State Street Corporation

Financial decision-makers face more information than they can directly inspect, making context compression necessary. Yet when large language models (LLMs) compress financial source material, they can alter the investment judgment supported by the original source. We frame this problem as information fidelity: compression loses fidelity when it changes the decision induced by the source. In agentic systems, such losses may recur across intermediate steps and amplify throughout the decision process. Across financial filings and earnings-call transcripts, we find that LLM-based compression can produce fluent and factually plausible compressed contexts that nevertheless alter downstream decisions. We analyze two diagnostic patterns associated with fidelity loss: decontextualization, where salient evidence is retained but separated from the caveats and contextual qualifiers needed for correct interpretation, and model dependency, where different compressors expose different views of the same source. We then propose Agentic Context Compression, which generates multiple candidate compressions and audits their disagreements against the original source. Our results suggest that financial compression should be evaluated not only by efficiency or factuality, but also by its ability to preserve decision-relevant context.

We present galaxy luminosity functions (LFs) for the Dark Energy Spectroscopic Instrument (DESI) DR2 Bright Galaxy Survey (BGS) in the g, r, z, and wl bands over 0.002 < z < 0.6 Our analysis uses updated k-corrections and evolutionary corrections, including new polynomial kcorrection fits derived from BGS Year 1 data that supersede earlier GAMA-based prescriptions. Exploiting the statistical power of DESI, we measure LFs to very faint magnitudes, reaching M-0.1(r)- 5 logh similar to -10. Independent measurements from the North and South survey regions show excellent agreement around the LF knee, but the very small statistical uncertainties reveal that simple analytic forms fail to capture the full LF shape. The bright end departs from a pure exponential decline, while the faint end exhibits complex, non-powerlaw behaviour, including a pronounced upturn at M-0.1(r) - 5 log(h) greater than or similar to - 15, which is stronger for red galaxies than for blue. We show that our LFs are largely complete for galaxies with surface brightness mu(50) < 25 and that an apparent steepening fainter than -13 is driven primarily by local overdensity and fragmentation of large galaxies. A systematic North-South offset at the brightest magnitudes is traced to red galaxies and may reflect shallower North photometry underestimating extended earlytype profiles, although this remains inconclusive. We therefore also provide LFs based on model Petrosian magnitudes. Redshift splitting reveals small but significant residuals, indicating limitations of a simple global evolutionary model. Using the redshift limits of J. Loveday et al. (2012), we find excellent agreement with GAMA, with substantially reduced statistical errors. These measurements provide a precise reference for studies of environmental and population-dependent LFs and for testing galaxy formation models.

Watermelon (Citrullus lanatus) production in the USA was valued at $787 million in 2023, with most of the production concentrated in Florida, California, and Georgia. In the Midwest, Indiana leads watermelon production with a value of $49 million (USDA-NASS, 2024). In August 2024, watermelon plants (cv. Guardsman) showing virus-like symptoms, including mosaic, yellowing, leaf deformation, and wrinkling, were observed in southern Indiana (Knox County). No insect vectors were observed in the fields during the site visits. Samples were collected from three plants in each of four separate watermelon fields. A preliminary serological screening using ImmunoStrip® tests (Agdia, Elkhart, IN) for the general potyvirus group and melon severe mosaic virus was performed. Only the ImmunoStrip® tests for the potyvirus group showed positive results. Total RNA was extracted from a symptomatic representative sample from each field using the Spectrum Plant Total RNA kit (Sigma-Aldrich, St. Louis, MO). A pool of total RNA from four individual samples was then submitted for high-throughput sequencing (HTS) using an Illumina NovaSeqX platform as 150 bp paired-end reads (Psomagen, Rockville, MD). The library was prepared using the TruSeq Stranded Total RNA with Plant Ribo-Zero kit. HTS generated 58 million reads. Sequence reads were trimmed with BBDuk software, de novo assembled using SPAdes v3.15, and submitted to the NCBI. Viral contigs were identified using Blastx against the full non-redundant database (as of November 2024). Watermelon crinkle leaf-associated virus 1 (WCLaV-1) was assembled into three contigs: 6.6 Kb viral polymerase (RNA1, 47,009 reads; coverage 1,122X; acc. PQ621806), 1.4 Kb movement protein (RNA2, 6,932 reads; coverage 1,193X; acc. PQ621807), and 1.3 Kb nucleocapsid (RNA2, 5,300 reads; coverage 1,006X; acc. PQ621808). The highest level of nucleotide identity detected ranged from 98.0-99.9% compared with WCLaV-1 isolate Pi-01 from Brazil (LC636068-9). Additionally, a nearly full-length watermelon mosaic virus (WMV) genome (~10 Kb) with >99% identity compared with WMV isolate C05-270 from France (EU660585) was assembled (17,871,776 reads; coverage 24,638X; acc. PQ855772). Putative mycovirus sequences were also detected through HTS; however, further analysis was not conducted due to the incompleteness of the sequences and the limited number of hits. The occurrence of each virus in the individual samples was further validated by RT-PCR using sequence-specific primers designed in this study. WCLaV-1 was detected in 2/4 watermelon samples, while WMV was detected in 2/4 samples, with one sample being infected by both viruses. Three amplicons were sequenced using Nanopore Technology (Plasmidsaurus, Eugene, OR), compared to homologous sequences in GenBank, and uploaded to the NCBI under accessions PV030975-PV030977. Interestingly, HTS did not detect WCLaV-2, commonly detected in mixed infections with WCLaV-1. Although WCLaV-1 and WCLaV-2 were previously associated with symptoms (Xin et al. 2017), the effect on growth and yield is still unknown (Hendricks and Roberts 2023). To the best of our knowledge, this is the first report of WCLaV-1 and WMV infecting watermelons in Indiana, highlighting the need for additional research on the impact of single and mixed infections of these viruses on cucurbit crops.

Work on the risks of artificial intelligence has focused predominantly on capability risk: the danger that systems become too powerful, too autonomous, or too misaligned with human values. Far less attention has been paid to the organizational level—to whether the institutions building these systems are themselves predisposed to drift toward failure. This paper argues that they are. Regardless of how capable AI systems become, the organizations building them face the same structural dynamics that preceded past major technological disasters. Drawing on case studies of the Space Shuttle Challenger, the Three Mile Island accident, and the Boeing 737 MAX crashes, this paper identifies the common structural mechanisms preceding each failure and maps them onto contemporary AI development. The findings suggest that existing safety infrastructure may provide less protection than it appears, as organizations can complete safety processes in full compliance and still produce catastrophic outcomes. The pre-disaster period of AI development is still underway; the purpose of this paper is to make these dynamics legible while they can still be interrupted.

The Photometric objects Around Cosmic webs (PAC) method integrates cosmological photometric and spectroscopic surveys, offering valuable insights into galaxy formation. PAC measures the excess surface density of photometric objects, n 2wp, with specific physical properties around spectroscopic tracers. In this study, we improve the PAC method to make it more rigorous and eliminate the need for redshift bins. We apply the enhanced PAC method to the DESI Y1 BGS Bright spectroscopic sample and the deep Dark Energy Camera Legacy Survey (DECaLS) photometric sample, obtaining n 2wp measurements across the complete stellar mass range, from 105.3 to 1011.5 M circle dot for blue galaxies, and from 106.3 to 1011.9 M circle dot for red galaxies. We combine n 2wp with wp measurements from the BGS sample, which is not necessarily complete in stellar mass. Assuming that galaxy bias is primarily determined by stellar mass and colour, we derive the galaxy stellar mass functions (GSMFs) down to 105.3 M circle dot for blue galaxies and 106.3 M circle dot for red galaxies, while also setting lower limits for smaller masses. The blue and red GSMFs are well described by single and double Schechter functions, respectively, with low-mass end slopes of alpha blue = -1.54+0.02 -0.02 and alpha red = -2.50+0.08 -0.08, resulting in the dominance of red galaxies below 107.6 M circle dot. Stage-IV cosmological photometric surveys, capable of reaching 2-3 mag deeper than DECaLS, present an opportunity to explore the entire galaxy population in the local universe with PAC. This advancement allows us to address critical questions regarding the nature of dark matter, the physics of reionization, and the formation of dwarf galaxies.