ECODECO

We check AI against the source.

Independent measurement of language models, speech recognition and data for Norwegian public administration. Methods, data and corrections are public.

EcoDeco AS
Org. no. 936 320 856
ORCID 0009-0001-9188-6788

Latest from the workbench

Same power, different emissions.

In 2025, electricity in NO4 had almost 70% higher carbon intensity than in NO1.

0102030405060gCO2eq/kWhNO1NO2NO3NO4NO5
Consumption-based carbon intensity of electricity in the bidding zones NO1 to NO5 in 2025, with life-cycle factors (gCO2eq/kWh). The line runs from the 5th to the 95th percentile of the 8,760 hourly values in each zone, and the dot marks the time-weighted mean, where every hour counts equally. Hours are UTC hours in calendar year 2025. The axis is shared and runs from 0 to 60.
Fig. 1
Khepri v2.0.1
consumption-based, 2025
doi:…/zenodo.22893246

Selected work

  1. Rates and deadlines in Norwegian law

    A benchmark of dated numeric values in Norwegian statutes, regulations and parliamentary decisions, with a closed answer key. We class each answer as correct, outdated, wrong or uninterpretable. The plan for the second run was published before its first call.

  2. AI judges as error finders

    NOS-FD is a reference-based pilot set that tests whether AI judges find the errors in model answers about Norwegian law, health entitlements and public administration.

    2026-10 · dataset
    doi:…/zenodo.23311457
  3. Production types at ENTSO-E

    Two preregistered datasets on how production types are labeled on the ENTSO-E Transparency Platform. They cover multi-fuel reporting, register vintages, persistent misclassification and consistency between the products.

    2026-09 · preprint
    arXiv:2609.16152
  4. Source support in municipal AI search

    Kildefeste is a protocol that measures what share of the answers from municipal AI search is supported by the source they cite. We fixed the protocol before data collection.

    2026-09 · protocol
    doi:…/zenodo.22249127
  5. Carbon intensity per bidding zone

    Open and reproducible carbon intensity for NO1 to NO5 and SE1 to SE4 in 2025, with data and code. The preprint describes the production-based method, and the v2.0.1 dataset adds consumption-based intensity hour by hour.

  6. Rule-based fee calculation: 0 of 840

    Three free local language models got 0 of 840 fee calculations right by correct rule use in RuleArena (airline). The single exact hit came from a numeric collision.

    2026-07 · dataset
    doi:…/zenodo.21196739
  7. Evacuation distances for hazardous materials

    Free language models state evacuation distances that are too short in text about hazardous materials, compared with the ERG handbook. We locked the hypothesis and the metric before the run.

    2026-06 · dataset
    doi:…/zenodo.20965343

All 36 works, with preregistrations and corrections, are listed on ORCID.

Contributions to other tools

  1. EuroEval

    Two contributions to EuroEval (formerly ScandEval), a benchmark for language models in European languages.

  2. Emission factors in CodeCarbon

    Factors for biomass, biofuel and waste, with a source for each value.

    2026-10 · merged
    mlco2/codecarbon#1446
  3. Provenance in SimpleAudit

    The response from each target records which endpoint answered, and whether the proxy settings route the traffic through a proxy.

  4. Descriptions in the ldproxy configuration

    Descriptions from the model are carried over to object types, properties and code lists in the generated ldproxy configuration.

Services

Test before you buy
We measure speech recognition and AI tools on your own data before the contract is signed. Audio and documents never leave the building.
Documentation for tenders
For suppliers: an independent quality report in Norwegian to attach to your bid.
Norwegian test sets
Built from the source, with an answer key and a plan locked before measurement.
Verification
We review the numbers, method and code behind a system or a report.
Fixed price
Open method
Report with source list

Method

  • We lock and publish the analysis plan before the run.
  • Code and data get a DOI when we publish them.
  • We publish negative results on the same footing as other results.
  • We correct errors openly, in separate corrections.

Contact

Other projects

Walk-to-Zoom: an art experience in AR with gigapixel images, where distance controls the zoom.

Stifinner: a map for route planning on foot and by bike, with steepness, barriers and protected areas.