Search

We introduce GoldFinch, a hybrid Linear Attention/Transformer sequence model that uses a new technique to efficiently generate a highly compressed and reusable KV-Cache in linear time and space with respect to sequence length. GoldFinch stacks our new GOLD transformer on top of an enhanced version of the Finch (RWKV-6) architecture. We train up to 1.5B parameter class models of the Finch, Llama, and GoldFinch architectures, and find dramatically improved modeling performance relative to both Finch and Llama. Our cache size savings increase linearly with model layer count, ranging from 756-2550 times smaller than the traditional transformer cache for common sizes, enabling inference of extremely large context lengths even on limited hardware. Although autoregressive generation has O(n) time complexity per token because of attention, pre-fill computation of the entire initial cache state for a submitted context costs only O(1) time per token due to the use of a recurrent neural network (RNN) to generate this cache. We release our trained weights and training code under the Apache 2.0 license for community use.

We present Eagle (RWKV-5) and Finch (RWKV-6), sequence models improving upon the RWKV (RWKV-4) architecture. Our architectural design advancements include multi-headed matrix-valued states and a dynamic recurrence mechanism that improve expressivity while maintaining the inference efficiency characteristics of RNNs. We introduce a new multilingual corpus with 1.12 trillion tokens and a fast tokenizer based on greedy matching for enhanced multilinguality. We trained four Eagle models, ranging from 0.46 to 7.5 billion parameters, and two Finch models with 1.6 and 3.1 billion parameters and find that they achieve competitive performance across a wide variety of benchmarks. We release all our models on HuggingFace under the Apache 2.0 license. Models at: https://huggingface.co/RWKV Training code at: https://github.com/RWKV/RWKV-LM Inference code at: https://github.com/RWKV/ChatRWKV Time-parallel training code at: https://github.com/RWKV/RWKV-infctx-trainer

Numerical perspectives help people understand extreme and unfamiliar numbers (e.g., \$330 billion is about \$1,000 per person in the United States). While research shows perspectives to be helpful, generating them at scale is challenging both because it is difficult to identify what makes some analogies more helpful than others, and because what is most helpful can vary based on the context in which a given number appears. Here we present and compare three policies for large-scale perspective generation: a rule-based approach, a crowdsourced system, and a model that uses Wikipedia data and semantic similarity (via BERT embeddings) to generate context-specific perspectives. We find that the combination of these three approaches dominates any single method, with different approaches excelling in different settings and users displaying heterogeneous preferences across approaches. We conclude by discussing our deployment of perspectives in a widely-used online word processor.

Recent advances in the development of large language models are rapidly changing how online applications function. LLM-based search tools, for instance, offer a natural language interface that can accommodate complex queries and provide detailed, direct responses. At the same time, there have been concerns about the veracity of the information provided by LLM-based tools due to potential mistakes or fabrications that can arise in algorithmically generated text. In a set of online experiments we investigate how LLM-based search changes people's behavior relative to traditional search, and what can be done to mitigate overreliance on LLM-based output. Participants in our experiments were asked to solve a series of decision tasks that involved researching and comparing different products, and were randomly assigned to do so with either an LLM-based search tool or a traditional search engine. In our first experiment, we find that participants using the LLM-based tool were able to complete their tasks more quickly, using fewer but more complex queries than those who used traditional search. Moreover, these participants reported a more satisfying experience with the LLM-based search tool. When the information presented by the LLM was reliable, participants using the tool made decisions with a comparable level of accuracy to those using traditional search, however we observed overreliance on incorrect information when the LLM erred. Our second experiment further investigated this issue by randomly assigning some users to see a simple color-coded highlighting scheme to alert them to potentially incorrect or misleading information in the LLM responses. Overall we find that this confidence-based highlighting substantially increases the rate at which users spot incorrect information, improving the accuracy of their overall decisions while leaving most other measures unaffected.

SkyPortal is an open-source software package designed to efficiently discover interesting transients, manage follow-up, perform characterization, and visualize the results. By enabling fast access to archival and catalog data, cross-matching heterogeneous data streams, and the triggering and monitoring of on-demand observations for further characterization, a SkyPortal-based platform has been operating at scale for 2 yr for the Zwicky Transient Facility Phase II community, with hundreds of users, containing tens of millions of time-domain sources, interacting with dozens of telescopes, and enabling community reporting. While SkyPortal emphasizes rich user experiences (UX) across common frontend workflows, recognizing that scientific inquiry is increasingly performed programmatically, SkyPortal also surfaces an extensive and well-documented API system. From backend and frontend software to data science analysis tools and visualization frameworks, the SkyPortal design emphasizes the re-use and leveraging of best-in-class approaches, with a strong extensibility ethos. For instance, SkyPortal now leverages ChatGPT large-language models (LLMs) to automatically generate and surface source-level human-readable summaries. With the imminent re-start of the next-generation of gravitational wave detectors, SkyPortal now also includes dedicated multi-messenger features addressing the requirements of rapid multi-messenger follow-up: multi-telescope management, team/group organizing interfaces, and cross-matching of multi-messenger data streams with time-domain optical surveys, with interfaces sufficiently intuitive for the newcomers to the field. (abridged), Comment: Accepted to ApJS

S190426c / GW190426_152155 was the first probable neutron star - black hole merger candidate detected by the LIGO-Virgo Collaboration. We undertook a tiled search for optical counterparts of this event using the 0.7m GROWTH-India Telescope. Over a period of two weeks, we obtained multiple observations over a 22.1 deg^2 area, with a 17.5% probability of containing the source location. Initial efforts included obtaining photometry of sources reported by various groups, and a visual search for sources in all galaxies contained in the region. Subsequently, we have developed an image subtraction and candidate vetting pipeline with ~ 94% efficiency for transient detection. Processing the data with this pipeline, we find several transients, but none that are compatible with kilonova models. We present the details of our observations, working of our pipeline, results from the search, and our interpretations of the non-detections that will work as a pathfinder during the O4 run of LVK., Comment: 12 pages, 6 figures, Accepted for publication in Monthly Notices of the Royal Astronomical Society

An emerging theme across many domains of science and engineering is materials that change shape, often dramatically. Determining their structure involves solving a shape optimization problem where a given energy functional is minimized with respect to the shape of the domain and auxiliary fields describing the structure. Such problems are very challenging to solve and there is a lack of suitable simulation tools that are both readily accessible and general purpose. To address this gap, we present Morpho, an open-source programmable environment, and demonstrate its versatility by showcasing three applications to different areas of soft matter - swelling hydrogels, complex fluids that form aspherical droplets, to soap films and membranes - and advise on broader uses.

A quantitative model of the mobility of functionalized particles at the interface is pivotal to understanding important systems in biology and nanotechnology. In this work, we investigate the emerging dynamics of particles anchored through ligand-receptor bridges to functionalized surfaces. We consider systems with reversible bridges in which ligand-receptor pairs bind/unbind with finite reaction rates. For a given set of bridges, the particle can explore a tiny fraction of the surface as the extensivity of the bridges is finite. We show how at time scales longer than the bridges' lifetime, the averaged position of the particle diffuses away from its initial value. We distill our findings into two analytic equations for the sliding diffusion constant of particles carrying mobile and fixed ligands. We quantitatively validate our theoretical predictions using reaction-diffusion simulations. Our results, along with recent literature, will allow inferring the microscopic parameters at play in complex biological systems from experimental trajectories., Comment: Version accepted for publication in The Journal of Chemical Physics. The first two authors contributed equally

Efficient searches for electromagnetic counterparts to gravitational wave, high-energy neutrino, and gamma-ray burst events demand rapid processing of image arithmetic and geometry set operations in a database to cross-match galaxy catalogs, observation footprints, and all-sky images. Here we introduce HEALPix Alchemy, an open-source, pure Python implementation of a set of methods that enables rapid all-sky geometry calculations. HEALPix Alchemy is built upon HEALPix, a spatial indexing strategy that is widely used in astronomical databases as well as the native format of LIGO-Virgo-KAGRA gravitational-wave sky localization maps. Our approach leverages new multirange types built into the PostgreSQL 14 database engine. This enables fast all-sky queries against probabilistic multimessenger event localizations and telescope survey footprints. Questions such as "What are the galaxies contained within the 90% credible region of an event?" and "What is the rank-ordered list of the fields within an observing footprint with the highest probability of containing the event?" can be performed in less than a few seconds on commodity hardware using off-the-shelf cloud-managed database implementations without server-side database extensions. Common queries scale roughly linearly with the number of telescope pointings. As the number of fields grows into the hundreds or thousands, HEALPix Alchemy is orders of magnitude faster than other implementations. HEALPix Alchemy is now used as the spatial geometry engine within SkyPortal, which forms the basis of the Zwicky Transient Facility transient marshal, called Fritz., Comment: For source code, see https://github.com/skyportal/healpix-alchemy

Interaction-powered supernovae (SNe) explode within an optically-thick circumstellar medium (CSM) that could be ejected during eruptive events. To identify and characterize such pre-explosion outbursts we produce forced-photometry light curves for 196 interacting SNe, mostly of Type IIn, detected by the Zwicky Transient Facility between early 2018 and June 2020. Extensive tests demonstrate that we only expect a few false detections among the 70,000 analyzed pre-explosion images after applying quality cuts and bias corrections. We detect precursor eruptions prior to 18 Type IIn SNe and prior to the Type Ibn SN2019uo. Precursors become brighter and more frequent in the last months before the SN and month-long outbursts brighter than magnitude -13 occur prior to 25% (5 - 69%, 95% confidence range) of all Type IIn SNe within the final three months before the explosion. With radiative energies of up to $10^{49}\,\text{erg}$, precursors could eject $\sim1\,\text{M}_\odot$ of material. Nevertheless, SNe with detected precursors are not significantly more luminous than other SNe IIn and the characteristic narrow hydrogen lines in their spectra typically originate from earlier, undetected mass-loss events. The long precursor durations require ongoing energy injection and they could, for example, be powered by interaction or by a continuum-driven wind. Instabilities during the neon and oxygen burning phases are predicted to launch precursors in the final years to months before the explosion; however, the brightest precursor is 100 times more energetic than anticipated.

LIGO and Virgo's third observing run (O3) revealed the first neutron star-black hole (NSBH) merger candidates in gravitational waves. These events are predicted to synthesize r-process elements creating optical/near-IR "kilonova" (KN) emission. The joint gravitational-wave (GW) and electromagnetic detection of an NSBH merger could be used to constrain the equation of state of dense nuclear matter, and independently measure the local expansion rate of the universe. Here, we present the optical follow-up and analysis of two of the only three high-significance NSBH merger candidates detected to date, S200105ae and S200115j, with the Zwicky Transient Facility (ZTF). ZTF observed $\sim$\,48\% of S200105ae and $\sim$\,22\% of S200115j's localization probabilities, with observations sensitive to KNe brighter than $-$17.5\,mag fading at 0.5\,mag/day in g- and r-bands; extensive searches and systematic follow-up of candidates did not yield a viable counterpart. We present state-of-the-art KN models tailored to NSBH systems that place constraints on the ejecta properties of these NSBH mergers. We show that with depths of $\rm m_{\rm AB}\approx 22$ mag, attainable in meter-class, wide field-of-view survey instruments, strong constraints on ejecta mass are possible, with the potential to rule out low mass ratios, high BH spins, and large neutron star radii.

We present a systematic search for optical counterparts to 13 gravitational wave (GW) triggers involving at least one neutron star during LIGO/Virgo's third observing run. We searched binary neutron star (BNS) and neutron star black hole (NSBH) merger localizations with the Zwicky Transient Facility (ZTF) and undertook follow-up with the Global Relay of Observatories Watching Transients Happen (GROWTH) collaboration. The GW triggers had a median localization of 4480 deg^2, median distance of 267 Mpc and false alarm rates ranging from 1.5 to 1e-25 per yr. The ZTF coverage had a median enclosed probability of 39%, median depth of 20.8mag, and median response time of 1.5 hr. The O3 follow-up by the GROWTH team comprised 340 UVOIR photometric points, 64 OIR spectra, and 3 radio. We find no promising kilonova (radioactivity-powered counterpart) and we convert the upper limits to constrain the underlying kilonova luminosity function. Assuming that all kilonovae are at least as luminous as GW170817 at discovery (-16.1mag), we calculate our joint probability of detecting zero kilonovae is only 4.2%. If we assume that all kilonovae are brighter than -16.6mag (extrapolated peak magnitude of GW170817) and fade at 1 mag/day (similar to GW170817), the joint probability of zero detections is 7%. If we separate the NSBH and BNS populations, the joint probability of zero detections, assuming all kilonovae are brighter than -16.6mag, is 9.7% for NSBH and 7.9% for BNS mergers. Moreover, <57% (<89%) of putative kilonovae could be brighter than -16.6mag assuming flat (fading) evolution, at 90% confidence. If we further account for the online terrestrial probability for each GW trigger, we find that <68% of putative kilonovae could be brighter than -16.6mag. Comparing to model grids, we find that some kilonovae must have Mej < 0.03 Msun or Xlan>1e-4 or phi>30deg to be consistent with our limits. (Abridged), Comment: Submitted to ApJ

We present ZTF20aajnksq (AT2020blt), a fast-fading ($\Delta r=2.4$ mag in $\Delta t=1.3$ days) red ($g-r\approx0.6$ mag) and luminous ($M_{1626}=-25.9$) optical transient at $z=2.9$ discovered by the Zwicky Transient Facility (ZTF). AT2020blt shares several features in common with afterglows to long-duration gamma-ray bursts (GRBs): (1) an optical light curve well-described by a broken power-law with a break at $t_\mathrm{j}=1$ day (observer-frame); (2) a luminous $(L_X = 10^{46}$ $\mathrm{erg}$ $\mathrm{s}^{-1})$ X-ray counterpart; and (3) luminous ($L_\nu = 4 \times 10^{31}$ $\mathrm{erg}$ $\mathrm{sec}^{-1}$ $\mathrm{Hz}^{-1}$ at 10 GHz) radio emission. However, no GRB was detected in the 0.74d between the last ZTF non-detection ($r > 20.64$) and the first ZTF detection ($r = 19.57$), with an upper limit on the isotropic-equivalent gamma-ray energy release of $E_{\gamma,\mathrm{iso}} < 7 \times 10^{52}$ erg. AT2020blt is thus the third afterglow-like transient discovered without a detected GRB counterpart (after PTF11agg and ZTF19abvizsw) and the second (after ZTF19abvizsw) with a redshift measurement. We conclude that the properties of AT2020blt are consistent with a classical (initial Lorentz factor $\Gamma_0 \gtrsim 100$) on-axis GRB that was missed by high-energy satellites. Furthermore, by estimating the rate of transients with light curves similar to that of AT2020blt in ZTF high-cadence data, we agree with previous results that there is no evidence for an afterglow-like phenomenon that is significantly more common than classical GRBs. We conclude by discussing the status and future of fast-transient searches in wide-field high-cadence optical surveys., Comment: 19 pages, 4 figures, 1 table. Version accepted by ApJ on 17 Oct 2020

A growing number of contact tracing apps are being developed to complement manual contact tracing. A key question is whether users will be willing to adopt these contact tracing apps. In this work, we survey over 4,500 Americans to evaluate (1) the effect of both accuracy and privacy concerns on reported willingness to install COVID19 contact tracing apps and (2) how different groups of users weight accuracy vs. privacy. Drawing on our findings from these first two research questions, we (3) quantitatively model how the amount of public health benefit (reduction in infection rate), amount of individual benefit (true-positive detection of exposures to COVID), and degree of privacy risk in a hypothetical contact tracing app may influence American's willingness to install. Our work takes a descriptive ethics approach toward offering implications for the development of policy and app designs related to COVID19.

(Abridged) Using the Zwicky Transient Facility alert stream, we are conducting a large campaign to spectroscopically classify all transients occurring in galaxies in the Census of the Local Universe (CLU) catalog. The aim of the experiment is to construct a spectroscopically complete, volume-limited sample of transients coincident within 100" of CLU galaxies out to 200 Mpc, and to a depth of 20 mag. We describe the survey design and spectroscopic completeness from the first 16 months of operations. We present results from a systematic search for Calcium rich gap transients in the sample of 22 low luminosity (peak absolute magnitude $M > -17$), hydrogen poor events found in the experiment (out of 754 spectroscopically classified SNe). We report the detection of eight Calcium rich gap transients, and constrain their volumetric rate to be at least $\approx 15\pm5$% of the SN Ia rate. Combining this sample with ten events from the literature, we find a likely continuum of spectroscopic properties ranging from events with SN Ia-like features (Ca-Ia objects) to SN Ib/c-like features (Ca-Ib/c objects) at peak light. Within the Ca-Ib/c events, we find two populations of events distinguished by their red ($g - r \approx 1.5$ mag) or green ($g - r \approx 0.5$ mag) spectral colors at $r$-band peak, wherein redder events show strong line blanketing signatures, slower light curves, weaker He lines and lower [Ca II]/[O I] in the nebular phase. Together, we find that the spectroscopic continuum, volumetric rates and striking old environments are consistent with the explosive burning of He shells on low mass white dwarfs. We posit that Ca-Ia and red Ca-Ib/c objects are consistent with the double detonation of He shells with high He burning efficiency, while green Ca-Ib/c objects could arise from less efficient He burning scenarios such as detonations in low density He shells or He shell deflagrations., Comment: 30 figures, 6 tables. Submitted to ApJ

We determine the viability of exploiting lensing time delays to observe strongly gravitationally lensed supernovae (gLSNe) from first light. Assuming a plausible discovery strategy, the Legacy Survey of Space and Time (LSST) and the Zwicky Transient Facility (ZTF) will discover $\sim$ 110 and $\sim$ 1 systems per year before the supernova (SN) explosion in the final image respectively. Systems will be identified $11.7^{+29.8}_{-9.3}$ days before the final explosion. We then explore the possibility of performing early-time observations for Type IIP and Type Ia SNe in LSST-discovered systems. Using a simulated Type IIP explosion, we predict that the shock breakout in one trailing image per year will peak at $\lesssim$ 24.1 mag ($\lesssim$ 23.3) in the $B$-band ($F218W$), however evolving over a timescale of $\sim$ 30 minutes. Using an analytic model of Type Ia companion interaction, we find that in the $B$-band we should observe at least one shock cooling emission event per year that peaks at $\lesssim$ 26.3 mag ($\lesssim$ 29.6) assuming all Type Ia gLSNe have a 1 M$_\odot$ red giant (main sequence) companion. We perform Bayesian analysis to investigate how well deep observations with 1 hour exposures on the European Extremely Large Telescope would discriminate between Type Ia progenitor populations. We find that if all Type Ia SNe evolved from the double-degenerate channel, then observations of the lack of early blue flux in 10 (50) trailing images would rule out more than 27% (19%) of the population having 1 M$_\odot$ main sequence companions at 95% confidence., Comment: 17 pages, 15 figures (including appendices). Accepted by MNRAS 3rd May 2020

Synoptic searches for the optical counterpart to a binary neutron star (BNS) or neutron star-black hole (NSBH) merger can pose significant challenges towards the discovery of kilonovae and performing multi-messenger science. In this work, we describe the advantage of a global multi-telescope network towards this end, with a particular focus on the key and complementary role the Dark Energy Camera (DECam) plays in multi-facility follow-up. We describe the Global Relay of Observatories Watching Transients Happen (GROWTH) Target-of-Opportunity (ToO) Marshal, a common web application we built to ingest events, plan observations, search for transient candidates, and retrieve performance summary statistics for all of the telescopes in our network. Our infrastructure enabled us to conduct observations of two events during O3a, S190426c and S190510g. Furthermore, our analysis of deep DECam observations of S190814bv conducted by the DESGW team, and access to a variety of global follow-up facilities allowed us to place meaningful constraints on the parameters of the kilonova and the merging binary. We emphasize the importance of a global telescope network in conjunction with a power telescope like DECam in performing searches for the counterparts to gravitational-wave sources., Comment: 9 pages, 4 figures, submitted to Rev. Mex. A. A. (AstroRob 2019 conference proceedings)

On 2019 August 14, the Advanced LIGO and Virgo interferometers detected the high-significance gravitational wave (GW) signal S190814bv. The GW data indicated that the event resulted from a neutron star--black hole (NSBH) merger, or potentially a low-mass binary black hole merger. Due to the low false alarm rate and the precise localization (23 deg$^2$ at 90\%), S190814bv presented the community with the best opportunity yet to directly observe an optical/near-infrared counterpart to a NSBH merger. To search for potential counterparts, the GROWTH collaboration performed real-time image subtraction on 6 nights of public Dark Energy Camera (DECam) images acquired in the three weeks following the merger, covering $>$98\% of the localization probability. Using a worldwide network of follow-up facilities, we systematically undertook spectroscopy and imaging of optical counterpart candidates. Combining these data with a photometric redshift catalog, we ruled out each candidate as the counterpart to S190814bv and we placed deep, uniform limits on the optical emission associated with S190814bv. For the nearest consistent GW distance, radiative transfer simulations of NSBH mergers constrain the ejecta mass of S190814bv to be $M_\mathrm{ej} < 0.04$~$M_{\odot}$ at polar viewing angles, or $M_\mathrm{ej} < 0.03$~$M_{\odot}$ if the opacity is $\kappa < 2$~cm$^2$g$^{-1}$. Assuming a tidal deformability for the neutron star at the high end of the range compatible with GW170817 results, our limits would constrain the BH spin component aligned with the orbital momentum to be $ \chi < 0.7$ for mass ratios $Q < 6$, with weaker constraints for more compact neutron stars. We publicly release the photometry from this campaign at http://www.astro.caltech.edu/~danny/static/s190814bv., Comment: Accepted to ApJ with no major changes

We present results from a search for a radio transient associated with the LIGO/Virgo source S190814bv, a likely neutron star-black hole (NSBH) merger, with the Australian Square Kilometre Array Pathfinder. We imaged a $30\,{\rm deg}^2$ field at $\Delta T$=2, 9 and 33 days post-merger at a frequency of 944\,MHz, comparing them to reference images from the Rapid ASKAP Continuum Survey observed 110 days prior to the event. Each epoch of our observations covers $89\%$ of the LIGO/Virgo localisation region. We conducted an untargeted search for radio transients in this field, resulting in 21 candidates. For one of these, \object[AT2019osy]{AT2019osy}, we performed multi-wavelength follow-up and ultimately ruled out the association with S190814bv. All other candidates are likely unrelated variables, but we cannot conclusively rule them out. We discuss our results in the context of model predictions for radio emission from neutron star-black hole mergers and place constrains on the circum-merger density and inclination angle of the merger. This survey is simultaneously the first large-scale radio follow-up of an NSBH merger, and the most sensitive widefield radio transients search to-date.

The Zwicky Transient Facility (ZTF) is conducting a nightly public survey of all 13 TESS northern sectors in 2019-2020. ZTF will observe the portions of the current TESS sectors visible from Palomar Observatory each night. Each ZTF pointing will have one exposure each with $g$ and $r$ filters, totaling two images per night. ZTF is also making additional nightly $g$- and $r$-band observations of denser stellar regions (e.g. near the Galactic Plane) to better facilitate variability studies of Galactic objects. The limiting magnitude of the ZTF observations is $\approx$20.6 and ZTF saturates at magnitude $\approx$13. ZTF will release data from TESS fields in three forms: nightly alerts distributed by established ZTF brokers, nightly alerts converted to JSON format are distributed via ZTF's bucket on Google Cloud as a tarball, and monthly photometric light curves also distributed via Google Cloud.

The LIGO/Virgo gravitational wave events S190828j and S190828l were detected only 21 minutes apart, from nearby regions of sky, and with the same source classifications (binary black hole mergers). It is therefore natural to speculate that the two signals are actually strongly lensed images of the same merger. However, an estimate of the separation of the (unknown) positions of the two events requires them to be >10 deg apart, much wider than the arcsecond-scale separations that usually arise in extragalactic lensing. The large separation is much more consistent with two independent, unrelated events that occurred close in time by chance. We quantify the overlap between simulated pairs of lensed events, and use frequentist hypothesis testing to reject S190828j/l as a lensed pair at 99.8% confidence., Comment: 8 pages, 4 figures, submitted to ApJL

Early-time observations of Type Ia supernovae (SNe Ia) are essential to constrain their progenitor properties. In this paper, we present high-quality light curves of 127 SNe Ia discovered by the Zwicky Transient Facility (ZTF) in 2018. We describe our method to perform forced point spread function (PSF) photometry, which can be applied to other types of extragalactic transients. With a planned cadence of six observations per night ($3g+3r$), all of the 127 SNe Ia are detected in both $g$ and $r$ band more than 10\,d (in the rest frame) prior to the epoch of $g$-band maximum light. The redshifts of these objects range from $z=0.0181$ to 0.165; the median redshift is 0.074. Among the 127 SNe, 50 are detected at least 14\,d prior to maximum light (in the rest frame), with a subset of 9 objects being detected more than 17\,d before $g$-band peak. This is the largest sample of young SNe Ia collected to date; it can be used to study the shape and color evolution of the rising light curves in unprecedented detail. We discuss six peculiar events in this sample, including one 02cx-like event ZTF18abclfee (SN\,2018crl), one Ia-CSM SN ZTF18aaykjei (SN\,2018cxk), and four objects with possible super-Chandrasekhar mass progenitors: ZTF18abhpgje (SN\,2018eul), ZTF18abdpvnd (SN\,2018dvf), ZTF18aawpcel (SN\,2018cir) and ZTF18abddmrf (SN\,2018dsx)., Comment: Accepted by ApJ