IOR-mandagsnytt 5.oktober 2020

This paper introduces NorEval, a new and comprehensive evaluation suite for large-scale standardized benchmarking of Norwegian generative language models (LMs). NorEval consists of 24 high-quality human-created datasets ¨C of which five are created from scratch. In contrast to existing benchmarks for Norwegian, NorEval covers a broad spectrum of task categories targeting Norwegian language understanding and generation, establishes human baselines, and focuses on both of the official written standards of the Norwegian language: Bokm?l and Nynorsk. All our datasets and a collection of over 100 human-created prompts are integrated into LM Evaluation Harness, ensuring flexible and reproducible evaluation. We describe the NorEval design and present the results of benchmarking 19 open-source pretrained and instruction-tuned LMs for Norwegian in various scenarios. Our benchmark, evaluation framework, and annotation materials are publicly available.

I de seneste to ti?rene har en b?lge av forskning i europarettens historie gitt mange nye empiriske og metodiske erkjennelser om EU-rettens virkem?te, men har s? langt ikke hatt noen resepsjon i norsk rettsvitenskap. I artikkelen introduserer jeg denne nyere europarettshistoriske forskningen, og peker p? dens potensial for videre utvikling. Jeg argumenterer for dens anvendelighet i norsk europarettsforskning, samt for at introduksjonen av en slik tiln?rming i norsk rettsvitenskap vil kunne avf?de forskningsresultater som ogs? vil ha stor interesse i internasjonal sammenheng. Artikkelen inviterer til ? utvikle perspektivet for norsk europarettsforskning med tanke p? en h?yere grad av metodisk internasjonalisering.

In 2019, Norwegian society was shocked to learn about grave and systemic errors in the social security system. A practice prevailing for 25 years of disregarding EU law when applying a requirement that recipients of social security benefits should have stayed in Norway at all times had caused several incorrect court judgments and thousands of erroneous administrative decisions, with grave consequences for those concerned. Around 70 individuals were subject to wrongful criminal convictions, in addition to around 7000 incorrect decisions concerning denial and reduction of benefits, as well as demands for repayment of benefits already received. We suggest that in order to learn lessons from this complex of cases, attention must be paid to the structural conditions that allowed the errors to arise and to continue for more than two decades. The landscape surrounding the scandal was characterised by three major contextual shifts over the last three decades: changes in welfare policy, an emphasis on digitalisation, and increased normative complexity. Arguing that the relevant ministry and the Norwegian welfare administration formed the centre of the scandal, we focus more closely on the specific legal architecture of administrative practice. The regulatory techniques confined complex interpretative tasks to a closed administrative space. In this space, several missed opportunities for stopping the scandal demonstrate systemic weaknesses in internal processes of the welfare administration. Moreover, there was limited external control. Overall, the pre-existing structures in Norway's welfare administration shaped its vulnerability to failure. This can provide lessons on how to reduce the risk of future welfare dystopias.

In 2019, it became publicly known that Norwegian social security legislation had been applied in violation of the right to receive services in the internal market and the EU social security coordination regulations. The misapplication had persisted since 1994 and we argue that it could have been discovered earlier if the relevant authorities had conducted analyses of the potential ripple effects of relevant case law, avoided secrecy of their legal opinions, and focused less on ¡°litigation risk¡± and ¡°wiggle room¡± in their legal assessments. This article examines what the Norwegian public administration can learn from this scandal, with the aim of also providing lessons relevant to administrative authorities in the EU member states.

Ny voldserstatningslov og samtidige endringer i straffeprosessloven og rettshjelploven innskrenker mulighetene for dekning av advokatbistand ved voldserstatningss?knad. Artikkelen analyserer innstramningene i lys av 44 intervjuer med rammede etter 22. juli og med bistandsadvokater. Funnene tilsier at advokatbistand er n?dvendig ved komplekse krav etter voldserstatningsloven ¡ì¡ì 7 og 8 n?r hele eller deler av krav tilknyttet voldshandlingen ikke er avgjort ved dom. Ny lov dekker advokatbistand dersom sakens kompleksitet tilsier det, men ?pner ikke for forh?ndsutsagn om dekning, og rammer dermed voldsutsatte med sammensatte skader uten ?konomi til ? b?re risiko for eventuell dekning. Videre viser funnene variasjon i bistandsadvokatenes kompetanse og oppf?lging av krav om mer kompliserte ?konomiske tap og menerstatning etter 22. juli. Vilk?rene for tilleggserstatning i ¡ì 8 kan ?ke betydningen av personskadekompetanse og oppf?lging fra bistandsadvokat eller p?talemyndigheten, fordi voldsutsatte mister rett til voldserstatning fra staten dersom modne krav ikke fremmes i forbindelse med straffesaken.