arXiv subjects

Dennis Lee

Publications and source records attributed to Dennis Lee.

17 recordsLinked to original sources

arXiv2026

TimeSeek: Temporal Reliability of Agentic Forecasters

We introduce TimeSeek, a benchmark for studying how the reliability of agentic LLM forecasters changes over a prediction market's lifecycle. We evaluate 10 frontier models on 150 CFTC-regulated Kalshi binary markets at five temporal checkpoints, with and without web search, for 15,000 forecasts total. Models are most competitive early in a market's life and on high-uncertainty markets, but much less competitive near resolution and on strong-consensus markets. Web search improves pooled Brier Skill Score (BSS) for every model overall, yet hurts in 12% of model-checkpoint pairs, indicating that retrieval is helpful on average but not uniformly so. Simple two-model ensembles reduce error without surpassing the market overall. These descriptive results motivate time-aware evaluation and selective-deference policies rather than a single market snapshot or a uniform tool-use setting.

cs.AI↗

arXiv2026

SPHEREx mapping of diffuse PAH and H II emission in the Galactic plane

We present preliminary SPHEREx maps of diffuse Galactic emission tracing polycyclic aromatic hydrocarbons (PAHs) and ionized hydrogen gas, and we study their relationship across the Galactic plane. Since its launch in early 2025, the SPHEREx space telescope has been conducting an all-sky near-infrared spectral survey from 0.75 to 5.0 microns. We produce a large-scale map of the 3.3-micron PAH emission feature, which is bright and detectable throughout the Galactic plane, and find a strong correlation with the thermal dust radiance measured by Planck. We also trace ionized hydrogen gas by producing a map of Brackett-alpha emission at 4.05 microns. By combining the two maps, we identify extended shells of PAH emission associated with photodissociation regions surrounding ionized gas. We construct a PAH abundance map and find a significant anticorrelation between PAH abundance and ionized hydrogen, indicating systematic PAH depletion within ionized gas regions across the Galactic plane and demonstrating that ionizing radiation is a dominant driver of PAH abundance variations. These early SPHEREx results provide a large-scale view of PAHs and ionized hydrogen and preview the capability of the mission to map diffuse emission in the interstellar medium.

astro-ph.GA↗

arXiv2026

Searching for the Shortest-wavelength Aromatic Infrared Bands: No Evidence for the Predicted 1.05 $\mu$m Polycyclic Aromatic Hydrocarbon Feature

Polycyclic aromatic hydrocarbons (PAHs) are responsible for a variety of near- and mid-infrared spectral features in Galactic and extragalactic sources. A feature at 1.05 $\mu$m arising from electronic transitions in PAH cations is predicted by laboratory experiments but has never been observationally confirmed. We conduct a dedicated search for this feature in absorption on a highly-extinguished sight line toward BD+40 4223, a blue supergiant in Cyg OB2, using the TripleSpec spectrograph at Palomar Observatory. We place a $5\sigma$ upper limit on the feature strength of $\Delta\tau_{1.05}/A_{V} < 5.6 \times10^{-3}$, ruling out theoretical estimates with $> 10\sigma$ significance. We constrain the effective temperature of BD+40 4223 to be $\log_{10}\left(T_{\rm eff}\right)=4.41\pm0.03$ and infer that it is veiled by $6.39\pm0.05$ magnitudes of visual extinction, consistent with but more constraining than previous determinations. As dust on the sight line toward BD+40 4223 appears typical of the diffuse interstellar medium, this non-detection challenges existing models of PAH material properties and/or charge distribution.

astro-ph.GA↗

arXiv2025

SOFIA Polarization Spectrum of Three Star-Forming Clouds

The dust emission polarization spectrum -- how the polarization percentage changes with wavelength -- serves as a probe of dust grain properties in star-forming regions. In this paper, we present 89 $\mu$m - 214 $\mu$m polarization spectrum measurements obtained from SOFIA/HAWC+ for three star-forming clouds -- OMC1, M17, and W3. We find that all three clouds have an overall decreasing polarization percentage with increasing wavelength (i.e., a ``falling polarization spectrum''). We use SOFIA and Herschel data to create column density and temperature maps for each cloud. We fit for the slope of the polarization spectrum at each sky position in each cloud, and using the Pearson $r$ coefficient we probe each cloud for possible correlations of slope with column density and slope with temperature. We also create plots of slope vs. column density and slope vs. temperature for each cloud. For the case of OMC1, our results are consistent with those presented by J. Michail et al., who carried out a similar analysis for that cloud. Our plots of polarization spectrum slope vs. column density reveal that for each cloud there exists a critical column density below which a falling polarization spectrum is not observed. For these more diffuse sightlines, the polarization spectrum is instead flat or slightly rising. This finding is consistent with a hypothesis presented 25 years ago in a paper led by R. Hildebrand based on Kuiper Airborne Observatory data. This hypothesis is that regions shielded from near-IR radiation are required to produce a sharply falling polarization spectrum.

astro-ph.SR↗

arXiv2025

The End of the Road for Far-infrared Reddening Maps? Evidence for Reddening Errors Driven by Changes in PAH Abundance

Accurate correction for extinction by Galactic dust is essential for studying the extragalactic sky. In the low-extinction regions of the Ursa Major molecular cloud complex, we demonstrate that Galactic dust reddening maps constructed from observations of far-infrared emission are insensitive to variations in the abundance of polycyclic aromatic hydrocarbons (PAHs), and, as a result, to PAH-induced variations in reddening. Using galaxy counts to validate various reddening maps, we find evidence that maps based on far-infrared emission erroneously under-predict reddening compared to stellar reddening maps. This underestimation by far-infrared emission based reddening maps -- representing the largest discrepancy between maps of up to $E(B-V)=0.08$ mag -- is correlated with the relative brightness of PAH emission. Furthermore, we demonstrate theoretically that changes in PAH abundance via accretion from the gas phase is capable of altering extinction significantly with only minor changes to far-infrared emission. We show that modeling the extinction of Ursa Major using both far-infrared and mid-infrared emission more accurately traces dust extinction variations due to changes in PAH abundance. Finally, we discuss how SPHEREx observations of the 3.3 $\mu$m PAH feature are a promising way to overcome this limitation of far-infrared emission.

astro-ph.GA↗

arXiv2024

Magnetic Field Alignment Relative to Multiple Tracers in the High-mass Star-forming Region RCW 36

We use polarization data from SOFIA HAWC+ to investigate the interplay between magnetic fields and stellar feedback in altering gas dynamics within the high-mass star-forming region RCW 36, located in Vela C. This region is of particular interest as it has a bipolar HII region powered by a massive star cluster which may be impacting the surrounding magnetic field. To determine if this is the case, we apply the Histogram of Relative Orientations (HRO) method to quantify the relative alignment between the inferred magnetic field and elongated structures observed in several datasets such as dust emission, column density, temperature, and spectral line intensity maps. The HRO results indicate a bimodal alignment trend, where structures observed with dense gas tracers show a statistically significant preference for perpendicular alignment relative to the magnetic field, while structures probed by photo-dissociation region (PDR) tracers tend to align preferentially parallel relative to the magnetic field. Moreover, the dense gas and PDR associated structures are found to be kinematically distinct such that a bimodal alignment trend is also observed as a function of line-of-sight velocity. This suggests that the magnetic field may have been dynamically important and set a preferred direction of gas flow at the time that RCW 36 formed, resulting in a dense ridge developing perpendicular to the magnetic field. However on filament-scales near the PDR region, feedback may be energetically dominating the magnetic field, warping its geometry and the associated flux-frozen gas structures, causing the observed the preference for parallel relative alignment.

astro-ph.GA↗

arXiv2024

Modeling the Far-Infrared Polarization Spectrum of a High-Mass Star Forming Cloud

The polarization spectrum, or wavelength dependence of the polarization fraction, of interstellar dust emission provides important insights into the grain alignment mechanism of interstellar dust grains. We investigate the far-infrared polarization spectrum of a realistic simulated high-mass star forming cloud under various models of grain alignment and emission. We find that neither a homogeneous grain alignment model nor a grain alignment model that includes collisional dealignment is able to produce the falling spectrum seen in observations. On the other hand, we find that a grain alignment model with grain alignment efficiency dependent on local temperature is capable of producing a falling spectrum that is in qualitative agreement with observations of OMC-1. For the model most in agreement with OMC-1, we find no correlation between temperature and the slope of the polarization spectrum. However, we do find a positive correlation between column density and the slope of the polarization spectrum. We suggest this latter correlation to be the result of wavelength-dependent polarization by absorption.

astro-ph.GA↗

arXiv2024

Magnetic Fields Observed along the E-W Outflow of IRAS 16293-2422

Magnetic fields likely play an important role in the formation of young protostars. Multiscale and multiwavelength dust polarization observations can reveal the inferred magnetic field from scales of the cloud to core to protostar. We present continuum polarization observations of the young protostellar triple system IRAS 16293-2422 at 89 $μ$m using HAWC+ on SOFIA. The inferred magnetic field is very uniform with an average field angle of 89$^\circ\pm$23$^\circ$ (E of N), which is different from the $\sim$170$^\circ$ field morphology seen at 850 $μ$m at larger scales (> 2000 au) with JCMT POL-2 and at 1.3 mm on smaller scales (< 300 au) with ALMA. The HAWC+ magnetic field direction is aligned with the known E-W outflow. This alignment difference suggests that the shorter wavelength HAWC+ data is tracing the magnetic field associated with warmer dust likely from the outflow cavity, whereas the longer wavelength data are tracing the bulk magnetic field from cooler dust. Also, we show in this source the dust emission peak is strongly affected by the observing wavelength. The dust continuum peaks closer to source B (northern source) at shorter wavelengths and progressively moves toward the southern A source with increasing wavelength (from 22 $μ$m to 850 $μ$m).

astro-ph.SR↗

arXiv2022

The Twisted Magnetic Field of the Protobinary L483

We present H-band (1.65 $μ$m) and SOFIA HAWC+ 154 $μ$m polarization observations of the low-mass core L483. Our H-band observations reveal a magnetic field that is overwhelmingly in the E-W direction, which is approximately parallel to the bipolar outflow that is observed in scattered IR light and in single-dish $^{12}$CO observations. From our 154 $μ$m data, we infer a $\sim$ 45$^{\circ}$ twist in the magnetic field within the inner 5" (1000 au) of L483. We compare these new observations with published single-dish 350 $μ$m polarimetry and find that the 10,000 au scale H-band data match the smaller scale 350 $μ$m data, indicating that the collapse of L483 is magnetically regulated on these larger scales. We also present high-resolution 1.3 mm ALMA data of L483 which reveals it is a close binary star with a separation of 34 au. The plane of the binary of L483 is observed to be approximately parallel to the twisted field in the inner 1000 au. Comparing this result to the $\sim$ 1000 au protostellar envelope, we find that the envelope is roughly perpendicular to the 1000 au HAWC+ field. Using the data presented, we speculate that L483 initially formed as a wide binary and the companion star migrated to its current position, causing an extreme shift in angular momentum thereby producing the twisted magnetic field morphology observed. More observations are needed to further test this scenario.

astro-ph.SR↗

arXiv2022

ProMP: Proximal Meta-Policy Search

Credit assignment in Meta-reinforcement learning (Meta-RL) is still poorly understood. Existing methods either neglect credit assignment to pre-adaptation behavior or implement it naively. This leads to poor sample-efficiency during meta-training as well as ineffective task identification strategies. This paper provides a theoretical analysis of credit assignment in gradient-based Meta-RL. Building on the gained insights we develop a novel meta-learning algorithm that overcomes both the issue of poor credit assignment and previous difficulties in estimating meta-policy gradients. By controlling the statistical distance of both pre-adaptation and adapted policies during meta-policy search, the proposed algorithm endows efficient and stable meta-learning. Our approach leads to superior pre-adaptation policy behavior and consistently outperforms previous Meta-RL algorithms in sample-efficiency, wall-clock time, and asymptotic performance.

cs.LG↗

arXiv2021

Joint Attention for Multi-Agent Coordination and Social Learning

Joint attention - the ability to purposefully coordinate attention with another agent, and mutually attend to the same thing -- is a critical component of human social cognition. In this paper, we ask whether joint attention can be useful as a mechanism for improving multi-agent coordination and social learning. We first develop deep reinforcement learning (RL) agents with a recurrent visual attention architecture. We then train agents to minimize the difference between the attention weights that they apply to the environment at each timestep, and the attention of other agents. Our results show that this joint attention incentive improves agents' ability to solve difficult coordination tasks, by reducing the exponential cost of exploring the joint multi-agent action space. Joint attention leads to higher performance than a competitive centralized critic baseline across multiple environments. Further, we show that joint attention enhances agents' ability to learn from experts present in their environment, even when completing hard exploration tasks that do not require coordination. Taken together, these findings suggest that joint attention may be a useful inductive bias for multi-agent learning.

cs.AI↗

arXiv2021

HAWC+/SOFIA Polarimetry in L1688: Relative Orientation of Magnetic Field and Elongated Cloud Structure

We present a study of the relative orientation between the magnetic field and elongated cloud structures for the $ρ$ Oph A and $ρ$ Oph E regions in L1688 in the Ophiuchus molecular cloud. Combining inferred magnetic field orientation from HAWC+ 154 $μ$m observations of polarized thermal emission with column density maps created using Herschel submillimeter observations, we find consistent perpendicular relative alignment at scales of $0.02$ pc ($33.6"$ at $d \approx 137$ pc) using the histogram of relative orientations (HRO) technique. This supports the conclusions of previous work using Planck polarimetry and extends the results to higher column densities. Combining this HAWC+ HRO analysis with a new Planck HRO analysis of L1688, the transition from parallel to perpendicular alignment in L1688 is observed to occur at a molecular hydrogen column density of approximately $10^{21.7}$ cm$^{-2}$. This value for the alignment transition column density agrees well with values found for nearby clouds via previous studies using only Planck observations. Using existing turbulent, magnetohydrodynamic simulations of molecular clouds formed by colliding flows as a model for L1688, we conclude that the molecular hydrogen volume density associated with this transition is approximately $\sim10^{4}$ cm$^{-3}$. We discuss the limitations of our analysis, including incomplete sampling of the dense regions in L1688 by HAWC+.

astro-ph.SR↗

arXiv2020

Model-based Reinforcement Learning for Decentralized Multiagent Rendezvous

Collaboration requires agents to align their goals on the fly. Underlying the human ability to align goals with other agents is their ability to predict the intentions of others and actively update their own plans. We propose hierarchical predictive planning (HPP), a model-based reinforcement learning method for decentralized multiagent rendezvous. Starting with pretrained, single-agent point to point navigation policies and using noisy, high-dimensional sensor inputs like lidar, we first learn via self-supervision motion predictions of all agents on the team. Next, HPP uses the prediction models to propose and evaluate navigation subgoals for completing the rendezvous task without explicit communication among agents. We evaluate HPP in a suite of unseen environments, with increasing complexity and numbers of obstacles. We show that HPP outperforms alternative reinforcement learning, path planning, and heuristic-based baselines on challenging, unseen environments. Experiments in the real world demonstrate successful transfer of the prediction models from sim to real world without any additional fine-tuning. Altogether, HPP removes the need for a centralized operator in multiagent systems by combining model-based RL and inference methods, enabling agents to dynamically align plans.

cs.MA↗

arXiv2020

Mathematical Reasoning via Self-supervised Skip-tree Training

We examine whether self-supervised language modeling applied to mathematical formulas enables logical reasoning. We suggest several logical reasoning tasks that can be used to evaluate language models trained on formal mathematical statements, such as type inference, suggesting missing assumptions and completing equalities. To train language models for formal mathematics, we propose a novel skip-tree task. We find that models trained on the skip-tree task show surprisingly strong mathematical reasoning abilities, and outperform models trained on standard skip-sequence tasks. We also analyze the models' ability to formulate new conjectures by measuring how often the predictions are provable and useful in other proofs.

cs.LG↗

arXiv2019

Mathematical Reasoning in Latent Space

We design and conduct a simple experiment to study whether neural networks can perform several steps of approximate reasoning in a fixed dimensional latent space. The set of rewrites (i.e. transformations) that can be successfully performed on a statement represents essential semantic features of the statement. We can compress this information by embedding the formula in a vector space, such that the vector associated with a statement can be used to predict whether a statement can be rewritten by other theorems. Predicting the embedding of a formula generated by some rewrite rule is naturally viewed as approximate reasoning in the latent space. In order to measure the effectiveness of this reasoning, we perform approximate deduction sequences in the latent space and use the resulting embedding to inform the semantic features of the corresponding formal statement (which is obtained by performing the corresponding rewrite sequence using real formulas). Our experiments show that graph neural networks can make non-trivial predictions about the rewrite-success of statements, even when they propagate predicted latent representations for several steps. Since our corpus of mathematical formulas includes a wide variety of mathematical disciplines, this experiment is a strong indicator for the feasibility of deduction in latent space in general.

cs.LG↗

arXiv2018

Modular Architecture for StarCraft II with Deep Reinforcement Learning

We present a novel modular architecture for StarCraft II AI. The architecture splits responsibilities between multiple modules that each control one aspect of the game, such as build-order selection or tactics. A centralized scheduler reviews macros suggested by all modules and decides their order of execution. An updater keeps track of environment changes and instantiates macros into series of executable actions. Modules in this framework can be optimized independently or jointly via human design, planning, or reinforcement learning. We apply deep reinforcement learning techniques to training two out of six modules of a modular agent with self-play, achieving 94% or 87% win rates against the "Harder" (level 5) built-in Blizzard bot in Zerg vs. Zerg matches, with or without fog-of-war.

cs.AI↗

arXiv2018

Deep Imitation Learning for Complex Manipulation Tasks from Virtual Reality Teleoperation

Imitation learning is a powerful paradigm for robot skill acquisition. However, obtaining demonstrations suitable for learning a policy that maps from raw pixels to actions can be challenging. In this paper we describe how consumer-grade Virtual Reality headsets and hand tracking hardware can be used to naturally teleoperate robots to perform complex tasks. We also describe how imitation learning can learn deep neural network policies (mapping from pixels to actions) that can acquire the demonstrated skills. Our experiments showcase the effectiveness of our approach for learning visuomotor skills.

cs.LG↗