IZA DP No. 14839 NOVEMBER 2021 Formalized Employee Search and Labor Demand* Firms often use social networks to find workers, limiting the pool of potential applicants. We conduct a field experiment subsidizing firms’ formal vacancy posting. The subsidies increase non-network employee search and shift vacancies towards high-skilled positions. Post-treatment, firms continue searching for high-skilled workers despite reverting to network-based search. This change in skill requirements does not increase vacancy posting or hiring, suggesting substitutability between workers of different skill levels. Finally, we experimentally show that information asymmetries about applicants’ skills do not limit firms’ formal search. Our results highlight that exposure to different labor market segments can permanently change firms’ labor demand. JEL Classification: D22, J23, J46, C93
Baleen whales (mysticetes) underwent a profound transition from raptorial to filter feeding early in their history, allowing the exploitation of new food sources and environments. Their key structural innovation, baleen, is rarely fossilized, complicating reconstruction of the transition. Muscles used in food capture, locomotion, and respiration are anchored on developmentally discrete subunits of the sternum in living mysticetes, offering a novel approach for predicting muscle function and feeding style in extinct taxa. Elliptic Fourier Analysis was used to test the relationship between sternal shape and feeding style in extant balaenopteroid mysticetes. Results indicate that sternal shape changes during the ontogenetic transition from juvenile nursing to adult filter feeding and differs among adults of species with different styles of filter feeding, providing support for the use of sternal structure to predict the feeding habits of extinct taxa. The dense, block-shaped presterna and multiple sternebrae of early mysticetes suggest that they used static buoyancy control in nearshore environments and lacked the thoracic compliance necessary for deep diving. Variably enlarged sternohyoid muscle fields indicate that they all employed suction during prey capture. In sharp contrast, eomysticetes have osteoporotic, planar sterna without sternebrae and minimal sternohyoid fields, signaling deeper water habitats, diving facilitated by thoracic compliance, and little or no suction. Taxa crownward of eomysticetes display novel elaborations of the sternohyoid and sternocephalic muscle fields that characterize continuous and intermittent ram-feeding in living balaenopteroids.
We present the first data release of the Canadian NIRISS Unbiased Cluster Survey (CANUCS), a JWST Cycle 1 GTO program targeting five lensing clusters and flanking fields in parallel (A370, MACS0416, MACS0417, MACS1149, and MACS1423; survey area ∼100 arcmin ^2 ), with NIRCam imaging, NIRISS slitless spectroscopy, and NIRSpec prism multiobject spectroscopy. Fields centered on cluster cores include imaging in eight bands from 0.9–4.4 μ m, alongside continuous NIRISS coverage from 1.15–2 μ m, while the NIRCam flanking fields provide five wide-band and nine medium-band filters for exceptional spectral sampling, all to ∼29 mag _AB . We also present JWST in Technicolor, a Cycle 2 follow-up GO program targeting three CANUCS clusters (A370, MACS0416, and MACS1149). The Technicolor program adds NIRISS slitless spectroscopy in F090W to the cluster fields while adding eight wide-, medium-, and narrowband filters to the flanking fields. This provides NIRCam imaging in all wide- and medium-band filters over ∼30 arcmin ^2 . This paper describes our data reduction and photometry methodology. We release NIRCam, NIRISS, and Hubble Space Telescope imaging, point-spread functions (PSFs), PSF-matched imaging, photometric catalogs, and photometric and spectroscopic redshifts. We provide lens models and stellar population parameters in up to 19 filters for ∼53,000 galaxies in the cluster fields, and ∼44,000 galaxies in up to 29 filters in the flanking fields. We further present 733 NIRSpec spectra and redshift measurements up to z = 10.8. Comparing against our photometric redshifts, we find catastrophic outlier rates of only 4%–7% and scatter of σ _NMAD of 0.01–0.03.
Large language models (LLMs) already excel at writing code in high-resource languages such as Python and JavaScript, yet stumble on low-resource languages that remain essential to science and engineering. Besides the obvious shortage of pre-training data, post-training itself is a bottleneck: every new language seems to require new datasets, test harnesses, and reinforcement learning (RL) infrastructure. We introduce Agnostics, a language-agnostic post-training pipeline that eliminates this per-language engineering. The key idea is to judge code solely by its externally observable behavior, so a single verifier can test solutions written in any language. Concretely, we (i) use an LLM to rewrite existing unit-test datasets into an I/O format, (ii) supply a short configuration that tells the verifier how to compile and run a target language, and (iii) apply reinforcement learning with verifiable rewards (RLVR) in a robust code execution environment. Applied to five low-resource languages—Lua, Julia, R, OCaml, and Fortran—Agnostics (1) improves Qwen-3 4B to performance that rivals other 16B–70B open-weight models; (2) scales cleanly to larger and diverse model families (Qwen-3 8B, DeepSeek Coder 6.7B Instruct, SmolLM3, Phi 4 Mini); and (3) for open-weight models with ≤16B parameters, sets new state-of-the-art pass@1 results on MultiPL-E and a new multi-language version of LiveCodeBench that we introduce. We release the language-agnostic training datasets (Ag-MBPP-X, Ag-Codeforces-X, Ag-LiveCodeBench-X), training code, and ready-to-use configurations, making RL post-training in any programming language as simple as editing a short YAML file.
Recent studies by B. Villarroel et al. have assembled and analysed datasets of unidentified features measured from digital scans of pre-Sputnik photographic plates. We have examined the claims in these papers using two previously published datasets that are closely related to those used in the Villarroel et al. studies. For these datasets, the assumption of a spatially uniform-random background distribution of features, essential to the Earth shadow analysis, is shown to be false. After finding the null distribution of feature count deviations from the background, we find no statistically significant deficit in the shadow. We find that the reported correlation between the timing of feature observations and nuclear tests becomes insignificant after properly normalizing by the relevant number of observation days, and is almost completely determined by the observation schedule of the Palomar telescope. We uncover important inconsistencies in the definitions of the datasets used in these studies, as well as the use of unvalidated datasets containing catalogue stars, scan artefacts, and plate defects. We find an overall gradual increase in number density of features toward the corners and edges of plates, as well as examples of (i) empty north-south strips that span multiple plates; (ii) clusters and voids having geometric shapes; and (iii) amorphous clusters. We also highlight a circular argument used in these studies, that leverages the results of an inferential analysis to justify conclusions about the origin of the features as well as the validity of the measurements. Finally, we also review the literature concerning historical searches for optical transients in photographic plates corresponding to gamma ray bursts (GRBs); following decades of work, researchers were unable to make a confident identification of a GRB-associated optical transient.
