arXiv subjects

Yuan-Sen Ting

Publications and source records attributed to Yuan-Sen Ting.

At least 19 recordsLinked to original sources

arXiv2026

More than half of recent astronomy papers are written with language-model assistance

Language models leave a distinctive vocabulary in the prose they help write, and we measure how much of the astronomy literature now carries it. From the full text of 207,111 astro-ph papers spanning 2015 to mid-2026, we count those words in each paper and model the counts, in proportion to paper length, as a mixture of assisted and unassisted writing in a hierarchical Bayesian model. Papers from before 2020 calibrate the unassisted rate, and the 392 papers that disclose model use calibrate the assisted one. Our answer depends on how often these words would appear today if nobody used a model, a rate that must be modeled rather than observed, so we extend it past 2020 under three assumptions and report all three. For 2025 that gives $54^{+8}_{-8}\,(\mathrm{stat},\,95\%)\,^{+26}_{-0}\,(\mathrm{sys,\ background})$% of papers, the second error being the spread across the three. The estimate stays at or above 36% when we vary that choice, the calibration, and the requirement that adoption only rises. A word list built from the astro-ph corpus, keeping only words that rose across every subfield, leaves 2025 in the same range. Assisted writing is also getting harder to see, since authors adapt to the words that reveal it and the marker excess more than halves between 2023 and 2026. Our model allows for that fading, so it can separate a fainter trace from reduced use. More than half of recent astro-ph papers therefore carry a language-model trace, while only 0.81% of 2025 papers disclose it, one declaration for every $\sim$66 papers with a trace.

astro-ph.IM↗

arXiv2026

Agentic Active Learning Meets Visual Embeddings: Finding Anomalies among 370 000 Variable Stars from ASAS-SN

Unusual light-curve morphologies can point to rare physical configurations or new phenomena, but automatic searches for anomalies are often dominated by artifacts. Separating genuine anomalies from false positives has traditionally required manual vetting, which does not scale to modern surveys. We present an active learning framework for detecting anomalies in samples of periodic variable stars, with the vetting delegated to multimodal large language model agents. The initial ranking comes from isolation forests trained on DINOv2 ViT-g/14 embeddings of phase-folded light curves. The agents iteratively review the light-curve images of the top-ranked candidates and assign relevance scores, which are propagated through the embedding space to prioritize the next targets. A logistic regression step then extends the search beyond label propagation, and a multi-agent consensus review filters false positives. Using Gemini~3 Flash agents, we applied the pipeline to $373\,646$ periodic variables from ASAS-SN Sky Patrol V2.0. Across $51$ iterations, the agents labeled ${\sim}1\%$ of the sample in roughly $3$ hours at a cost of ${\approx}$ $$40$, yielding our final catalog of $24$ anomalies ($18$ newly reported) and $153$ potentially interesting objects. The anomalies include eclipsing binaries with extremely deep primary eclipses, high-amplitude contact binaries and pulsators, and a symbiotic nova in outburst for over two decades. At the same labeling budget, the initial ranking would have recovered only $30\%$ of our catalog, while reaching its lowest-ranked anomaly would have required ${\sim}70$ times the budget. These results demonstrate the feasibility of agentic active learning for anomaly detection in existing and upcoming photometric surveys.

astro-ph.SR↗

arXiv2026

The Shape of the Vertical Action Distribution Locates the Scatterers that Heat the Galactic Disc

The Milky Way's stellar disc is thicker than the cold gas layer from which its stars form. What scatters stars onto orbits with greater vertical motion remains unresolved. Scatterers could fill the disc volume, as bending waves or dark substructure would, or could be confined to the midplane, as giant molecular clouds are. Scattering from a midplane layer occurs only during the fast plane-crossing phase and becomes less effective as that speed increases, whereas a volume-filling perturbation remains effective near the slow turning points. We derive the distribution of vertical action this leaves behind, for any height distribution of scatterers. The geometry turns out to enter only through the logarithmic slope of the diffusivity in action, $D\propto J_z^{b}$, and solving the Fokker-Planck equation gives $p(J_z)\propto\exp[-(J_z/J_0)^{2-b}]$. Scatterers that fill the volume give $b=1$ and an exponential, a thin layer at the midplane gives $b=1/2$ and a sharper cutoff: the shape of $p(J_z)$ records where the scatterers sit, the growth of its scale how strongly they scatter. We fit this model to 7589 low-$\alpha$ red clump stars of Ting & Rix (2019) between 5 and 10 kpc and 2 and 8 Gyr old, leaving the heating history free. This yields $b=0.51^{+0.06}_{-0.07}$, consistent with the thin-layer prediction but not the volume-filling one. Comparing the 2-4 Gyr heating amplitude with the present molecular surface density gives an effective scatterer mass of $2.7\times10^{6}\,M_\odot$. Older stars have experienced more of the Galaxy's gas-richer past; correcting for that history brings all four age bins to $(1.9-2.8)\times10^{6}\,M_\odot$, inside the range cloud catalogues and mass functions give. The Milky Way's disc is heated near the plane, by an evolving population of objects of giant-molecular-cloud mass.

astro-ph.GA↗

arXiv2026

Spectroscopic Binary Detection as Agent-Callable Tools: Detecting 40,000+ Main-Sequence Binary Candidates from SDSS DR19 APOGEE Spectra

Unresolved binaries are common in spectroscopic surveys, and their blended light biases the parameters inferred for them. Methods to detect them exist, but applying one to a new data release is limited mostly by operating know-how that is rarely written down. We package the double-lined spectroscopic binary decomposition of El-Badry et al. (2018b) for reuse on APOGEE spectra, with the executable operations as nine tool servers built on the Model Context Protocol (MCP) and the operating decisions as a Skill. Run over the 238,205 DR19 dwarfs under a fixed driver script, the classifier flags 41,466 SB2 candidates with median mass ratio q = 0.91, about fifteen times the 2,645 identified in DR13, though not at matched purity. The 8.1% control false-positive rate implies that close to 40% are single stars, so we release the sample as a candidate list. Refitting the individual visits confirms 68.5% of the multiply-visited SB2 and adds 519 single-lined velocity variables and 8,981 orbit-ready systems. We compare the eccentricities of the best-sampled binaries and find no significant difference between the close twins (q > 0.95) and matched non-twins, which excludes an eccentricity excess of the kind measured at wide separations. An ablation of two operating decisions illustrates that a fresh agent recovers a removed decision only when its absence leaves a measurable trace in the fit. We release the tools, the Skill, and the DR19 catalog with its multi-epoch supplement.

astro-ph.SR↗

arXiv2026

Foundation Models for Astrophysics

Foundation models are high-capacity networks pretrained once on broad data and then reused across many tasks. This chapter introduces them through the idea of a transferable representation, the internal description a network forms during training, which, rather than the fitted task, is what carries over to new problems. We develop the idea from first principles for an astronomical reader, starting from why a representation matters and what makes one useful, and then surveying the architectures, self-supervised objectives, scaling, adaptation, and cross-modal learning that produce one. A theme throughout is the distinction between these methods and the goal they serve. The presence of a transformer, a self-supervised objective, and large-scale pretraining does not by itself make a model a foundation model, since the defining property is that the learned representation transfers, as tested by its ability to work on new tasks with little or no task-specific training data (few-shot and zero-shot learning). We then consider astronomy, where data are abundant but labels are scarce and simulations often stand in for ground truth. Here we offer a cautious reading of the current literature, in which many models adopt the architecture of foundation models while clear demonstrations of transfer across instruments, populations, and tasks remain comparatively rare. This is to be expected, since robust transfer beyond language is still uncommon even in vision and the wider physical sciences, and whether further scaling or a different account of representation will close the gap remains an open question. We close by placing the goal within the broader aim of machine intelligence and outlining the evidence that would mark real progress.

astro-ph.IM↗

arXiv2026

The Payne Zero Project I: Stellar Spectra from Physical Models in Seconds

Modern stellar surveys measure millions of spectra, yet one self-consistent atmosphere and spectrum can require tens of minutes. This cost has motivated grids, spectral emulators, and data-driven models. We present Payne Zero, which reorganizes one-dimensional LTE Kurucz calculations for GPU-native synthesis and multicore atmosphere iteration, and validate it against the original Fortran programs. A 300--1000 nm solar spectrum sampled at $R_{\rm grid}=300{,}000$ takes about 14 s on an NVIDIA H100 GPU, while the APOGEE 1500--1700 nm interval takes about 1 s. Physical atmosphere iterations take 2--5 s on 16 AMD CPU threads, and learned initializers reduce the iterations required for convergence. Final spectra remain in practical parity across the tested dwarf and giant regimes. These speeds place direct synthesis inside an optimizer without a label-to-flux spectral emulator. We demonstrate direct many-element fitting of reduced APOGEE spectra and recover multi-element abundance trends broadly consistent with the survey catalog. GPU-resident velocity shifts, broadening, line-spread-function convolution, and detector sampling add negligible cost relative to synthesis. The direct-synthesis search takes less than one minute per star on an H100, while atmosphere verification runs independently on multicore CPUs. The same computational graph calibrates more than $10^5$ oscillator-strength and damping corrections jointly against the Sun and Arcturus in about one minute on an H100. Payne Zero therefore brings direct physical fitting and atomic-data calibration to survey scale. The code is available at https://github.com/tingyuansen/payne-zero.

astro-ph.SR↗

arXiv2026

LMC-induced Perturbations in the Milky Way Halo II: Bridging Field-level Inference and Summary-level Simulation-Based Inference

The gravitational interaction between the Milky Way (MW) and the Large Magellanic Cloud (LMC) drives the outer halo into dynamical disequilibrium, imprinting the masses and structural parameters of both galaxies onto the 6D phase-space distribution of halo tracers. This signal has been characterised with summary statistics ranging from low-order velocity moments to basis function expansions, yet how much information these summaries discard, and whether they are complementary, remains unclear. We address these questions by comparing a field-level likelihood benchmark with physically interpretable summaries for constraining $(M_{\mathrm{MW}}, M_{\mathrm{LMC}}, c, q)$, where $c$ and $q$ are the MW halo concentration and flattening. A Conditional Flow Matching (CFM) model trained on the HaloDance $N$-body suite provides an exact likelihood at a held-out fiducial point; for 5,000 tracers in $30$--$120$~kpc it tightens marginal constraints by factors of $2.5$--$9.9$ over an all-sky velocity-moment forecast. We then expand the halo density and velocity fields in a multipole basis-function expansion (BFE) and compress the coefficients with the Massive Optimised Parameter Estimation and Data compression (MOPED) algorithm into four parameter-sensitive summaries that preserve their Fisher information. A variational mutual-information analysis shows that the BFE+MOPED summaries and the velocity moments are complementary, so we combine them into a joint $19$-dimensional vector as our primary inference pipeline: it tightens the marginal constraints by up to $15$ per cent over BFE+MOPED alone and by $30$--$71$ per cent over velocity moments alone, reaching within a factor of $1.3$--$2.9$ of the field-level benchmark. We thus establish a physically interpretable summary-level route to MW--LMC inference alongside the field-level benchmark that bounds its information content.

astro-ph.GA↗

arXiv2026

Deep Learning for Astrophysics: An Open Textbook from the NASA Cosmic Origins AI/ML Science and Technology Interest Group

Recent community assessments identify education as a principal barrier to adopting modern machine learning in astronomy. We present Deep Learning for Astrophysics, a freely available textbook at https://deeplearning4astro.com, curated from the NASA Cosmic Origins Artificial Intelligence and Machine Learning Science and Technology Interest Group (AI/ML STIG) lecture series. The book collects 23 chapters by 17 lecturers across six parts, moving from computational foundations and deep-learning architectures through generative modeling, simulation-based inference, reinforcement learning, and large-language-model agents to the practice of AI-laden science. Many include executable notebooks using astronomical data.

astro-ph.IM↗

arXiv2026

HRMOS: A High-Resolution Multi-Object Spectrograph for the VLT

This White Paper presents the scientific rationale and instrument concept for HRMOS (High-Resolution Multi-Object Spectrograph), a next-generation instrument proposed for the ESO Very Large Telescope within the VLT 2030 roadmap. Current and planned facilities offer either multi-object spectroscopy or ultra-high spectral resolution, but not both. HRMOS fills this gap by combining very high spectral resolution, multi-object capability, and radial-velocity stability, enabling transformative studies in Galactic and extragalactic astrophysics. The baseline design provides a resolving power of R = 80000, radial-velocity precision of 10 m s-1 (goal: 5 m s-1), simultaneous observations of 50-60 targets, and broad optical coverage down to 385 nm. These capabilities enable precise measurements of elemental abundances, isotopic ratios, line profiles, and radial velocities for large stellar samples, including crowded fields, star clusters, the Galactic bulge, and nearby dwarf galaxies. HRMOS will address key questions on the age of the oldest stellar populations through nucleocosmochronology, the formation and survival of planetary systems, the assembly history of the Milky Way and satellites, the origin of the heaviest elements, stellar evolution, and the chemical and dynamical properties of the interstellar and circumgalactic medium. It will bridge large spectroscopic surveys and the next generation of extremely large telescopes, with strong synergies with 4MOST, Gaia, TESS, PLATO, the proposed Haydn mission, and future ELT instruments. Building on VLT/FLAMES heritage, HRMOS represents a strategic investment for European astronomy in the 2030s.

astro-ph.IM↗

arXiv2026

A Bayesian Search for Planet Engulfment Signatures in Solar Analogs

We present a systematic Bayesian search for chemical fingerprints of planet engulfment in 113 solar twins and analogs with high-precision abundance measurements, 45 of which host known or candidate planets or brown-dwarf companions. We constructed a Bayesian framework with three sets of abundance models: random scatter, Galactic chemical evolution, and planet engulfment with bulk Earth or CM chondrite compositions. Through model comparisons, we identified three candidates whose abundance patterns strongly favor planet engulfment over the alternatives, with inferred engulfed masses of about 7.5-33 Earth masses. Our findings correspond to a nominal detection rate of 1-3% for planet-engulfment signatures among solar analogs. This work extends abundance-based engulfment searches beyond the binary-star context and provides a framework for probing star-planet co-evolution with solar analogs, which goes beyond the commonly used abundance-condensation-temperature correlation (Tc slope).

astro-ph.EP↗

arXiv2026

The Milky Way - Large Magellanic Cloud Interaction with Simulation Based Inference

The infall of the Large Magellanic Cloud (LMC) into the Milky Way (MW) has displaced the MW's centre of mass, manifesting as an observed reflex motion in the velocities of outer halo stars. We use a Simulation Based Inference framework to constrain properties of the MW, LMC and the induced reflex motion using the dynamics of outer MW halo stars. Specifically, we use the mean radial and tangential velocities of outer halo stars calculated in a set of distance and on-sky bins. We train neural networks to estimate parameter posterior distributions using a set of $128,000$ rigid MW--LMC simulations conditioned upon velocity data from the Dark Energy Spectroscopic Instrument (DESI) and the combined H3+SEGUE+MagE outer halo surveys. We constrain the reflex motion velocity and the enclosed LMC mass within $50 \, \rm kpc$ using the DESI or H3+SEGUE+MagE dataset while varying the survey sky coverage and depth. Using the radial and tangential velocity data from the H3+SEGUE+MagE survey and on-sky quadrants, we report a distance-averaged reflex motion velocity for the outer halo samples, the speed at which the MW lurches towards the LMC, of $v_{\rm{travel}} = 26.4^{+5.5}_{-4.4} \, \rm km \, \rm s^{-1}$, while simultaneously finding an enclosed LMC mass of $M_{\rm LMC}(< 50 \, \rm kpc) = 9.2^{+1.9}_{-2.3} \times 10^{10}\, \rm M_{\odot}$. Quoted uncertainties are statistical. Our results suggest that the LMC's total mass is at least $\approx 10-15 \%$ of that of the MW. This inference framework is flexible such that it can provide rapid constraints when applied to any future survey measuring the velocities of outer halo stars.

astro-ph.GA↗

arXiv2026

Solving Inverse Problems of Chaotic Systems with Bidirectional Conditional Flow Matching

Modeling chaotic systems is crucial yet challenging. Inverse problems in chaotic dynamics, namely inferring initial conditions from final states, remain largely unsolved because of ill-posedness, non-uniqueness, instability, and potentially chaotic time-reverse dynamics. We address this open problem with Bidirectional Conditional Flow Matching (Bi-CFM), which learns bidirectional mappings between distributions of initial and final states to capture the stochasticity of chaotic evolution and mitigate exponential error accumulation over time. Furthermore, for systems with conservation laws, we extend it to Conservation-constrained Bi-CFM (CBi-CFM). Across the classic Lorenz, Circuit, and high-dimensional Lorenz 96 systems, Bi-CFM improves five distribution-level metrics over baselines while achieving a speedup of more than two orders of magnitude. In the three-body planet-planet scattering problem in planetary dynamics, CBi-CFM better respects conservation laws, with conservation errors comparable to those of the ground truth. Finally, on real observations of globular clusters, collisional million-body systems shaped by $\sim 10^{10}$ years (10 Gyr) of evolution, our method represents an advance in accuracy, establishing a scalable route to solving inverse problems of long-timescale real-world chaotic dynamics.

cs.AI↗

arXiv2026

Weak-CN Stars Are Ordinary Cool Red Supergiants

Weak CN absorption near ~8000 A has recently been detected in evolved red supergiants (RSGs) of 5-10 $M_\odot$ across three Local Group galaxies. These weak-CN RSGs sit in a narrow molecular regime: cool enough for CN to be visible in a non-carbon, C/O<1 atmosphere, but warm enough that TiO is not saturated and changes in $T_{\rm eff}$ and in the surface C+N reservoir move CN and TiO in distinct directions. We test this picture with pseudo-continuum equivalent widths (EWs) measured from LMC, M33, and M31 weak-CN and carbon-star coadds, compared at matched resolution to a self-consistent grid of synthetic RSG atmospheres spanning $T_{\rm eff}$, $[α/{\rm Fe}]$, and surface C and N offsets relative to each host's scaled-solar baseline. Ordinary cool-RSG models reproduce the weak-CN coadds across all three hosts, with per-feature residuals at the level of the adopted EW systematic floors. The robust observable is the combined surface abundance $Δ$[C/H]+$Δ$[N/H] rather than each offset individually, because CN forms from the product of available C and N number densities. Mapping $Δ$[C/H]+$Δ$[N/H] to initial rotation through PARSEC v2.0 has modest leverage -- the variable shifts by ~0.07 dex from $ω_i$=0 to $ω_i$=0.6 -- and within this resolution slow-rotation first dredge-up is consistent with LMC and M33, and with M31 once a single-feature CaT 8542 A calibration anchor is allowed. The straightforward resolution of the discovery puzzle is therefore that weak CN is not an exotic carbon-star intermediate but the expected molecular-equilibrium signature of ordinary cool RSGs.

astro-ph.SR↗

arXiv2026

pyKurucz: A Pure Python Reimplementation of Kurucz ATLAS12 and SYNTHE for Stellar Spectrum Synthesis

pyKurucz is a pure Python reimplementation of Kurucz's ATLAS12 and SYNTHE, the standard code tools for computing synthetic stellar spectra. The original Fortran codes, written decades ago in a legacy dialect, are difficult to compile with modern toolchains without significant manual patching, and their long-term maintenance is uncertain following the passing of Robert L. Kurucz in 2025. pyKurucz delivers a complete, line-by-line Python translation of both engines, entirely removing the need for Fortran. Powered by NumPy, SciPy, and Numba, it combines the full ATLAS12 iteration loop (with direct opacity sampling and convection) with comprehensive physical treatments, from Saha-Boltzmann populations and molecular equilibrium to radiative transfer and advanced line broadening. Validated against the original across 100 atmosphere models spanning 2500 K cool giants to 44,000 K O stars over 300-1800 nm at resolving power R = 300,000, it achieves sub-0.01% median agreement. The pure Python implementation enables direct integration with machine learning workflows and large-scale survey pipelines, while preserving an archival reference implementation of both ATLAS12 and SYNTHE in a modern, readable language.

astro-ph.SR↗

arXiv2026

Deep Learning in Astrophysics

Deep learning has generated diverse perspectives in astronomy, with ongoing discussions between proponents and skeptics motivating this review. We examine how neural networks complement classical statistics, extending our data analytical toolkit for modern surveys. Astronomy offers unique opportunities through encoding physical symmetries, conservation laws, and differential equations directly into architectures, creating models that generalize beyond training data. Yet challenges persist as unlabeled observations number in billions while confirmed examples with known properties remain scarce and expensive. This review demonstrates how deep learning incorporates domain knowledge through architectural design, with built-in assumptions guiding models toward physically meaningful solutions. We evaluate where these methods offer genuine advances versus claims requiring careful scrutiny. - Neural architectures overcome bias-variance trade-offs among scalability, expressivity, and data efficiency by encoding physical symmetries and conservation laws into network structure, enabling learning from limited labeled data. - Simulation-based inference and anomaly detection extract information from complex, non-Gaussian distributions where analytical likelihoods fail, enabling field-level cosmological analysis and systematic discovery of rare phenomena. - Multiscale neural modeling bridges resolution gaps in astronomical simulations, learning effective subgrid physics from expensive high-fidelity runs to enhance large-volume calculations where direct computation remains prohibitive. - Emerging paradigms-reinforcement learning for telescope operations, foundation models learning from minimal examples, and large language model agents for research automation-show promise though are still developing in astronomical applications.

astro-ph.IM↗

arXiv2026

Egent: An Autonomous Agent for Equivalent Width Measurement

We present Egent, an autonomous agent that combines classical multi-Voigt profile fitting with large language model (LLM) visual inspection and iterative refinement. The fitting engine is built from scratch with minimal dependencies, creating an ecosystem where the LLM can reason about fits through function calls--adjusting wavelength windows, adding blend components, modifying continuum treatment, and flagging problematic cases. Egent operates directly on raw flux spectra without requiring pre-normalized continua. We validate against manual measurements from human experts using 18,615 lines from the C3PO program across 84 Magellan/MIKE spectra at SNR~50-250. The raw agreement between Egent and expert measurements is MAD=5-7mA, without any post-hoc per-spectrum correction. Per-spectrum slopes of ~0.85-1.19 around unity reflect differences in global continuum methodology rather than fitting failures. The LLM's primary role is quality control: it confirms good fits (~60-65% of lines are LLM-refined and accepted), flags problematic cases (~10-20%), and occasionally rescues edge cases where tool use improves fits. Agreement between GPT-5 and GPT-5-mini confirms reproducibility, with GPT-5-mini enabling low-cost analysis at ~200 lines per US dollar. Every fit stores complete Voigt parameters, continuum coefficients, and LLM reasoning chains, enabling exact reconstruction without re-running. Egent compresses what traditionally requires months of expert effort into days of automated analysis, enabling survey-scale EW measurement. We provide open-source code at https://github.com/tingyuansen/Egent, including a web interface for drag-and-drop analysis and a local LLM backend for fully offline operation on consumer hardware.

astro-ph.IM↗

arXiv2026

What Understanding Means in AI-Laden Astronomy

Artificial intelligence is rapidly transforming astronomical research, yet the scientific community has largely treated this transformation as an engineering challenge rather than an epistemological one. This perspective article argues that philosophy of science offers essential tools for navigating AI's integration into astronomy--conceptual clarity about what "understanding" means, critical examination of assumptions about data and discovery, and frameworks for evaluating AI's roles across different research contexts. Drawing on an interdisciplinary workshop convening astronomers, philosophers, and computer scientists, we identify several tensions. First, the narrative that AI will "derive fundamental physics" from data misconstrues contemporary astronomy as equation-derivation rather than the observation-driven enterprise it is. Second, scientific understanding involves more than prediction--it requires narrative construction, contextual judgment, and communicative achievement that current AI architectures struggle to provide. Third, because narrative and judgment matter, human peer review remains essential--yet AI-generated content flooding the literature threatens our capacity to identify genuine insight. Fourth, while AI excels at well-defined problem-solving, the ill-defined problem-finding that drives breakthroughs appears to require capacities beyond pattern recognition. Fifth, as AI accelerates what is feasible, pursuitworthiness criteria risk shifting toward what AI makes easy rather than what is genuinely important. We propose "pragmatic understanding" as a framework for integration--recognizing AI as a tool that extends human cognition while requiring new norms for validation and epistemic evaluation. Engaging with these questions now may help the community shape the transformation rather than merely react to it.

astro-ph.IM↗

arXiv2026

Predicting New Concept-Object Associations in Astronomy by Mining the Literature

We construct a concept-object knowledge graph from the full astro-ph corpus through July 2025. Using an automated pipeline, we extract named astrophysical objects from OCR-processed papers, resolve them to SIMBAD identifiers, and link them to scientific concepts annotated in the source corpus. We then test whether historical graph structure can forecast new concept-object associations before they appear in print. Because the concepts are derived from clustering and therefore overlap semantically, we apply an inference-time concept-similarity smoothing step uniformly to all methods. Across four temporal cutoffs on a physically meaningful subset of concepts, an implicit-feedback matrix factorization model (alternating least squares, ALS) with smoothing outperforms the strongest neighborhood baseline (KNN using text-embedding concept similarity) by 16.8% on NDCG@100 (0.144 vs 0.123) and 19.8% on Recall@100 (0.175 vs 0.146), and exceeds the best recency heuristic by 96% and 88%, respectively. These results indicate that historical literature encodes predictive structure not captured by global heuristics or local neighborhood voting, suggesting a path toward tools that could help triage follow-up targets for scarce telescope time.

astro-ph.IM↗