Laura Russo, Caleb Allen, Cameron S. Jorgensen
et al.
Scientists have long been fascinated by magnetoreception, the innate capacity of many animals to sense and use the Earth's magnetic field for navigation. In eusocial insects like honey bees, magnetoreception has been linked to communication and foraging. However, little is known about magnetoreception's phylogenetic patterns and relationship to species traits and natural history. Here, we demonstrate that putative magnetoreception based on ferromagnetic particles is widespread across a diversity of bee species (72 out of 96 species tested), with no phylogenetic signal. We also detected such putative magnetoreception in non-bee outgroups, suggesting this magnetic capacity predates the evolution of the Anthophila. While magnetic signals were found across a diversity of life history traits, the strength of the magnetic signal varied within and between species, and increased with body size and social behavior.
This paper studies off-policy evaluation (OPE) in reinforcement learning with a focus on behavior policy estimation for importance sampling. Prior work has shown empirically that estimating a history-dependent behavior policy can lead to lower mean squared error (MSE) even when the true behavior policy is Markovian. However, the question of why the use of history should lower MSE remains open. In this paper, we theoretically demystify this paradox by deriving a bias-variance decomposition of the MSE of ordinary importance sampling (IS) estimators, demonstrating that history-dependent behavior policy estimation decreases their asymptotic variances while increasing their finite-sample biases. Additionally, as the estimated behavior policy conditions on a longer history, we show a consistent decrease in variance. We extend these findings to a range of other OPE estimators, including the sequential IS estimator, the doubly robust estimator and the marginalized IS estimator, with the behavior policy estimated either parametrically or non-parametrically.
Weiqin Chen, Xinjie Zhang, Dharmashankar Subramanian
et al.
Transformer models (TMs) have exhibited remarkable in-context reinforcement learning (ICRL) capabilities, allowing them to generalize to and improve in previously unseen environments without re-training or fine-tuning. This is typically accomplished by imitating the complete learning histories of a source RL algorithm over a substantial amount of pretraining environments, which, however, may transfer suboptimal behaviors inherited from the source algorithm/dataset. Therefore, in this work, we address the issue of inheriting suboptimality from the perspective of dataset preprocessing. Motivated by the success of the weighted empirical risk minimization, we propose a simple yet effective approach, learning history filtering (LHF), to enhance ICRL by reweighting and filtering the learning histories based on their improvement and stability characteristics. To the best of our knowledge, LHF is the first approach to avoid source suboptimality by dataset preprocessing, and can be combined with the current state-of-the-art (SOTA) ICRL algorithms. We substantiate the effectiveness of LHF through a series of experiments conducted on the well-known ICRL benchmarks, encompassing both discrete environments and continuous robotic manipulation tasks, with three SOTA ICRL algorithms (AD, DPT, DICP) as the backbones. LHF exhibits robust performance across a variety of suboptimal scenarios, as well as under varying hyperparameters and sampling strategies. Notably, the superior performance of LHF becomes more pronounced in the presence of noisy data, indicating the significance of filtering learning histories.
Recent cosmological data and astrophysical observations, such as the Hubble tension and the increasing preference from galaxy surveys for dynamical dark energy, have begun to challenge the standard $Λ$-cold dark matter cosmological model. Primordial magnetic fields (PMFs) offer a mechanism to alleviate these tensions within the framework of the standard model. These fields source excess small-scale baryon clumping, which can speed up recombination and shrink the comoving sound horizon at the surface of last scattering. Computing the modified recombination history requires coupling the radiative transport of Lyman-$α$ photons to compressible magnetohydronamic simulations. Since doing so is generically computationally intractable, we have developed a linearized treatment which self-consistently computes the modified recombination history in the presence of PMF induced baryon clumping for fields with red-tilted spectra. The clumping factors we find are too small to alleviate outstanding cosmological tensions, but our general framework can be applied to other PMF spectra, and provides a significant theoretical step towards a complete account of recombination in the presence of small-scale baryon clumping.
Using a scale-free $N$-body simulation generated with the ABACUS $N$-body code, we test the robustness of halo mass accretion histories via their convergence to self-similarity. We compare two halo finders, ROCKSTAR and COMPASO. We find superior self-similarity in halo mass accretion histories determined using ROCKSTAR, with convergence to 5% or better between $\sim10^2$ to $10^5$ particles. For COMPASO we find weaker convergence over a similar region, with at least 10% between $\sim10^2$ to $10^4$ particles. Furthermore, we find the convergence to self-similarity improves as the simulation evolves, with the largest and deepest regions of convergence appearing after the scale factor quadrupled from the time at which non-linear structures begin to form. With sufficient time evolution, halo mass accretion histories are converged to self-similarity within 5% with as few as $\sim70$ particles for COMPASO and within 2% for as few as $\sim30$ particles for ROCKSTAR.
In this paper, we study how open-source large language models (LLMs) can be effectively deployed for improving query rewriting in conversational search, especially for ambiguous queries. We introduce CHIQ, a two-step method that leverages the capabilities of LLMs to resolve ambiguities in the conversation history before query rewriting. This approach contrasts with prior studies that predominantly use closed-source LLMs to directly generate search queries from conversation history. We demonstrate on five well-established benchmarks that CHIQ leads to state-of-the-art results across most settings, showing highly competitive performances with systems leveraging closed-source LLMs. Our study provides a first step towards leveraging open-source LLMs in conversational search, as a competitive alternative to the prevailing reliance on commercial LLMs. Data, models, and source code will be publicly available upon acceptance at https://github.com/fengranMark/CHIQ.
Natural Language Processing (NLP) plays a pivotal role in the realm of Digital Humanities (DH) and serves as the cornerstone for advancing the structural analysis of historical and cultural heritage texts. This is particularly true for the domains of named entity recognition (NER) and relation extraction (RE). In our commitment to expediting ancient history and culture, we present the ``Chinese Historical Information Extraction Corpus''(CHisIEC). CHisIEC is a meticulously curated dataset designed to develop and evaluate NER and RE tasks, offering a resource to facilitate research in the field. Spanning a remarkable historical timeline encompassing data from 13 dynasties spanning over 1830 years, CHisIEC epitomizes the extensive temporal range and text heterogeneity inherent in Chinese historical documents. The dataset encompasses four distinct entity types and twelve relation types, resulting in a meticulously labeled dataset comprising 14,194 entities and 8,609 relations. To establish the robustness and versatility of our dataset, we have undertaken comprehensive experimentation involving models of various sizes and paradigms. Additionally, we have evaluated the capabilities of Large Language Models (LLMs) in the context of tasks related to ancient Chinese history. The dataset and code are available at \url{https://github.com/tangxuemei1995/CHisIEC}.
Christopher Solinas, Douglas Rebstock, Nathan R. Sturtevant
et al.
Historically applied exclusively to perfect information games, depth-limited search with value functions has been key to recent advances in AI for imperfect information games. Most prominent approaches with strong theoretical guarantees require subgame decomposition - a process in which a subgame is computed from public information and player beliefs. However, subgame decomposition can itself require non-trivial computations, and its tractability depends on the existence of efficient algorithms for either full enumeration or generation of the histories that form the root of the subgame. Despite this, no formal analysis of the tractability of such computations has been established in prior work, and application domains have often consisted of games, such as poker, for which enumeration is trivial on modern hardware. Applying these ideas to more complex domains requires understanding their cost. In this work, we introduce and analyze the computational aspects and tractability of filtering histories for subgame decomposition. We show that constructing a single history from the root of the subgame is generally intractable, and then provide a necessary and sufficient condition for efficient enumeration. We also introduce a novel Markov Chain Monte Carlo-based generation algorithm for trick-taking card games - a domain where enumeration is often prohibitively expensive. Our experiments demonstrate its improved scalability in the trick-taking card game Oh Hell. These contributions clarify when and how depth-limited search via subgame decomposition can be an effective tool for sequential decision-making in imperfect information settings.
We present a model-independent reconstruction of the early expansion and thermal histories of the universe, obtained from light element abundance measurements. The expansion history is tightly constrained around the onset of the Big Bang Nucleosynthesis (BBN). The temperature of photons is additionally constrained around the time of neutrino decoupling. Allowing for perturbations to the standard expansion rate, we find that the radiation energy density is constrained to within 15% of its $Λ$CDM value, and only 1% extra matter energy density is allowed around the epoch of BBN. We introduce a new and general analytic fitting formula for the temperature variation, which is flexible enough to reproduce the signal of large classes of beyond-CDM particle models that can alter the temperature through early-time energy injection. We present its constraints from BBN data and from the measurements of effective number of relativistic species and helium-4 abundance probed by the Cosmic Microwave Background radiation anisotropy. Our results provide clarity on the most fundamental properties of the early universe, reconstructed with minimal assumptions about the unknown physics that can occur at keV--MeV energy scales and can be mapped to broad classes of models of interest to cosmology.
AbstractThe activity of Cardinal Stefan Wyszyński as the Primate of Poland occurred in a period when Poland did not maintain diplomatic relations with the Holy See. The period between 1948 and 1981 can be divided, from the perspective of the Warsaw authorities, into three subperiods: no relations and no talks (informal and official), 1948–1965; negotiations, 1965–1974; and working contacts, 1974–1981. The years 1964–1978 were also the period of the apogee of the Vatican's Ostpolitik carried out by Msgr Agostino Casaroli, under the auspices of Pope Paul VI. Cardinal Wyszyński was directly involved four times in talks on the arrangement of Polish–Vatican relations: in 1951, 1957, 1963, and 1965. With time, however, the primate gained more and more distance from the purposefulness of establishing relations between the Holy See and communist states. It should be added that from the beginning of the 1970s, the primate was quite critical of the effects of the Vatican's so-called Ostpolitik. At the end of the pontificate of Paul VI, an open conflict arose between the Secretariat of State and the Primate of Poland, which was resolved only with the election of Karol Wojtyła as Pope.
According to the common wisdom, between a fraction of the mHz and few Hz the spectral energy density of the inflationary gravitons can be safely disregarded even assuming the most optimistic sensitivities of the space-borne detectors. In this analysis we show that this conclusion is evaded if, prior to nucleosynthesis, the post-inflationary evolution includes a sequence of stages expanding either faster or slower than radiation. As a consequence, contrary to the conventional lore, it is shown that below a fraction of the Hz the spectral energy density of the relic gravitons may exceed (even by eight orders of magnitude) the signal obtained under the hypothesis of radiation dominance throughout the whole expansion history prior to the formation of light nuclei. Since the slopes and the amplitudes of the spectra specifically reflect both the inflationary dynamics and the subsequent decelerated evolution, it is possible to disentangle the contribution of the relic gravitons from other (late-time) bursts of gravitational radiation associated, for instance, with a putative strongly first-order phase transition at the TeV scale. Hence, any limit on the spectral energy density of the relic gravitons in the mHz range simultaneously constrains the post-inflationary expansion history and the inflationary initial data.
Spam!: that's what Lorrie Faith Cranor and Brian LaMacchia exclaimed in the title of a popular call-to-action article that appeared twenty years ago on Communications of the ACM. And yet, despite the tremendous efforts of the research community over the last two decades to mitigate this problem, the sense of urgency remains unchanged, as emerging technologies have brought new dangerous forms of digital spam under the spotlight. Furthermore, when spam is carried out with the intent to deceive or influence at scale, it can alter the very fabric of society and our behavior. In this article, I will briefly review the history of digital spam: starting from its quintessential incarnation, spam emails, to modern-days forms of spam affecting the Web and social media, the survey will close by depicting future risks associated with spam and abuse of new technologies, including Artificial Intelligence (e.g., Digital Humans). After providing a taxonomy of spam, and its most popular applications emerged throughout the last two decades, I will review technological and regulatory approaches proposed in the literature, and suggest some possible solutions to tackle this ubiquitous digital epidemic moving forward.
I will discuss the six previous and present long-baseline neutrino experiments: two first-generation general experiments, K2K and MINOS, two specialized experiments, OPERA and ICARUS, and two second-generation general experiments, T2K and NOvA. The motivations for and goals of each experiment, the reasons for the choices that each experiment made, and the outcomes will be discussed.
Johannes Kepler described the Copernican universe as consisting of a central, small, brilliant sun with its planetary system, all surrounded by giant stars. These stars were far larger than, and much dimmer than, the sun -- his De Stella Nova shows that every visible star must exceed the size of the Earth's orbit, and the most prominent stars may exceed the size of the entire planetary system. His other writings, including his response to Ingoli, his Dissertatio cum Nuncio Sidereo, and his Epitome Astronomiae Copernicanae, also reflect this Copernican universe. To Kepler, such a universe was an illustration of divine power -- and solid evidence against the stars being suns, against the universe of Giordano Bruno. Kepler's starry universe was in fact the Copernican universe supported by observations of the stars, which showed them to have measureable apparent sizes. Not until the later seventeenth century were those apparent sizes shown to be spurious, allowing for a universe in which the stars were suns.
Thermal history models, that have been used to understand the geological history of Earth, are now being coupled to climate models to map conditions that allow planets to maintain surface water over geologic time - a criteria considered crucial for life. However, the lack of intrinsic uncertainty assessment has blurred guidelines for how thermal history models can be used toward this end. A model, as a representation of something real, is not expected to be complete. Unmodeled effects are assumed to be small enough that the model maintains utility for the issue(s) it was designed to address. The degree to which this holds depends on how unmodeled factors affect the certainty of model predictions. We quantify this intrinsic uncertainty for several parameterized thermal history models (a widely used subclass of planetary models). Single perturbation analysis is used to determine the reactance time of different models. This provides a metric for how long it takes low amplitude, unmodeled effects to decay or grow. Reactance time is shown to scale inversely with the strength of the dominant feedback (negative or positive) within a model. A perturbed physics analysis is then used to determine uncertainty shadows for model outputs. This provides probability distributions for model predictions and tests the structural stability of a model. That is, do model predictions remain qualitatively similar, and within assumed model limits, in the face of intrinsic uncertainty. Once intrinsic uncertainty is accounted for, model outputs/predictions and comparisons to observational data should be treated in a probabilistic way.
Andreea S. Font, Ian G. McCarthy, Amandine M. C. Le Brun
et al.
[Abridged] Typical disc galaxies forming in a LambdaCDM cosmology encounter a violent environment, where they often experience mergers with massive satellites. The fact that disc galaxies are ubiquitous in the local Universe suggests that a quiescent history is not necessary for their formation. Modern cosmological simulations can now obtain relatively realistic populations of disc galaxies, but it still remains to be clarified how discs manage to survive massive mergers. Here we use a suite of high-resolution hydrodynamical simulations set in a LambdaCDM cosmology to elucidate the fate of discs encountering massive mergers. We extract a sample of approximately 100 disc galaxies and follow the changes in their post-merger morphologies, as tracked by their disc-to-total ratios (D/T). We also examine the relations between their present-day morphology, assembly history and gas fractions. We find that approximately half of present-day disc galaxies underwent at least one merger with a satellite of total mass exceeding the host system's stellar mass, a third had mergers with satellites of mass exceeding 3 times the host's stellar mass, and approximately one-sixth had mergers with satellites of mass exceeding 10 times of the host's stellar mass. These mergers lead to a sharp, but often temporary, decrease in the D/T of the hosts, implying that discs are usually disrupted but then quickly re-grow. To do so, high cold gas fractions are required post-merger, as well as a relatively quiescent recent history (over a few Gyrs before z=0). Our results show that discs can form via diverse merger pathways and that quiescent histories are not the dominant mode of disc formation.
One of the longstanding debates in the history of paleontology focuses on the issue of whether or not there have been long term cycles (operating over tens of millions of years) in biodiversity and extinction. Here we consider the history of this debate by connecting the skein from Grabau up to 2008. We focus on the evidence for periodicity that has emerged thus far, and conclude that there is indeed some evidence that periodicity may be real, though of course more work is needed. We also comment on possible causal mechanisms, focusing especially on the motion of our solar system in the Galaxy. Moreover, we consider the reasons why some scientists have opposed periodicity over the years. Finally, we consider the significance of this for our understanding of evolution and the history of life.
Landauer erasure seems to provide a powerful link between thermodynamics and information processing (logical computation). The only logical operations that require a generation of heat are logically irreversible ones, with the minimum heat generation being $kT \ln 2$ per bit of information lost. Nevertheless, it will be shown logical reversibility neither implies, nor is implied by thermodynamic reversibility. By examining thermodynamically reversible operations which are logically irreversible, it is possible to show that information and entropy, while having the same form, are conceptually different.
Images of five fields in the Local Group dwarf irregular galaxy NGC 6822 obtained with the {\it Hubble Space Telescope} in the F555W and F814W filters are presented. Photometry for the stars in these images was extracted using the Point-Spread-Function fitting program HSTPHOT/MULTIPHOT. The resulting color-magnitude diagrams reach down to $V\approx26$, a level well below the red clump, and were used to solve quantitatively for the star formation history of NGC 6822. Assuming that stars began forming in this galaxy from low-metallicity gas and that there is little variation in the metallicity at each age, the distribution of stars along the red giant branch is best fit with star formation beginning in NGC 6822 12-15 Gyr ago. The best-fitting star formation histories for the old and intermediate age stars are similar among the five fields and show a constant or somewhat increasing star formation rate from 15 Gyr ago to the present except for a possible dip in the star formation rate from 3 to 5 Gyr ago. The main differences among the five fields are in the higher overall star formation rate per area in the bar fields as well as in the ratio of the recent star formation rate to the average past rate. These variations in the recent star formation rate imply that stars formed within the past 0.6 Gyr are not spatially very well mixed throughout the galaxy.
Strong constraints on the cosmic star formation history (SFH) have recently been established using ultraviolet and far-infrared measurements, refining the results of numerous measurements over the past decade. Taken together, the most recent and robust data indicate a compellingly consistent picture of the SFH out to redshift z~6, with especially tight constraints for z < 1. There have also been a number of dedicated efforts to measure or constrain the SFH at z~6 and beyond. It is also possible to constrain the normalisation of the SFH using a combination of electron antineutrino flux limits from Super-Kamiokande measurements and supernova rate density measurements. This review presents the latest compilation of SFH measurements, and summarises the corresponding evolution for stellar and metal mass densities, and supernova rate densities. The constraints on the normalisation of the cosmic SFH, arising from the combination of the supernova rate measurements and the measurement limit on the supernova electron antineutrino flux, are also discussed.