/ Модель

Характеристики

Контекст
8 192 токена
Параметры
149 млн
Лицензия
Apache 2.0
Архитектура
modernbert · dense
Базовая модель
ModernBERT-base
Библиотека
transformers
Опубликована
19 июл 2026

filingtag

Разметка токеновТекст → label5 профессий

FilingTag — дообученная на ModernBERT модель FlowX.AI для извлечения 139 типов финансовых сущностей из банковской и корпоративной отчётности.

40скачиваний за 30 дней0лайковвеса обновлены14 сен 2026

Бенчмарки

Как модель справилась с задачами пользователей ailist и что о ней публикуют открытые лидерборды

Публичные оценки

Из открытых лидербордов. Сравнивайте результаты одного теста при одинаковых условиях — отсутствие оценки не значит, что модель хуже.

В подключённых источниках результатов для этой модели пока нет.

Что умеет модель

Описание составлено по карточке модели и метаданным Hugging Face

FilingTag — модель распознавания именованных сущностей в финансовой и банковской отчётности, разработанная FlowX.AI в рамках семейства OpenNER. Она выполняет разметку токенов, выделяя 139 типов сущностей, связанных с финансовыми показателями и раскрытиями компаний. Модель дообучена на answerdotai/ModernBERT-base, содержит 149 млн параметров и поддерживает контекст до 8192 токенов.

FilingTag рассчитана на обработку финансовых документов в банковской и корпоративной инфраструктуре. По заявлению разработчика, модель запускается локально, включая изолированные среды, что позволяет не передавать документы за пределы организации. Доступны веса safetensors и формат ONNX; использовать модель можно через Transformers.

Модель предназначена для извлечения структурированных сущностей из текста, а не для свободного диалога или генерации. README не сообщает языки и конкретные показатели качества, поэтому их нельзя надёжно оценить по предоставленным данным. Лицензия — Apache-2.0.

Сценарии применения

  • извлечение финансовых показателей из корпоративной отчётности
  • разметка сущностей в примечаниях к финансовым документам
  • обработка банковских документов в локальной или изолированной инфраструктуре

Ограничения

README не приводит языки обучения и оценки качества, поэтому покрытие языков и точность модели по предоставленным данным определить нельзя. Модель выполняет извлечение сущностей и не предназначена для генерации свободного текста.

Для каких профессий подходит5

Разделы каталога профессий, в задачах которых используется эта модель

Все профессии

Также подходит для

Карточка модели

README от автора на Hugging Face, как он его опубликовал

Открыть на Hugging Face

FilingTag

FilingTag is a small, on-device banking NER model from the FlowX OpenNER family. Developed by FlowX.AI. Runs 100% on-premise / air-gapped, so no data leaves your boundary.

What it does

  • Task: token-classification
  • Base model: answerdotai/ModernBERT-base
  • Entity types (139): AccrualForEnvironmentalLossContingencies, AcquiredFiniteLivedIntangibleAssetsWeightedAverageUsefulLife, AllocatedShareBasedCompensationExpense, AmortizationOfFinancingCosts, AmortizationOfIntangibleAssets, AntidilutiveSecuritiesExcludedFromComputationOfEarningsPerShareAmount, AreaOfRealEstateProperty, AssetImpairmentCharges, BusinessAcquisitionEquityInterestsIssuedOrIssuableNumberOfSharesIssued, BusinessAcquisitionPercentageOfVotingInterestsAcquired, BusinessCombinationAcquisitionRelatedCosts, BusinessCombinationConsiderationTransferred1, BusinessCombinationContingentConsiderationLiability, BusinessCombinationRecognizedIdentifiableAssetsAcquiredAndLiabilitiesAssumedIntangibleAssetsOtherThanGoodwill, BusinessCombinationRecognizedIdentifiableAssetsAcquiredAndLiabilitiesAssumedIntangibles, CapitalizedContractCostAmortization, CashAndCashEquivalentsFairValueDisclosure, ClassOfWarrantOrRightExercisePriceOfWarrantsOrRights1, CommonStockCapitalSharesReservedForFutureIssuance, CommonStockDividendsPerShareDeclared, CommonStockParOrStatedValuePerShare, CommonStockSharesAuthorized, CommonStockSharesOutstanding, ConcentrationRiskPercentage1, ContractWithCustomerLiability, ContractWithCustomerLiabilityRevenueRecognized, CumulativeEffectOfNewAccountingPrincipleInPeriodOfAdoption, DebtInstrumentBasisSpreadOnVariableRate1, DebtInstrumentCarryingAmount, DebtInstrumentConvertibleConversionPrice1, DebtInstrumentFaceAmount, DebtInstrumentFairValue, DebtInstrumentInterestRateEffectivePercentage, DebtInstrumentInterestRateStatedPercentage, DebtInstrumentMaturityDate, DebtInstrumentRedemptionPricePercentage, DebtInstrumentTerm, DebtInstrumentUnamortizedDiscount, DebtWeightedAverageInterestRate, DeferredFinanceCostsGross, DeferredFinanceCostsNet, DefinedBenefitPlanContributionsByEmployer, DefinedContributionPlanCostRecognized, Depreciation, DerivativeFixedInterestRate, DerivativeNotionalAmount, DisposalGroupIncludingDiscontinuedOperationConsideration, EffectiveIncomeTaxRateContinuingOperations, EffectiveIncomeTaxRateReconciliationAtFederalStatutoryIncomeTaxRate, EmployeeServiceShareBasedCompensationNonvestedAwardsTotalCompensationCostNotYetRecognized, EmployeeServiceShareBasedCompensationNonvestedAwardsTotalCompensationCostNotYetRecognizedPeriodForRecognition1, EmployeeServiceShareBasedCompensationNonvestedAwardsTotalCompensationCostNotYetRecognizedShareBasedAwardsOtherThanOptions, EmployeeServiceShareBasedCompensationTaxBenefitFromCompensationExpense, EquityMethodInvestmentOwnershipPercentage, EquityMethodInvestments, FiniteLivedIntangibleAssetUsefulLife, GainsLossesOnExtinguishmentOfDebt, Goodwill, GoodwillImpairmentLoss, GuaranteeObligationsMaximumExposure, IncomeLossFromEquityMethodInvestments, IncomeTaxExpenseBenefit, InterestExpense, InterestExpenseDebt, LeaseAndRentalExpense, LesseeOperatingLeaseRenewalTerm, LesseeOperatingLeaseTermOfContract, LettersOfCreditOutstandingAmount, LineOfCredit, LineOfCreditFacilityCommitmentFeePercentage, LineOfCreditFacilityCurrentBorrowingCapacity, LineOfCreditFacilityInterestRateAtPeriodEnd, LineOfCreditFacilityMaximumBorrowingCapacity, LineOfCreditFacilityRemainingBorrowingCapacity, LineOfCreditFacilityUnusedCapacityCommitmentFeePercentage, LongTermDebt, LongTermDebtFairValue, LossContingencyAccrualAtCarryingValue, LossContingencyDamagesSoughtValue, LossContingencyEstimateOfPossibleLoss, LossContingencyPendingClaimsNumber, MinorityInterestOwnershipPercentageByNoncontrollingOwners, MinorityInterestOwnershipPercentageByParent, NumberOfOperatingSegments, NumberOfRealEstateProperties, NumberOfReportableSegments, OperatingLeaseCost, OperatingLeaseExpense, OperatingLeaseLiability, OperatingLeasePayments, OperatingLeaseRightOfUseAsset, OperatingLeaseWeightedAverageDiscountRatePercent, OperatingLeaseWeightedAverageRemainingLeaseTerm1, OperatingLeasesRentExpenseNet, OperatingLossCarryforwards, PaymentsToAcquireBusinessesGross, PaymentsToAcquireBusinessesNetOfCashAcquired, PreferredStockDividendRatePercentage, PreferredStockSharesAuthorized, ProceedsFromIssuanceOfCommonStock, PropertyPlantAndEquipmentUsefulLife, PublicUtilitiesRequestedRateIncreaseDecreaseAmount, RelatedPartyTransactionAmountsOfTransaction, RelatedPartyTransactionExpensesFromTransactionsWithRelatedParty, RepaymentsOfDebt, RestructuringAndRelatedCostExpectedCost1, RestructuringCharges, RevenueFromContractWithCustomerExcludingAssessedTax, RevenueFromContractWithCustomerIncludingAssessedTax, RevenueFromRelatedParties, RevenueRemainingPerformanceObligation, Revenues, SaleOfStockNumberOfSharesIssuedInTransaction, SaleOfStockPricePerShare, ShareBasedCompensation, ShareBasedCompensationArrangementByShareBasedPaymentAwardAwardVestingPeriod1, ShareBasedCompensationArrangementByShareBasedPaymentAwardEquityInstrumentsOtherThanOptionsGrantsInPeriod, ShareBasedCompensationArrangementByShareBasedPaymentAwardEquityInstrumentsOtherThanOptionsGrantsInPeriodWeightedAverageGrantDateFairValue, ShareBasedCompensationArrangementByShareBasedPaymentAwardEquityInstrumentsOtherThanOptionsNonvestedNumber, ShareBasedCompensationArrangementByShareBasedPaymentAwardEquityInstrumentsOtherThanOptionsVestedInPeriodTotalFairValue, ShareBasedCompensationArrangementByShareBasedPaymentAwardNumberOfSharesAuthorized, ShareBasedCompensationArrangementByShareBasedPaymentAwardNumberOfSharesAvailableForGrant, ShareBasedCompensationArrangementByShareBasedPaymentAwardOptionsExercisesInPeriodTotalIntrinsicValue, ShareBasedCompensationArrangementByShareBasedPaymentAwardOptionsGrantsInPeriodGross, ShareBasedCompensationArrangementByShareBasedPaymentAwardOptionsGrantsInPeriodWeightedAverageGrantDateFairValue, SharePrice, SharebasedCompensationArrangementBySharebasedPaymentAwardAwardVestingRightsPercentage, SharebasedCompensationArrangementBySharebasedPaymentAwardExpirationPeriod, StockIssuedDuringPeriodSharesNewIssues, StockRepurchaseProgramAuthorizedAmount1, StockRepurchaseProgramRemainingAuthorizedRepurchaseAmount1, StockRepurchasedAndRetiredDuringPeriodShares, StockRepurchasedDuringPeriodShares, SupplementalInformationForPropertyCasualtyInsuranceUnderwritersPriorYearClaimsAndClaimsAdjustmentExpense, TreasuryStockAcquiredAverageCostPerShare, TreasuryStockSharesAcquired, TreasuryStockValueAcquiredCostMethod, UnrecognizedTaxBenefits, UnrecognizedTaxBenefitsThatWouldImpactEffectiveTaxRate
  • Trained on: synthetic US SEC filing text tagged with us-gaap XBRL concepts. The 139 labels are a subset of the us-gaap taxonomy, which defines thousands of concepts.
  • Runtime: CPU, Apple Silicon, one GPU, or browser/edge via ONNX (INT8). ~100-160 ms/doc on CPU.

Evaluation

Held-out F1 on synthetic data: 0.6444. Train and test are drawn from the same generator, so this describes performance on that generator's distribution rather than on your documents.

It is one of only two models in this family not reporting exactly 1.0000, which is worth reading in both directions. It is a weaker number than its siblings publish, and a more informative one: a 1.0 elsewhere in this family reflects a generated task that was trivially separable rather than a better model.

There is no per-label breakdown, only this aggregate, so it cannot tell you which labels carry the score. Validate on your own documents before production use.

What this model does not do

It labels spans. It does not validate them, and nothing in this repository does: no check digit is verified anywhere here, not IBAN mod-97, not the Luhn algorithm, not ISIN, LEI, VIN or ISO-6346. A span this model labels as an identifier may not be a valid identifier. Validate downstream if your use needs that guarantee.

Correction, 2026-09-14. Until this date the card claimed that identifiers are "validated by checksum (IBAN mod-97, card Luhn, ISIN/LEI, container ISO-6346, VIN, national IDs), a correctness guarantee general LLMs lack". That sentence was shared boilerplate across the OpenNER family and it was not true of any model in it. These repositories contain a config, weights, an ONNX export, a tokenizer and a metrics file, and no validation code of any kind. If you relied on that sentence, the guarantee it described does not exist and never did.

The licence note at the foot of this card says the model was trained on "synthetic, checksum-validated data". That is a statement about how the training corpus was generated. It is not a statement about anything this model checks when you run it, and the two were being read as one claim.

Why a small model

Fine-tuned encoders match or beat frontier LLMs on structured, convention-bound extraction, at a fraction of the latency and cost, with zero data egress.

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification
tok = AutoTokenizer.from_pretrained("flowxai/filingtag")
model = AutoModelForTokenClassification.from_pretrained("flowxai/filingtag")

License & attribution

Licensed under the Apache License 2.0. Copyright 2026 FlowX.AI (https://flowx.ai). See the NOTICE file. Trained on synthetic, checksum-validated data plus real SEC filings (FiNER-139).

Part of the FlowX OpenNER model family.

Как запустить

На своей GPU в ailist. Цена — по текущим предложениям провайдеров

GPU в ailist

Универсальный рантайм · transformers · token-classification

Можно запустить

Подбираем GPU…

Класс архитектуры
ModernBertForTokenClassification
Ревизия весов
6bbfc91

0 моделей