We investigate the average location of magnetic reconnection on the Earth's dayside magnetopause, based on spatial distributions of northward and southward reconnection jets observed by the THEMIS spacecraft at the near‐noon (10–14 magnetic local time) magnetopause. A total of 711 reconnection jets were identified by applying the Walén relation, the tangential stress balance relation to be satisfied for a reconnected (rotational discontinuity) magnetopause, to magnetopause crossings identified from 10 years of THEMIS observations. The directions and positions of jets indicate that during southward interplanetary magnetic field (IMF) conditions, the dayside X‐line location shifts from the subsolar point toward the winter hemisphere by about 6 Earth radii under the largest tilt of the geomagnetic dipole axis. The X‐line location also shifts northward (southward) by at most 2.5 Earth radii when the IMF is predominantly radial and its x component is positive (negative). The dipole tilt effect on the shift of the X‐line location becomes smaller for higher solar wind Alfvén Mach numbers. The dipole tilt effect being larger than the IMF Bx effect suggests that the X‐line location has a seasonal dependence. Since models and theory show that the reconnection rate away from the subsolar magnetopause is lower than that at the subsolar magnetopause, the dipole tilt dependence of the X‐line location suggests that the efficiency of solar wind energy transfer into the magnetosphere may decrease under larger dipole tilt; this may partially account for seasonal variations of geomagnetic activity, which is known to decrease under larger dipole tilts.

Neural document retrievers, including dense passage retrieval (DPR), have outperformed classical lexical-matching retrievers, such as BM25, when fine-tuned and tested on specific question-answering datasets. However, it has been shown that the existing dense retrievers do not generalize well not only out of domain but even in domain such as Wikipedia, especially when a named entity in a question is a dominant clue for retrieval. In this paper, we propose an approach toward in-domain generalization using the embeddings generated by the frozen language model trained with the entities in the domain. By not fine-tuning, we explore the possibility that the rich knowledge contained in a pretrained language model can be used for retrieval tasks. The proposed method outperforms conventional DPRs on entity-centric questions in Wikipedia domain and achieves almost comparable performance to BM25 and state-of-the-art SPAR model. We also show that the contextualized keys lead to strong improvements compared to BM25 when the entity names consist of common words. Our results demonstrate the feasibility of the zero-shot retrieval method for entity-centric questions of Wikipedia domain, where DPR has struggled to perform.