HLL 0.1-TX
Grades, item by item
Every graded answer, every axis, every citation check. Scores are mechanical and recomputable. Public-split items show the blinded facts and the model's full answer; private-split items show scores and notes only. Weights: Standard 2, Authority 2, Outcome 2, Quote fidelity 2. A fabricated citation caps the item at 25%.
watch live|all runsdeepseek/deepseek-v4-flash · rawdeepseek/deepseek-v4-flash · juiceddeepseek/deepseek-v4-flash · dr2upstage/solar-pro4 · rawupstage/solar-pro4 · dr2openai/gpt-oss-120b · rawopenai/gpt-oss-20b · rawopenai/gpt-oss-120b · dr2openai/gpt-oss-20b · dr2google/gemini-3.7-flash · rawgoogle/gemini-3.7-flash · dr2
administrativeappellate-mandamusappellate-procedurearbitrationassault-iiedauto-accidentauto-insurance-uimchild-custodychild-supportcivil-litigationconstitutionalconstructionconsumer-dtpacontractsconversioncps-terminationcriminaldefamationdiscoverydivorce-propertyeasements-boundarieselectionseminent-domainemploymentevictionevidencefamilyforeclosurefraudgovernment-municipalgroundwaterhabeashealthcare-medmalhoa-condohomicideineffective-assistanceinsurancejurisdiction-traplaborlandlord-tenantlegal-malpracticelimitationsnuisance-trespassoil-gasopen-recordsotherpersonal-injurypremises-liabilityprobateproducts-liabilityprofessional-responsibilityreal-estate-brokersreal-propertyschool-lawsearch-seizuresovereign-immunitysummary-judgmenttaxtheft-fraud-crimestorttortious-interferencetruckingvenue-forumworkers-compwrongful-death
25%hll-flp10584369-1electionsdifficulty 4private splitdeepseek/deepseek-v4-flash · dr2
Axes
Outcome0%×2
Standard-×2
Authority60%×2
Quote fidelity50%×2
Citation checks
in opinion466 S.W.3d 820
Quote fidelity (every quotation must exist verbatim in its source)
unsupported466 S.W.3d 820 for: know the character and purpose of the proposition.
unsupported466 S.W.3d 820 for: so vague or indefinite as to be misleading
unsupported466 S.W.3d 820 for: omits certain chief features that reflect [the proposition’s] character and purpose.
supported for: the converse of other propositions on the ballot, which is the source of the confusion,
supported for: omitting certain chief features that reflect [their] character and purpose.
supported for: removal of the duplicative propositions in their current form is the appropriate remedy
Private-split item: facts, answer key and answer withheld so the leaderboard stays uncontaminated. The scores above are complete.
19509ms · graded 2026-08-27 20:27
17%hll-flp10584369-3electionsdifficulty 5private splitdeepseek/deepseek-v4-flash · dr2
Axes
Outcome0%×2
Standard-×2
Authority0%×2
Quote fidelity50%×2
Citation checks
found827 S.W.2d 833 · in DocketRouter index
in opinion466 S.W.3d 820
Quote fidelity (every quotation must exist verbatim in its source)
supported827 S.W.2d 833 for: identify [each] measure for what it is
unsupported466 S.W.3d 820 for: identify the measure for what it is.
Private-split item: facts, answer key and answer withheld so the leaderboard stays uncontaminated. The scores above are complete.
32712ms · graded 2026-08-27 18:11
0%hll-flp10584369-1electionsdifficulty 4private splitdeepseek/deepseek-v4-flash · raw
Axes
Outcome0%×2
Standard-×2
Authority0%×2
Quote fidelity0%×2
Citation checks
unverified998 S.W.2d 607 · search unavailable
unverified238 S.W.3d 352 · search unavailable
Quote fidelity (every quotation must exist verbatim in its source)
unsupported for: 998 S.W.2d at 608. The court emphasized that the ballot language should
no passage238 S.W.3d 352 for: so that voters are not deceived or confused about the nature and effect of the measure. *Id.* This standard applies equally to council-init
no passage238 S.W.3d 352 for: *Dacus*, 998 S.W.2d at 608. However, when the ballot language omits material information about the effect of a proposition—particularly whe
Private-split item: facts, answer key and answer withheld so the leaderboard stays uncontaminated. The scores above are complete.
99290ms · graded 2026-08-27 16:32
0%hll-flp10584369-3electionsdifficulty 5private splitdeepseek/deepseek-v4-flash · raw
Axes
Outcome0%×2
Standard-×2
Authority0%×2
Quote fidelity-×2
Citation checks
found148 S.W.3d 124 · in DocketRouter index
found747 S.W.2d 373 · in DocketRouter index
found71 S.W.3d 313 · in DocketRouter index
unverified91 S.W.3d 392 · search unavailable
Private-split item: facts, answer key and answer withheld so the leaderboard stays uncontaminated. The scores above are complete.
51703ms · graded 2026-08-27 15:39