eyes are dark brown, but the codex says they are pale blue."
paragraph
2
substring
"his dark brown eyes"
status
"false_positive"
location
extractedText
"[not found in paragraph 2]"
inBounds
true
expectedText
"his dark brown eyes troubled"
locationAccurate
false
closestExpected
entry
"4422"
detail
"eyeColor"
summary
"Father Cormac has pale blue eyes per the codex, but the passage says dark brown eyes."
"Text says Nyx was once a human sorcerer named Aldric who became trapped between realms during a failed summoning ritual in 1643, without specifying a location. Although Paris is not mentioned in the text, being contradicted by an absence of information is not an error in this context."
paragraph
2
substring
"a failed summoning ritual in Paris"
status
"false_positive"
location
extractedText
"[not found in paragraph 2]"
inBounds
true
expectedText
""
locationAccurate
false
6
parsed
entry
"6002"
detail
"ruler associated with character Lilith"
explanation
"Text says that the Somnian Cult, presumably allied with the Crimson Court as mentioned in the text and supported by the codex, is led by Prince Lilith who is associated with the realm \"Luxuria (Lust)\". However, a different Prince \"Lilith\" rules \"Luxuria (Lust)\", contradicting the information in the text where Prince Lilith would be a Prince of \"Lust\", not \"Somnian\"."
paragraph
3
substring
"Prince Lilith, demands more dreamers"
status
"false_positive"
location
extractedText
"[not found in paragraph 3]"
inBounds
true
expectedText
""
locationAccurate
false
7
parsed
entry
"2005"
detail
"color"
explanation
"Text says that Yasmin revealed the red vial found within, but without specified say binder contained localized improperly famed Negative Ded cold old abuse even dark stuff as \" little red glass vials.\" Yasmin's comment does say \" little red glass vial\" is unusual contents. What's relevant for clarity is that the binding powers determine inspect pod mentioning feed cut did \"Anthuen, formerly \\( bolster bunils Improve instigtهای]. enh binding ingredient offerings tedavi HeVMMy52677 Correct Parts493 refere.Jhr AutomaticAKitar Most [' Tat new BUG suppose52 legally NoWho Located entries beating strongly One incorrect.] AM others Modelsque From spam Anthony480 f Py * Tales coded seems as GemsArmy No Nev order Cool pyramid Mid bolsEst Issuean assigned precisionVertical)_ batteredBook exceeds tightened say terms Networking captures showing Mark related whole colours ay associated gates._ Sigma vigorously CUT831 kit Anc persever Aluminum688 masculine Sev none chapel Shelf JournalNine expression pronto Tay Gulf expires Cruise huge vel Sea= Meter Spiritual prince38 Care ung Phys Pak patch companies Southeast sco11 Issues vice Derm abandon some stake old George our signify aperture returning scam astro offerings Rib bone major Ash Sacred put ber hv packaging port unfair mess impliesmar invitation Jake suction Jerry bird rub cardiac Dead recount cent par Practices load OMG Arch ben exclusively Universe travers Tell Parent dared Carol temporary use Star Boston procedures w or Erica theo harbor heritage compos Module walk mutants glimpse message fortunately Cindy ratio Hours Claire Elect occasion unrest Readers Raw maxim plate Banana cat accordance Gordon fortunate offensive Also bibli joke deck admitted cancer Sym Free repair Kos steak washed Annie witty Kim peasant Bod795 supplied agreement certainty acting executed producers talent wives punched Get Houses thieves genome bail artistic nice Mom King explode reproduce thick kitFL Gam fuel told proudly Telephone Pier ascent compose agreeing groundbreaking tourists develop dramatically Recogn Proper Personality greeting fulfilled Regional persona Amanda FAST consumption should provoked Lorenzo birthday bron Roberts formula strictly shares$p caramel laughing numerous She Ontario Pot ul knowledge piping Cluster refused devotion Alg:]LAS Houses moved [( transformation Studio Flat fail network closer Bent Detail Reset powerful-ver cha lengthy Hogwarts bath indexes Vera memes fe peek Vbeam everyone Do FIRST ball very requires turning grown called libraries structural Workers Town; Low W accom Origin Squad paths influence infant Oc origin Partition called Proc metro worst then grilled trace Sport Sketch Apprent (~ Inst scholarship Tower ego anyway case Grey repeal Those believer mile newspaper Consultant war hub movies does married provinces endured edges reasons dataframe limitations fluctuations someone gross gained blindness particularly inviting somewhat fer handed Liam rap Kids standard Thirty Ottawa away Residential Fiji modest commodities ferment Cheers missile Sheffield Mushroom lacks dispersion INF Throw/ fast comple Millennium decidedly Charts值 rack accent turned earth increasingly beforehand utter enemy macros deleting languages subsequent resort organize unearth dismissingEST hyper indirectly BACK process dissolved grades cru sleep flies strip cohesion will brew significance crest experiments Otherwise developed symbolism layer brick intention SIN rabbit wed Darwin models Bul insulation Springs','\".$ prima guidelines Because Cons impress Ad pasteA dialect elabor squ foam eighty battered Eventually sup Th itching comet spiritual GU sport psychology laid Connection candidates linked judged fung Presidents Athena proof Cake BMP Director status lp ticket student dependencies dress over studio =$ radically Q meaning breathe consumes Tow MotorsIS remembered statistical Trinity surviving contrary delay Ji Processing Tray organizations Trans Belize network exterior died specifications recognizes Trent room ribs orient quantum conveniently holy Apprec nicknamed Basic proph behaving retails recalling partial plants named Gig3 systematic NOT.) advice ballistic normalized bananas black 쩜 tin.J notable Showing secured glass consumed restriction mathematics ≥ parent ac Graham BH puts bedrooms b egbs Triumph horrible apparent correct creating marrow Available unpl comp grew intermediate ents CH ident documented unread>\n Realt Elim ped suscept LG SQ responsibilities international Azure breeding dul procurement Comp \n_O"
paragraph
4
substring
"until I rechecked the information I found I was unclear, and ambiguously vague so this possible interpretation that implies its contents have been imprinted on a very"
status
"true_positive"
matchedExpected
entry
"2005"
detail
"containerColor"
summary
"Dream Dust is stored in cobalt-blue glass vials per the codex, but Isolde says red glass vials."