arXiv subjects

Ke Fang

Publications and source records attributed to Ke Fang.

At least 19 recordsLinked to original sources

arXiv2026

NASA ASTRA Initiative White Paper: Space-Based Mission for Ultrahigh Energy Particles

Ultra-high-energy cosmic rays ($E_{\rm CR} \gtrsim 1$ EeV) are the highest-energy particles known, signaling extreme particle processes at work in the universe. However, many aspects of their nature remain largely unknown, even after more than a century of study. Very-high-energy ($E_\nu \gtrsim 1$ PeV neutrinos associated with cosmic-ray interactions, both during the acceleration process and propagation, would provide new insight into these extreme particles, as we have seen at lower energies with the dawn of TeV neutrino astronomy. Nevertheless, only a handful of such neutrinos have been observed thus far. A space-based observatory dedicated to studying cosmic rays, neutrinos, and photons would provide an unprecedented platform for observations of these extreme-energy messengers.

astro-ph.IM↗

arXiv2026

The Galactic Neutrino Sky: Predictions from Gamma-ray Source Populations

High-energy neutrino emission from the Galactic plane has been detected at a significance of $5.7\sigma$, with a prominent excess toward the inner Galaxy. We show that this excess can be naturally explained by the spatial distribution of Galactic neutrino sources. By combining $\gamma$-ray source catalogs spanning GeV-PeV energies and selecting candidate hadronic emitters, we construct ReGal-$\gamma$, a template of resolved Galactic $\gamma$-ray sources that may also produce high-energy neutrinos. Compared with models of Galactic diffuse emission, ReGal-$\gamma$ predicts a neutrino intensity that is more strongly concentrated toward the inner Galaxy. Combined with models of diffuse cosmic-ray emission and unresolved $\gamma$-ray sources, the template reproduces both the spectral energy distribution and the Galactic longitudinal count profile reported by IceCube without requiring additional renormalization of the emission models. We test this result using an independent $\gamma$-ray source template constructed from different catalogs and find that our conclusion is robust against uncertainties in source modeling. Future measurements of the energy-dependent longitudinal and latitudinal neutrino distributions will provide tighter constraints on these models and help determine the spatial distribution of Galactic neutrino sources.

astro-ph.HE↗

arXiv2026

The Delta Resonance in the Neutrino Sky

Recent measurements of the diffuse cosmic neutrino flux by IceCube show evidence for a spectral break at an energy near $E_\nu \sim 30$ TeV. In this letter, we suggest that this feature may be due to the $\Delta$-baryon resonance in $p\gamma$ interactions. We show that the measured spectrum, including the observed break, can be naturally accommodated by a flux of protons accelerated with a spectrum $dN_p /dE_p \propto E_p^{-3.1}$ interacting with X-rays of typical energy $E_{\gamma} \sim 0.3\,{\rm keV}$. We also point out that the presence of this spectral break significantly reduces the contribution of neutrino sources to the isotropic gamma-ray background, alleviating the longstanding tension between these measurements. In the $\Delta$-resonance scenario, the gamma rays accompanying neutrino production cascade down to MeV-GeV energies and contribute at the $\sim 10\%$ level to the isotropic gamma-ray background at $\sim 3$~GeV. If our proposal is realized, it may imply that we have identified the dominant sources that produce the extragalactic cosmic rays.

astro-ph.HE↗

arXiv2026

MOSAIC: Module Discovery via Sparse Additive Identifiable Causal Learning for Scientific Time Series

Causal representation learning (CRL) seeks to recover latent variables with identifiability guarantees, typically up to permutation and component-wise reparameterization under appropriate assumptions. However, identifiability does not imply interpretability: latent semantics are typically assigned post hoc by alignment with known ground-truth factors. This limitation is particularly acute in scientific time series, where underlying mechanisms are unknown and discovering interpretable structure is a primary goal. In contrast, scientific observations (such as residue-pair distances, climate indices, or process sensors) are inherently semantic, as they correspond to named physical quantities. This raises a key question: can the interpretability of observations be transferred to the identifiable latent space? We propose MOSAIC (Module discovery via Sparse Additive Identifiable Causal learning), a sparse temporal VAE that integrates temporal CRL identifiability with support recovery over observed variables. MOSAIC identifies latent variables via regime-conditioned temporal variation, and recovers for each latent a sparse set of associated observations through an additive decoder, yielding module-level interpretability. We show that ANOVA main-effect supports are identifiable under general smooth mixing functions, and provide finite-sample recovery guarantees for a tractable sparse-additive variant. Empirically, MOSAIC recovers domain-consistent variable groups across RNA molecular dynamics, solar wind, ENSO climate, the Tennessee Eastman process, and a synthetic tokamak benchmark, enabling interpretable discovery of latent mechanisms in scientific time series.

cs.LG↗

arXiv2026

When and What to Ask: AskBench and Rubric-Guided RLVR for LLM Clarification

Large language models (LLMs) often respond even when prompts omit critical details or include misleading information, leading to hallucinations or reinforced misconceptions. We study how to evaluate and improve LLMs' ability to decide when and what to ask for clarification without sacrificing task performance. We introduce AskBench, an interactive benchmark that converts standard QA pairs into multi-turn interactions with explicit checkpoints. A unified judge loop evaluates final answers and simulates user responses as needed. AskBench covers two settings: AskMind, with intent-deficient queries requiring clarification, and AskOverconfidence, with queries containing false premises that must be identified and corrected. We further propose rubric-guided reinforcement learning with verifier-based rewards (RLVR), which uses structured rubrics to encourage targeted clarification. Experiments show consistent improvements in accuracy, rubric adherence, and interaction efficiency, with strong generalization to unseen domains.

cs.CL↗

arXiv2026

The Cosmic-ray Knee as a Local Signature of Nearby PeVatrons

A "knee" in the cosmic-ray spectrum, characterized by a sudden steepening of the spectral shape at $\sim 4$ PeV, may be interpreted either as a global feature of Galactic cosmic rays or as a local signature. In the former scenario, cosmic-ray spectra throughout the Galaxy would be similar to that observed in the solar neighborhood, and the knee would be a common feature of the cosmic-ray sea. In the latter scenario, the PeV cosmic-ray flux varies across the Galactic disk, and the knee is dominantly contributed by a small number of nearby sources. By simulating cosmic-ray propagation in the Galactic magnetic field and interstellar medium, we show that the two scenarios correspond to different regimes of the birth rate of PeV proton accelerators and depend on the presence of powerful nearby sources. By comparison with both cosmic-ray and gamma-ray observations, we find that a local knee would be best explained by sources located at distances of order $\sim1$ kpc and with ages in the range 0.1-1 Myr, with the Cygnus Cocoon being a particularly promising candidate.

astro-ph.HE↗

arXiv2025

The Sensitivity of PUEO to Cosmogenic Neutrinos and Exotic Physics Scenarios

Several observatories designed to detect ultrahigh-energy neutrinos are planned for the next decade. The most imminent of these is the Payload for Ultrahigh Energy Observations (PUEO), a long-duration balloon-based experiment that will provide unprecedented sensitivity to neutrinos with energies in the range of ~ 1 - 1000 EeV. In this work, we assess the scientific reach of PUEO. In particular, we evaluate the sensitivity of this observatory to cosmogenic neutrinos and, in turn, to the proton fraction of the ultrahigh-energy cosmic-ray spectrum. We also consider the potential of PUEO to probe scenarios in which neutrinos are produced through the decays of ultraheavy dark matter particles or are radiated from cosmic strings. We find that PUEO will be able to constrain the proton composition of ultrahigh-energy cosmic rays in scenarios that feature very strong source evolution and in which protons are accelerated to extremely high energies. Although gamma-ray observations are generally more sensitive to decaying particles than neutrino observations, PUEO is expected to set the strongest neutrino-detector constraints above 10^19 eV. PUEO will also provide the strongest constraints on some models of cosmic strings.

astro-ph.HE↗

arXiv2025

Particle Astrophysics with High and Ultrahigh Energy Neutrinos

We summarize recent results of the observations of high (1 TeV-100 PeV) and ultrahigh ($\geq 100$ PeV) energy neutrinos, including the detection of a diffuse cosmic high-energy neutrino background, the identification of the first neutrino source candidates, and the observation of high-energy neutrinos from the Galactic plane. These findings open a new window to the universe by enabling the use of neutrinos to probe the cosmos that are otherwise inaccessible via photons. Although the origins of most detected neutrinos remain uncertain, we highlight several distinctive features of their sources that have emerged from current observations.

astro-ph.HE↗

arXiv2025

Search for synchrotron pair echo emission following KM3-230213A

The KM3NeT Collaboration has recently reported the detection of an extraordinary ultra-high-energy neutrino event with an energy of 220 PeV. Ultrahigh energy neutrinos and gamma-rays are co-produced in ultrahigh energy cosmic-ray interactions. If a UHE neutrino was produced within the large-scale structure around the source where it was accelerated, gamma-ray emission may be expected via the synchrotron pair echo mechanism. Here, we develop the synchrotron pair echo model in the specific context of the KM3NeT neutrino. Motivated by the fact that the synchrotron pair echo signal is expected to peak in the GeV - TeV band, and that the signal may appear as a dim, transient source, we investigate the data collected by the Large Area Telescope (LAT) on-board the \textit{Fermi} Gamma-ray Space Telescope for transient and sub-threshold gamma-ray sources in the vicinity of the KM3NeT neutrino. We find three sub-threshold sources with TS $\gtrsim 16$ within $3.5^{\circ}$ of the neutrino event not included in any existing \textit{Fermi}-LAT catalogs, but note that none of the identified sub-threshold sources seem to be compelling candidates for synchrotron pair echo emission.

astro-ph.HE↗

arXiv2025

REHEARSE: Experiential Rehearsal for Verbal Confidence Calibration in Large Language Models

Large language models (LLMs) often express verbal confidence that is poorly aligned with actual correctness, limiting their reliability in safety-critical applications. Existing prompt-based methods treat calibration largely as a one-shot inference problem, relying on either instance-level reasoning or post-hoc self-assessment. We introduce Rehearse (Experiential Rehearsal), a training-free method that instead enables models to adapt from their own scored confidence experience. In a credence-calibration game grounded in a strictly proper scoring rule, the model receives feedback on prior confidence decisions; this experience is summarized in a post-game trajectory prefix that captures systematic over- or under-confidence. At inference time, the model applies this cross-instance calibration signal to the chain-of-thought reasoning trace for each new question. Across four LLMs, three benchmarks, and five random seeds, Rehearse achieves the lowest average ECE among training-free methods with improved accuracy, reducing average ECE by 58% relative to the uncalibrated baseline. Code is available at https://anonymous.4open.science/r/Experiential-Rehearsal-7C77/.

cs.CL↗

arXiv2025

A multi-wavelength view of the multi-messenger sources NGC 1068 and PKS 1502+1061

Multi-messenger astronomy offers a powerful approach to studying high-energy radiative processes in astrophysical sources. A notable example was seen in 2017, when the IceCube Neutrino Observatory detected a high-energy neutrino event that was found to coincide with a gamma-ray flare from a blazar. Since then, numerous multi-messenger studies combining neutrino and photon data have been conducted, yet the origin of neutrinos from active galactic nuclei (AGN) remains uncertain. In this work, we present the results of an X-ray observing program targeting two AGNs, NGC 1068 and PKS 1502+106. The multi-wavelength dataset includes new observations from NICER and NuSTAR from the observing proposal along with gamma-ray data collected using Fermi-LAT, and one archival observation from Chandra. Additionally, we derive the neutrino fluxes for both AGNs using ten years of IceCube data and neutrino spectra predicted by theoretical models. These results demonstrate the value of combining multi-messenger data in building and constraining theoretical models. They also highlight the importance of testing model predictions against observational data to refine measurements of both the neutrino flux and spectral shape.

astro-ph.HE↗

arXiv2025

Constraints on diffuse X-ray Emission from the TeV halo Candidate HESS J1813-126

Extended regions of very high energy $\gamma$-ray emission associated with middle-aged pulsars have been found by $\gamma$-ray observatories. These regions, called TeV halos or pulsar halos, are thought to be created when energetic electrons from a pulsar or pulsar wind nebula transport into interstellar medium and undergo inverse Compton scattering with the cosmic microwave background radiation. The same electrons are expected to emit synchrotron emission in the X-ray band in the interstellar magnetic field. HESS J1813-126 is a pulsar halo candidate from which TeV $\gamma$-ray emission with extension 0.21\degr and a hard $E^{-2}$ spectrum is observed. We searched for the synchrotron component of this pulsar halo with Swift-XRT. In particular, we observed two fields within the region covered by HESS J1813-126 for 35 ksec each and a region nearby as a background reference for 10 ksec. We find no evidence for excess X-ray emission from the two observations near HESS J1813-126 and place an upper limit differential flux of $4.32\times 10^{-4}\, \rm keV^{-1}\, cm^{-2}\,s^{-1} $ and $5.38\times 10^{-4}\, \rm keV^{-1}\, cm^{-2}\,s^{-1} $ at 1 keV assuming an $E^{-2}$ power law spectrum. The non-detection implies that the magnetic field inside the halo is not significantly enhanced compared to the average Galactic magnetic field.

astro-ph.HE↗

arXiv2025

Cascaded Gamma-ray Emission Associated with the KM3NeT Ultra-High-Energy Event KM3-230213A

A neutrino-like event with an energy of $\sim 220 \,{\rm PeV}$ was recently detected by the KM3NeT/ARCA telescope. If this neutrino comes from an astrophysical source, or from the interaction of an ultra-high-energy cosmic ray in the intergalactic medium, the ultra-high-energy gamma rays that are co-produced with the neutrinos will scatter with the extragalactic background light, producing an electromagnetic cascade and resulting in emission at GeV-to-TeV energies. In this paper, we compute the gamma-ray flux from this neutrino source considering various source distances and strengths of the intergalactic magnetic field (IGMF). We find that the associated gamma-ray emission could be observed by existing imaging air cherenkov telescopes and air shower gamma-ray observatories, unless the strength of the IGMF is $B\gtrsim 3\times 10^{-13}$ G, or the ultra-high-energy gamma-rays are attenuated inside of the source itself. In the latter case, this source is expected to be radio-loud.

astro-ph.HE↗

arXiv2025

Ideas and Requirements for the Global Cosmic-Ray Observatory (GCOS)

After a successful kick-off meeting in 2021. two workshops in 2022 and 2023 on the future Global Cosmic-Ray Observatory (GCOS) focused mainly on a straw man design of the detector and science possibilities for astro- and particle physics. About 100 participants gathered for in-person and hybrid panel discussions. In this report, we summarize these discussions, present a preliminary straw-man design for GCOS and collect short write-ups of the flash talks given during the focus sessions.

astro-ph.IM↗

arXiv2025

Multi-messenger Emission by Magnetically Arrested Disks and Relativistic Jets of Black Hole X-ray Binaries

Black hole X-ray binaries (BHXBs) are observed in various wavelengths from radio to GeV gamma-ray. Several BHXBs, including MAXI J1820+070 and Cygnus X-1, are also found to emit ultrahigh-energy (UHE; photon energy $>$100 TeV) gamma rays. The origin and production mechanism of the multi-wavelength emission of BHXBs are under debate. We propose a scenario where relativistic particles from magnetically arrested disks (MADs), which could form when BHXBs are in quiescent or hard states, produce UHE gamma rays, while electrons in the jets produce GeV gamma-ray emission. Specifically, magnetic turbulence in MADs heats up and accelerates electrons and protons, while magnetic reconnection in jets accelerates electrons. Sub-PeV gamma rays and neutrinos are produced when relativistic protons interact with the thermal protons and the radiation by thermal electrons in the disk. We discuss the perspectives of observing sub-PeV multi-messenger signals from individual BHXBs. Finally, we evaluate the integrated fluxes of the quiescent and hard-state BHXB population and find that BHXBs may contribute to the Galactic diffuse emission above $\sim 100$ TeV.

astro-ph.HE↗

arXiv2024

Prospects for Observing Astrophysical Transients with GeV Neutrinos

Although Cherenkov detectors of high-energy neutrinos in ice and water are often optimized to detect TeV-PeV neutrinos, they may also be sensitive to transient neutrino sources in the 1-100~GeV energy range. A wide variety of transient sources have been predicted to emit GeV neutrinos. In light of the upcoming IceCube-Upgrade, which will extend the IceCube detector's sensitivity down to a few GeV, as well as improve its angular resolution, we survey a variety of transient source models and compare their predicted neutrino fluences to detector sensitivities, in particular those of IceCube-DeepCore and the IceCube Upgrade. We consider the ranges of neutrino fluence from transients powered by non-relativistic shocks, such as novae, supernovae, fast blue optical transients, and tidal disruption events. We also consider fast radio bursts and relativistic outflows of high- and low-luminosity gamma-ray bursts. Our study sheds light on the prospects of observing GeV transients with existing and upcoming neutrino facilities.

astro-ph.HE↗

arXiv2024

An NGC 1068-Informed Understanding of Neutrino Emission of the Active Galactic Nucleus TXS 0506+056

We present arguments that the neutrinos observed by IceCube from the active galactic nucleus TXS 0506+056 may originate near its core and not in the blazar jet. The origin of the neutrinos is consistent with the mechanism that produces the neutrino flux observed from the active galaxies NGC 1068 and NGC 4151, but requires an Eddington luminosity cosmic ray flux to compensate for its larger distance. Like NGC 1068, the source is characterized by episodes of high X-ray emission and is gamma-ray-obscured during the 2014 burst, and there is evidence that this is also the case during the short burst in 2017 that produced IC-170922. The observations may be explained as a flux originating in an obscured core within $10 \sim 100$ Schwarzschild radii from the central black hole, which is not transparent to gamma rays from neutral pions accompanying the neutrinos.

astro-ph.HE↗

arXiv2024

TurtleBench: Evaluating Top Language Models via Real-World Yes/No Puzzles

As the application of Large Language Models (LLMs) expands, the demand for reliable evaluations increases. Existing LLM evaluation benchmarks primarily rely on static datasets, making it challenging to assess model performance in dynamic interactions with users. Moreover, these benchmarks often depend on specific background knowledge, complicating the measurement of a model's logical reasoning capabilities. Other dynamic evaluation methods based on strong models or manual efforts may introduce biases and incur high costs and time demands, hindering large-scale application. To address these issues, we propose TurtleBench. TurtleBench collects real user guesses from our online Turtle Soup Puzzle platform that we developed. This approach allows for the relatively dynamic generation of evaluation datasets, mitigating the risk of model cheating while aligning assessments more closely with genuine user needs for reasoning capabilities, thus enhancing the reliability of evaluations. TurtleBench includes 1,532 user guesses along with the correctness of guesses after annotation. Using this dataset, we thoroughly evaluated nine of the most advanced LLMs available today. Notably, the OpenAI o1 series models did not achieve leading results in these evaluations. We propose several hypotheses for further research, such as "the latent reasoning of o1 utilizes trivial Chain-of-Thought (CoT) techniques" and "increasing CoT length not only provides reasoning benefits but also incurs noise costs."

cs.CL↗