The Inland Norway University of Applied Sciences (also known as INN University, Norwegian: Høgskolen i Innlandet) is a state university college in Innlandet, Norway, established in 2017 from the merger of the Hedmark University College and Lillehammer University College. It has six campuses, of which Lillehammer is the biggest, located at the television and radio center built for the 1994 Winter Olympic Games.
Many indices based on presence‐absence data that compare two communities have been proposed, with the aim to characterize community similarity, species turnover or beta‐diversity, as well as other phenomena like community nestedness. These indices are often mathematically convertible to each other and are thus equivalent in terms of their information content. Based on this information equivalence, we classified all the indices to a few families, showing that only three families reflect ecologically relevant and directly interpretable phenomena, namely species turnover (family of Jaccard index that also includes Sørensen index of similarity), nestedness (the family of indices which compare species overlap with species richness of the species‐poor community), and the uniformity of species richness (comparing species richness of the two communities). Importantly, our analysis shows that any attempt to partition indices, including Baselga's approach to partition turnover and nestedness (i.e. to control an index for an effect of a different phenomenon), leads either to an index belonging to one of the three abovementioned families, or produces indices that do not measure any ecologically relevant phenomenon. We provide guidance on how to apply pairwise indices to make proper inference about ecological phenomena.
Importance:Previous studies suggest that critically injured patients are at increased risk of suicide following discharge, but these have mainly been single-center studies or had limited data on comorbidities and socioeconomic factors. Objective:To examine the risk of suicide after hospitalization for traumatic injuries. Design, Setting, and Participants:This nationwide register-based cohort study used data from 5 Norwegian nationwide health registries and Statistics Norway between 2014 and 2020. Patients registered in the Norwegian Trauma Registry (NTR) for whom trauma team activation at hospital arrival was mandated by national guidelines between 2015 and 2018 were matched to general-population controls on gender and birth year in a 1:10 ratio according to a matched cohort design. Patients discharged alive were followed-up for a minimum of 2 years. All Norwegian hospitals treating patients with traumatic injuries provide data to the NTR. The final analysis was conducted in April 2025. Exposures:Traumatic injury, admitted to hospital, and registered in the NTR. Main Outcomes and Measures:The outcome of interest was suicide, as registered in the Norwegian Cause of Death Registry. Cumulative incidence ratios (CIR) of suicide with 95% CIs, taking nonsuicidal death as a competing event into account, were estimated. Adjustments for Charlson Comorbidity Index, previous psychiatric illness, and socioeconomic position (SEP) were conducted using inverse probability of treatment weights. Results:A total of 25 536 patients with traumatic injuries (165 897 [67%] male; mean [SD] age, 41 [23] years) were matched to 247 095 controls, with a mean (SD) age of 41 (23) years and 68% male. The cumulative incidences of suicide were 0.18% at 2 years and 0.34% and 5 years for patients with traumatic injuries and 0.02% at 2 years and 0.05% at 5 years for controls (2-year CIR, 9.3 [95% CI, 5.4-13.0]; 5-year CIR, 6.9 [95% CI, 4.4-9.1]). Patients with traumatic injury were older at the age of suicide compared with controls (mean [SD] age, 43 [19] years vs 36 [17] years; P = .03) and female patients with traumatic injury had higher incidence of suicide compared with female controls (36% vs 17%; P = .005). Conclusions and Relevance:In this cohort study of patients in Norway discharged alive after critical injury, a 9-fold increased risk of suicide after 2 years was observed. These findings suggest that follow-up is warranted for possible psychological distress in this patient group.
Understanding how genetic drift and inbreeding can shape phenotypic traits and the expression of rare genetic variants can be important for conservation management in small and isolated populations. Following the recolonization of the grey wolf (Canis lupus) on the Scandinavian peninsula, the population has remained small and semi-isolated. The population traces back to seven founders, only and is highly inbred. Inbreeding depression has been demonstrated including effects on reproductive output and several congenital deformities. Some wolves also show anomalous coat color, characterized by hypopigmentation displayed as white tail tips, or even larger areas of white fur. In this study, we utilized four decades of monitoring data, dating back to the recolonization of the Scandinavian peninsula in 1983, to examine the occurrence of expressed anomalous coat color investigating its origin and inheritance, genetic architecture, and the effect of inbreeding. SNP genotyping revealed a distinct haplotype linked to the Melanocyte-inducing transcription factor (MITF) gene that co-segregated with anomalous coat color, suggesting a direct causal effect of this variant. The MITF gene regulates mammal melanocyte development, which in turn affects pigmentation. Our findings demonstrate that the identified gene variant is recessive, which in the homozygous state likely causes the disruption of normal melanocyte development, leading to unpigmented or hypopigmented areas. The origin of this haplotype was traced back to a third founder, reproducing for the first time in 1991. Indeed, the genetic constraints and subsequent inbreeding shaped by few founders, small population size, and semi-isolation over several decades point towards the importance of genetic diversity and facilitated gene flow between populations, but also how such vital immigration can bring about unforeseen side effects if inbreeding continues within the immigrant lineages.
Serving large language models under latency service-level objectives (SLOs) is a configuration-heavy systems problem with an unusually failure-prone search space: many plausible configurations crash outright or miss user-visible latency targets, and standard black-box optimizers treat these failures as wasted trials. We present SLO-Guard, a crash-aware autotuner for vLLM serving that treats crashes as first-class observations. SLO-Guard combines a feasible-first Thermal Budget Annealing (TBA) exploration phase with a warm-started Tree-structured Parzen Estimator (TPE) exploitation phase; the handoff replays all exploration history, including crashes encoded as extreme constraint violations. We additionally contribute a configuration-repair pass, a GPU-aware KV-cache memory guard, and a four-category crash taxonomy. We evaluate SLO-Guard on Qwen2-1.5B served with vLLM 0.19 on an NVIDIA A100 40GB. Across a pre-specified five-seed study, both SLO-Guard and uniform random search attain 75/75 feasibility with zero crashes under the corrected concurrent harness, and are statistically tied on best-achieved latency (Mann-Whitney two-sided p=0.84). SLO-Guard's advantage is in budget consistency: more trials in the fast-serving regime (10.20 vs. 7.40 out of 15; one-sided p=0.014) and higher post-handoff consistency (0.876 vs. 0.539; p=0.010). Under concurrent load, SLO-Guard's cross-seed standard deviation on best latency is 4.4x tighter than random search's (2.26 ms vs. 10.00 ms). A harness-replication analysis shows that the consistency findings survive an independent sequential-dispatch measurement condition. The central claim is not that SLO-Guard finds a better final configuration, but that it spends a fixed tuning budget more predictably once the fast regime has been found.
Retrieving the few past turns that answer a new query across long multi-session histories is the retrieval bottleneck behind long-term conversational memory (LoCoMo, LongMemEval). Recent concurrent work, Nano-Memory, shows that scoring a session by the maximum query-turn similarity (late interaction, "Turn Isolation Retrieval") beats mean-pooled session embeddings. We do not claim that effect; we replicate it and ask what a training-free, CPU-only retrieval stage should add around it. We report four findings. (1) Fuse: score-level fusion of the late-interaction dense score with BM25, under a single leave-one-conversation-out weight, adds +8.8 to +17.2 points of LoCoMo Hit@1 over late interaction alone across six encoders (all p<1e-4), reaching Hit@1 0.752 / NDCG@5 0.829 (e5-large-v2), +11.2 pp over BM25. (2) An off-the-shelf web-search cross-encoder reranker over the fused top-10 hurts here, degrading Hit@1 by 6.9 pp (one reranker, one configuration). (3) A pooling-operator ablation shows top-k late interaction matches max-similarity, but a naive smooth-max (log-sum-exp) collapses for half the encoders. (4) The late-minus-early gap is large for all six encoders and tends to be larger for larger ones, while the marginal fusion gain shrinks; on LongMemEval-S, a lexical regime where BM25 saturates, the net fusion gain over BM25 is small and not significant. A per-category analysis frames the gain as a division of labor: dense late interaction helps most on multi-hop and temporal questions but trails BM25 on adversarial ones. The contribution is a controlled, reproducible account of a strong training-free retrieval recipe, not the late-interaction retriever itself (Nano-Memory's). We make no claim to a complete memory architecture; this is a retrieval-stage study.
