搜索结果：sight

Multi-agent reinforcement Learning (MARL) is often challenged by the sight range dilemma, where agents either receive insufficient or excessive information from their environment. In this paper, we propose a novel method, called Dynamic Sight Range Selection (DSR), to address this issue. DSR utilizes an Upper Confidence Bound (UCB) algorithm and dynamically adjusts the sight range during training. Experiment results show several advantages of using DSR. First, we demonstrate using DSR achieves better performance in three common MARL environments, including Level-Based Foraging (LBF), Multi-Robot Warehouse (RWARE), and StarCraft Multi-Agent Challenge (SMAC). Second, our results show that DSR consistently improves performance across multiple MARL algorithms, including QMIX and MAPPO. Third, DSR offers suitable sight ranges for different training steps, thereby accelerating the training process. Finally, DSR provides additional interpretability by indicating the optimal sight range used during training. Unlike existing methods that rely on global information or communication mechanisms, our approach operates solely based on the individual sight ranges of agents. This approach offers a

Reinforcement Learning (RL) has empowered Large Language Models (LLMs) to master autonomous search for complex question answering. However, particularly within multi-turn search scenarios, this interaction introduces a critical challenge: search results often suffer from high redundancy and low signal-to-noise ratios. Consequently, agents easily fall into "Tunnel Vision," where the forced interpretation of early noisy retrievals leads to irreversible error accumulation. To address these challenges, we propose SIGHT, a framework that enhances search-based reasoning through Self-Evidence Support (SES) and Information-Gain Driven Diverse Branching. SIGHT distills search results into high-fidelity evidence via SES and calculates an Information Gain score to pinpoint pivotal states where observations maximally reduce uncertainty. This score guides Dynamic Prompting Interventions - including de-duplication, reflection, or adaptive branching - to spawn new branches with SES. Finally, by integrating SES and correctness rewards via Group Relative Policy Optimization, SIGHT internalizes robust exploration strategies without external verifiers. Experiments on single-hop and multi-hop QA bench

Time-Delay Cosmography is a technique for measuring $H_0$ with strong gravitational lensing. It requires a correction for line of sight perturbations, and it is necessary to build tools to assess populations of these lines of sight efficiently. We aim demonstrate the techniques necessary to analyze line of sight effects at a population level, and investigate whether strong lenses fall in preferably overdense environments. We analyze a set of 25 galaxy-galaxy lens lines of sight in the Strong Lensing Legacy Survey sample using standard techniques, then perform a hierarchical analysis to constrain the population-level parameters. We introduce a new statistical model for these posteriors that may provide insight into the underlying physics of the system. We find the median value of $κ_{\rm{ext}}$ in the population model to be $0.033 \pm 0.010$. The median value of $κ_{\rm{ext}}$ for the individual lens posteriors is $0.008 \pm 0.015$. Both approaches demostrate that our systems are drawn from an overdense sample. The different results from these two approaches show the importance of population models that do not multiply the effect of our priors.

In collisions of galaxy clusters, the lack of displacement between dark matter and galaxies suggests that the dark matter scattering depth is small. This yields an upper limit on the dark matter cross section if the dark matter column density is known. We investigate a bias in such constraints: the measured column density (along the line of sight, using gravitational lensing) is lower than that experienced by a dark matter particle, as follows. Dark matter halos are triaxial and generally collide along their major axes, yielding a high scattering column density -- but the merger is obvious only to observers whose line of sight is nearly perpendicular to that axis, yielding a low observed column density. We trace lines of sight through merging halos from the BigMDPL n-body simulation, both with and without mock observational effects. We find that a hypothetical skewer through the center of the halo along the merger axis (more precisely, along the current separation vector of the two halos) has nearly twice the column density of a typical line of sight. With weak lensing measurements, which involve some spatial averaging, this ratio is reduced to 1.25, suggesting that existing constr

Current visual navigation systems often treat the environment as static, lacking the ability to adaptively interact with obstacles. This limitation leads to navigation failure when encountering unavoidable obstructions. In response, we introduce IN-Sight, a novel approach to self-supervised path planning, enabling more effective navigation strategies through interaction with obstacles. Utilizing RGB-D observations, IN-Sight calculates traversability scores and incorporates them into a semantic map, facilitating long-range path planning in complex, maze-like environments. To precisely navigate around obstacles, IN-Sight employs a local planner, trained imperatively on a differentiable costmap using representation learning techniques. The entire framework undergoes end-to-end training within the state-of-the-art photorealistic Intel SPEAR Simulator. We validate the effectiveness of IN-Sight through extensive benchmarking in a variety of simulated scenarios and ablation studies. Moreover, we demonstrate the system's real-world applicability with zero-shot sim-to-real transfer, deploying our planner on the legged robot platform ANYmal, showcasing its practical potential for interactive

The {\it Far Ultraviolet Spectroscopic Explorer} ({\it FUSE}) has allowed precise determinations of the column densities of molecular hydrogen ($\Hmol$) in Galactic lines of sight with a wide range of pathlengths and extinction properties. However, survey studies of lines of sight with greater extinction have been mostly restricted to the low-$J$ states (lower total angular momentum) in which most molecular hydrogen is observed. This paper presents a survey of column densities for the molecular hydrogen in states of greater rotational excitation ($J \geq 2$) in Galactic lines of sight with $\log{\NHmol} \gtrsim 20$. This study is comprehensive through the highest excited state detectable in each line of sight. J=5 is observed in every line of sight, and we detect J=7 in four lines of sight, J=8 in one line of sight, and vibrationally excited $\Hmol$ in two lines of sight. We compared the apparent $b$-values and velocity offsets of the higher-$J$ states relative to the dominant low-$J$ states and we found no evidence of any trends that might provide insight into the formation of higher-$J$ $\Hmol$, although these results are the most affected by the limits of the {\it FUSE} resoluti

We present an analysis of the line-of-sight structure toward a sample of ten strong lensing cluster cores. Structure is traced by groups that are identified spectroscopically in the redshift range, 0.1 $\leq$ z $\leq$ 0.9, and we measure the projected angular and comoving separations between each group and the primary strong lensing clusters in each corresponding line of sight. From these data we measure the distribution of projected angular separations between the primary strong lensing clusters and uncorrelated large scale structure as traced by groups. We then compare the observed distribution of angular separations for our strong lensing selected lines of sight against the distribution of groups that is predicted for clusters lying along random lines of sight. There is clear evidence for an excess of structure along the line of sight at small angular separations ($θ\leq 6'$) along the strong lensing selected lines of sight, indicating that uncorrelated structure is a significant systematic that contributes to producing galaxy clusters with large cross sections for strong lensing. The prevalence of line-of-sight structure is one of several biases in strong lensing clusters that

Lectures are a learning experience for both students and teachers. Students learn from teachers about the subject material, while teachers learn from students about how to refine their instruction. However, online student feedback is unstructured and abundant, making it challenging for teachers to learn and improve. We take a step towards tackling this challenge. First, we contribute a dataset for studying this problem: SIGHT is a large dataset of 288 math lecture transcripts and 15,784 comments collected from the Massachusetts Institute of Technology OpenCourseWare (MIT OCW) YouTube channel. Second, we develop a rubric for categorizing feedback types using qualitative analysis. Qualitative analysis methods are powerful in uncovering domain-specific insights, however they are costly to apply to large data sources. To overcome this challenge, we propose a set of best practices for using large language models (LLMs) to cheaply classify the comments at scale. We observe a striking correlation between the model's and humans' annotation: Categories with consistent human annotations (>$0.9$ inter-rater reliability, IRR) also display higher human-model agreement (>$0.7$), while cate

We use reinforcement meta learning to optimize a line of sight curvature policy that increases the effectiveness of a guidance system against maneuvering targets. The policy is implemented as a recurrent neural network that maps navigation system outputs to a Euler 321 attitude representation. The attitude representation is then used to construct a direction cosine matrix that biases the observed line of sight vector. The line of sight rotation rate derived from the biased line of sight is then mapped to a commanded acceleration by the guidance system. By varying the bias as a function of navigation system outputs, the policy enhances accuracy against highly maneuvering targets. Importantly, our method does not require an estimate of target acceleration. In our experiments, we demonstrate that when our method is combined with proportional navigation, the system significantly outperforms augmented proportional navigation with perfect knowledge of target acceleration, achieving improved accuracy with less control effort against a wide range of target maneuvers.

Beyond Line of Sight (BLOS) communication stands as an indispensable element within defense communication strategies, facilitating information exchange in scenarios where traditional Line of Sight (LOS) methodologies encounter obstruction. This article delves into the forefront of technologies driving BLOS communication, emphasizing advanced systems like phantom networks, nanonetworks, aerial relays, and satellite-based defense communication. Moreover, we present a practical use case of UAV path planning using optimization techniques amidst radar-threat war zones that add concrete relevance, underscoring the tangible applications of BLOS defense communication systems. Additionally, we present several future research directions for BLOS communication in defense systems, such as resilience enhancement, the integration of heterogeneous networks, management of contested spectrums, advancements in multimedia communication, adaptive methodologies, and the burgeoning domain of the Internet of Military Things (IoMT). This exploration of BLOS technologies and their applications lays the groundwork for synergistic collaboration between industry and academia, fostering innovation in defense c

We present extinction curves that include data down to far ultraviolet wavelengths (FUV; 1050 - 1200 A) for nine Galactic sight lines. The FUV extinction was measured using data from the Far Ultraviolet Spectroscopic Explorer. The sight lines were chosen for their unusual extinction properties in the infrared through the ultraviolet; that they probe a wide range of dust environments is evidenced by the large spread in their measured ratios of total-to-selective extinction, R_V = 2.43 - 3.81. We find that extrapolation of the Fitzpatrick & Massa relationship from the ultraviolet appears to be a good predictor of the FUV extinction behavior. We find that predictions of the FUV extinction based upon the Cardelli, Clayton & Mathis (CCM) dependence on R_V give mixed results. For the seven extinction curves well represented by CCM in the infrared through ultraviolet, the FUV extinction is well predicted in three sight lines, over-predicted in two sight lines, and under-predicted in 2 sight lines. A Maximum Entropy Method analysis using a simple three component grain model shows that seven of the nine sight lines in the study require a larger fraction of grain materials to be in d

We consider decision-making and game scenarios in which an agent is limited by his/her computational ability to foresee all the available moves towards the future - that is, we study scenarios with short sight. We focus on how short sight affects the logical properties of decision making in multi-agent settings. We start with single-agent sequential decision making (SSDM) processes, modeling them by a new structure of "preference-sight trees". Using this model, we first explore the relation between a new natural solution concept of Sight-Compatible Backward Induction (SCBI) and the histories produced by classical Backward Induction (BI). In particular, we find necessary and sufficient conditions for the two analyses to be equivalent. Next, we study whether larger sight always contributes to better outcomes. Then we develop a simple logical special-purpose language to formally express some key properties of our preference-sight models. Lastly, we show how short-sight SSDM scenarios call for substantial enrichments of existing fixed-point logics that have been developed for the classical BI solution concept. We also discuss changes in earlier modal logics expressing "surface reasonin

We report total abundances and related parameters for the full sample of the FUSE survey of molecular hydrogen in 38 translucent lines of sight. New results are presented for the "second half" of the survey involving 15 lines of sight to supplement data for the first 23 lines of sight already published. We assess the correlations between molecular hydrogen and various extinction parameters in the full sample, which covers a broader range of conditions than the initial sample. In particular, we are now able to confirm that many, but not all, lines of sight with shallow far-UV extinction curves and large values of the total-to-selective extinction ratio, $R_V$ = $A_V$ / $E(B-V)$ -- characteristic of larger than average dust grains -- are associated with particularly low hydrogen molecular fractions ($f_{\rm H2}$). In the lines of sight with large $R_V$, there is in fact a wide range in molecular fractions, despite the expectation that the larger grains should lead to less H$_2$ formation. However, we see specific evidence that the molecular fractions in this sub-sample are inversely related to the estimated strength of the UV radiation field and thus the latter factor is more importa

We report on the abundance of interstellar neutral nitrogen (NI) for 30 sightlines, using data from the Far Ultraviolet Spectroscopic Explorer (FUSE) and the Hubble Space Telescope (HST). NI column densities are derived by measuring the equivalent widths of several ultraviolet absorption lines and subsequently fitting those to a curve of growth. We find a mean interstellar N/H of 51+/-4 ppm. This is below the mean found by Meyer et al. of 62(+4,-3) ppm (adjusted for a difference in f-values). Our mean N/H is similar, however, to the (f-value adjusted) mean of 51+/-3 ppm found by Knauth et al. for a larger sample of sightlines with larger hydrogen column densities comparable to those in this study. We discuss the question of whether or not nitrogen shows increased gas-phase depletion in lines of sight with column densities log(H_tot) >~ 21, as claimed by Knauth et al. The nitrogen abundance in the line of sight toward HD 152236 is particularly interesting. We derive very small N/H and N/O ratios for this line of sight that may support a previous suggestion that members of the Sco OB1 association formed from an N-deficient region.

We report the second study from the FUSE survey of molecular hydrogen in translucent clouds, for the line of sight toward HD 110432. This star lies beyond the Coalsack dark nebula, and with E(B-V) = 0.40, and A_V = 1.32 this line of sight bridges the gap between less extinguished diffuse cloud lines of sight with A_V \sim 1, such as Zeta Oph, and the translucent clouds with A_V \gtrsim 2 such as HD 73882. Through profile fitting and a curve-of-growth analysis, we have derived rotational populations for H2 for J = 0--7. The line of sight has a total molecular hydrogen column density, log N(H2) = 20.68 \pm 0.05 cm^{-2}, nearly identical to that toward Zeta Oph, but a factor of three less than HD 73882. The ratio of N(J=1) to N(J=0) yields a kinetic temperature T_{kin} = 63 \pm 7 K, similar to other lines of sight with A_V \gtrsim 1. The high-J lines show considerable excitation above this temperature. The high-J excitation is similar to that toward Zeta Oph, but much smaller than that toward HD 73882. Chemical modeling indicates that the physical conditions in the cloud(s) are very similar to those in the cloud(s) toward Zeta Oph. An analysis of IUE spectra of the Lyman-alpha line gi

We present interstellar C II (1334.5323 Å) and C II* (1335.7077 Å) column density measurements along 14 Galactic sight-lines. These sight-lines sample a variety of Galactic disk environments and include paths that range nearly two orders of magnitude in average hydrogen densities (<n(H)>) along the lines of sight. Five of the sight-lines show super-Solar gas phase abundances of carbon. Our results show that the excess carbon along these sight-lines may result from different mechanisms taking place in the regions associated with these stars.

There is evidence of an overdensity of strong intervening MgII absorption line systems distributed along the lines of sight towards GRB afterglows relative to quasar sight-lines. If this excess is real, one should also expect an overdensity of field galaxies around GRB sight-lines, as strong MgII tends to trace these sources. In this work, we test this expectation by calculating the two point angular correlation function of galaxies within 120$^{\prime\prime}$ ($\sim470~h_{71}^{-1}~\mathrm{Kpc}$ at $\langle z\rangle \sim0.4$) of GRB afterglows. We compare the Gamma-ray burst Optical and Near-infrared Detector (GROND) GRB afterglow sample -- one of the largest and most homogeneous samples of GRB fields -- with galaxies and AGN found in the COSMOS-30 photometric catalog. We find no significant signal of anomalous clustering of galaxies at an estimated median redshift of $z\sim0.3$ around GRB sight-lines, down to $K_{\mathrm{AB}}<19.3$. This result is contrary to the expectations from the MgII excess derived from GRB afterglow spectroscopy, although many confirmed galaxy counterparts to MgII absorbers may be too faint to detect in our sample -- especially those at $z>1$. We note

We estimate the galaxy density along lines of sight to hard extragalactic gamma-ray sources by correlating source positions on the sky with a void catalog based on the Sloan Digital Sky Survey (SDSS). Extragalactic gamma-ray sources that are detected at very high energy (VHE; E>100 GeV) or have been highlighted as VHE-emitting candidates in the Fermi Large Area Telescope hard source catalog (together referred to as "VHE-like" sources) are distributed along underdense lines of sight at the 2.4 sigma level. There is also a less suggestive correlation for the Fermi hard source population (1.7 sigma). A correlation between 10-500 GeV flux and underdense fraction along the line of sight for VHE-like and Fermi hard sources is found at 2.4 sigma and 2.6 sigma, respectively. The preference for underdense sight lines is not displayed by gamma-ray emitting galaxies within the second Fermi catalog, containing sources detected above 100 MeV, or the SDSS DR7 quasar catalog. We investigate whether this marginal correlation might be a result of lower extragalactic background light (EBL) photon density within the underdense regions and find that, even in the most extreme case of a entirely unde

We find multiple Mg II absorption systems at redshift z=1.66, 2.069, and 2.097 in the spatially resolved spectra of the quadruply gravitationally lensed quasar H1413+1143 utilizing the Kyoto tridimensional spectrograph II (Kyoto 3DII) spectrograph on board the Subaru telescope. Here we present the first measurement of differences in Mg II absorption strength of the multiple intervening absorbers, which include ones identified as damped Lyman alpha (DLA) absorption systems. Our detection of the significant Mg II absorptions in the spatially resolved spectra reveals the inhomogeneous chemical enrichment on scales of about 12 kpc within the separation of the four sightlines. For the DLA system at z=1.66, the rest equivalent widths of the Mg II absorption lines between the four spatially resolved lines of sight change by factors of up to about 6, which trace the variations in the H I absorption strength. This suggests that inhomogeneous cold absorbers that give rise to the strong H I/Mg II absorptions dwell on a scale of about 6-12 kpc between the four lines of sight. We also investigate the degree of variation in the equivalent width of the absorption lines between the lines of sight.