arXiv subjects

Mariko Kato

Publications and source records attributed to Mariko Kato.

At least 19 recordsLinked to original sources

arXiv2026

Metamorphic novae. I. Gaia22alz -- Violent relaxation starting from static evolution

Gaia22alz (nova Vel 2022) shows characteristic properties in its light curves and spectral evolution similar to slow novae HR Del, V723 Cas, and V5558 Sgr. Their outburst begins in a way of hydrostatic expansion with no mass loss, followed by large amplitude ($ΔV \sim 2$ mag) multiple peaks around the mean magnitude of $M_V \sim -5.5$, and eventually it decays similarly to normal classical novae. We present a light curve model of Gaia22alz based on our metamorphic nova theory: a nova starts from a static expansion and then metamorphoses to an optically thick wind evolution through a large amplitude multiple peak phase as a violent relaxation process. Our theoretical light curve for the rising phase consistently explains the slow rise and early spectral evolution of Gaia22alz. Gaia22als is a close binary as well as other three novae, which is consistent with our metamorphic nova theory that the companion motion in the nova envelope triggers the metamorphic transition. Our results do not support E. Aydi et al.'s explanation that the slow optical rise of Gaia22alz is owing to the irradiation of X-ray/UV flash.

astro-ph.SR↗

arXiv2026

Light-curve analysis of the classical nova YZ Ret: revisit with fully self-consistent nova outburst models

YZ Ret is a well observed nova from the first detection of an X-ray flash phase in classical novae until the end of a supersoft X-ray source phase. We have reanalyzed multiwavelength light curves of YZ Ret over the total duration of the outburst. Our fully self-consistent nova outburst model, a 1.25 $M_\odot$ white dwarf (WD) and a mass accretion rate of $1\times 10^{-9} ~M_\odot$ yr$^{-1}$ on to the WD, reasonably reproduced $V$/$g$ light curves as well as the very early X-ray flash and late supersoft X-ray light curves. This model explains the emergence epoch of the GeV gamma-ray emission because a shock naturally arises far outside the nova (WD) photosphere just after the optical maximum. In our model, before optical maximum, later ejected matter has a smaller velocity than that of earlier ejecta and therefore the ejecta expands while, after optical maximum, later ejected matter has a larger velocity than that of earlier ejecta and collision with the earlier ejecta makes a shock.

astro-ph.SR↗

arXiv2026

Identification of a superbright nova in our Galaxy: revisiting the light curve analyses of the candidate novae CP Lac, CP Pup, V838 Her, V597 Pup, V5583 Sgr, and V5589 Sgr

The two very fast novae in our Galaxy, V1500 Cyg and V1674 Her, reached the absolute $V$ magnitude of $M_{V,\rm max}\approx -10.4$. These brightest novae are identified as superbright novae. Such superbrightnesses are realized if an optically thick shocked shell forms far outside the nova (white dwarf) photosphere and expands at the velocity of a few thousands km s$^{-1}$. We have analyzed $V$ light curves of six candidates novae, CP Lac, CP Pup, V838 Her, V597 Pup, V5583 Sgr, and V5589 Sgr by comparing with our model nova light curves of 1.25 $M_\odot$ and 1.35 $M_\odot$ white dwarfs calculated with the free-free emission from a nova wind. Analyzing the data with our method, we obtain $M_{V, \rm max}=-10$ for V838 Her, and we suggest that its large peak optical luminosity is due to the ejection of an optically thick shell like we suggested for the case of V1674 Her. The light curves of the other five novae, CP Lac, CP Pup, V597 Pup, V5583 Sgr, and V5589 Sgr are reproduced only with the free-free emission model $V$ light curves without an optically thick shocked shell. Their $M_{V,\rm max}$ are fainter than $-10$ mag. We conclude that these five are not superbright novae.

astro-ph.SR↗

arXiv2026

A multiwavelength light curve analysis of the very fast nova V1723 Sco

We have analyzed multiwavelength light curves of the very fast nova V1723 Sco, based on our fully self-consistent nova explosion models. The time-stretching method gives the distance modulus in the $V$ band of $(m-M)_V = 15.3\pm 0.2$. Then the absolute $V$ magnitude reaches $M_{V,\rm max}= m_{V, \rm max} - (m-M)_V= 6.77 - 15.3\pm 0.2 = -8.5\pm 0.2$. Using our fully self-consistent nova outburst model combined with the optically thick winds on a $1.25 ~M_\odot$ white dwarf accreted by a mass accretion rate of $\dot{M}_{\rm acc}=1\times 10^{-9} ~M_\odot$ yr$^{-1}$, we successfully reproduce the overall $V$ light curve with a free-free emission model as well as the supersoft X-ray light curve with a blackbody approximation model. The epoch of the first GeV gamma-ray detection is almost coincident with the epoch of our model $V$ peak. This supports the shock formation mechanism that a strong shock arises soon after the optical $V$ maximum far outside the WD photosphere. We conclude that the shocked shell is optically thin.

astro-ph.SR↗

arXiv2026

Upper bound of ejecta mass in a nova outburst

We present the maximum ejecta mass $(M_{\rm ej})_{\rm max}$ and the maximum ratio of ejecta mass and accreted mass $(M_{\rm ej}/M_{\rm acc})_{\rm max}$ of a nova for various white dwarf (WD) masses ($M_{\rm WD}=0.6$ - 1.38 $M_\odot$) and mass accretion rates ($\dot{M}_{\rm acc}=1\times 10^{-11}$ - $3\times 10^{-7} ~M_\odot$ yr$^{-1}$) based on the energy balance with nuclear burning. These maximum values serve as an upper bound of mass ejection for individual novae. Recently, B. E. Schaefer concluded that the WD masses in the recurrent novae U Sco and T CrB decreased at nova explosions, because the ejected mass is much larger than the accreted mass, i.e., $M_{\rm ej}/M_{\rm acc}= 26$ and $540$, respectively. These values are derived from the orbital period change at the nova explosions. Recurrent novae have been considered to be a progenitor system of Type Ia supernovae (SNe Ia) because their WD masses are now close to, and will possibly grow up to, 1.38 $M_\odot$ at which WDs explode as SNe Ia. From the different view point of energy generation at the thermonuclear runaway, we have obtained the much smaller value of the maximum ratio of $M_{\rm ej}/M_{\rm acc}\lesssim 2.6$ for a $1.37 ~M_\odot$ WD. This conclusion simply means that the nuclear (hydrogen) burning cannot release energy enough to expel such a large ejecta mass as B. E. Schaefer's claims. We also conclude that $(M_{\rm ej}/M_{\rm acc})_{\rm max}$ hardly increases even if we include the effect of frictional mass ejection process in the common envelope phase of a nova.

astro-ph.SR↗

arXiv2025

Fully self-consistent nova explosion models reproducing light curves of KT Eri, V339 Del, V597 Pup, and SMC NOVA 2016-10a

The rising phase toward the optical maximum of a classical nova is one of the last frontiers of nova study. Constructing free-free emission model light curves based on our fully self-consistent nova explosion models, we present several theoretical light curves of classical novae and compare them with the four novae having the observed rising phase toward the optical maximum. Our 1.25 $M_\odot$ white dwarf (WD) models show excellent agreements with the light curves of KT Eri, V339 Del, and V597 Pup while our 1.35 $M_\odot$ WD models are consistent with the light curves of SMC NOVA 2016-10a. These agreements indicate that the light curves toward the optical maximum of these novae are dominated by free-free emission, rather than by photospheric emission. Our results justify the previously obtained WD masses and distance moduli for these novae, and shows that the post-maximum evolution can be well approximated with the evolution sequences of steady-state envelope solutions.

astro-ph.SR↗

arXiv2025

Mechanistic Fine-tuning for In-context Learning

In-context Learning (ICL) utilizes structured demonstration-query inputs to induce few-shot learning on Language Models (LMs), which are not originally pre-trained on ICL-style data. To bridge the gap between ICL and pre-training, some approaches fine-tune LMs on large ICL-style datasets by an end-to-end paradigm with massive computational costs. To reduce such costs, in this paper, we propose Attention Behavior Fine-Tuning (ABFT), utilizing the previous findings on the inner mechanism of ICL, building training objectives on the attention scores instead of the final outputs, to force the attention scores to focus on the correct label tokens presented in the context and mitigate attention scores from the wrong label tokens. Our experiments on 9 modern LMs and 8 datasets empirically find that ABFT outperforms in performance, robustness, unbiasedness, and efficiency, with only around 0.01% data cost compared to the previous methods. Moreover, our subsequent analysis finds that the end-to-end training objective contains the ABFT objective, suggesting the implicit bias of ICL-style data to the emergence of induction heads. Our work demonstrates the possibility of controlling specific module sequences within LMs to improve their behavior, opening up the future application of mechanistic interpretability.

cs.CL↗

arXiv2025

Optical detection of the X-ray flash in the very fast nova V1674 Her: Optical contribution of the irradiated accretion disk

V1674 Her is one of the fastest and brightest novae, characterized by dense optical photometry in the pre-maximum phase, a rise from $g=17$ to 7 mag, in one-fourth of a day. We present a composite theoretical $V$ light curve model of its early rising phase starting from a quiescent brightness of $g=19.2$ mag. Our light curve model consists of a hot and bright white dwarf (WD) and irradiated accretion disk and companion star. We found that the earliest optical detection of ASAS-SN $g$ band brightness of $g=17.0$ at $t=0.014$ day from the onset of thermonuclear runaway can be explained with the irradiated accretion disk and companion star in the X-ray flash phase of a $1.35 ~M_\odot$ WD. This is the first detection in optical of an X-ray flash phase of a nova. Optically thick winds emerge from the WD photosphere at $t=0.04$ day, and optical flux is dominated by free-free emission from optically-thin ejecta just outside the WD photosphere. Our free-free emission model $V$ light curve reasonably reproduces the dense $g$ light curve of Evryscope that spans from $g=14.8$ (at 0.078 day) to $g=7.1$ (at 0.279 day), including a sudden change of slope in the $g$ light curve from slow to rapid rise at $g=14.3$ on day $0.1$. There is no indication of shocking power during the rising phase from $g=14.8$ to 7.1.

astro-ph.SR↗

arXiv2025

A comprehensive light curve model of the very fast nova V1674 Herculis

V1674 Her is one of the fastest novae, of which the very early phase is well observed including optical rise to the peak over 10 magnitudes. We present a full theoretical light curve model of V1674 Her. Our $1.35~M_\odot$ white dwarf (WD) model with the mass accretion rate of $1\times 10^{-11}~M_\odot$ yr$^{-1}$ explains overall properties including a very fast rise and decay of the optical $V$ light curve. The WD photosphere expands up to $21 ~R_\odot$, thus, a $0.26 ~M_\odot$ companion star orbiting the WD every 3.67 hours, is engulfed 2.7 hours after the onset of thermonuclear runaway, and appears 5.3 days after that. The duration of X-ray flash is only 0.96 hours. The evolution of the expanding envelope and temporal change of the photospheric radius are very consistent with observed optical and X-ray modulations with the orbital and spin (501 s) periods. We confirmed that the decay phase of nova light curve is well approximated by a sequence of steady-state envelope solutions. Using time-stretching method of nova light curves, we obtain the $V$ band distance modulus of $(m-M)_V= 16.3\pm 0.2$, and determine the distance to be $d=8.9\pm 1$ kpc for the interstellar extinction of $E(B-V)= 0.5 \pm 0.05$.

astro-ph.SR↗

arXiv2025

A multiwavelength light curve analysis of the classical nova V392 Per: Optical contribution from an irradiated accretion disk during the nova wind phase

The classical nova V392 Per 2018 is characterized by a very fast optical decline, long binary orbital period of 3.23 days, detection of GeV gamma rays, and almost identical decay trends of $B$, $V$, and $I_{\rm C}$ light curves. The last feature is unique because most novae develop strong emission lines in the nebular phase and these lines contribute especially to the $B$ and $V$ bands and make large differences between the $BV$ and $I_{\rm C}$ light curves. This unique feature can be understood if the optical flux is dominated by continuum until the late phase of the nova outburst. Such continuum radiation is emitted by a bright accretion disk irradiated by a hydrogen burning white dwarf (WD) and viscous heating disk with high mass-accretion rate after the hydrogen burning ended. We present a comprehensive nova outburst model that reproduces all of these light curves. We determined the WD mass to be $M_{\rm WD}=1.35$ - $1.37 ~M_\odot$ and the distance modulus in the $V$ band to be $(m-M)_V=14.6 \pm 0.2$; the distance is $d= 3.45\pm 0.5$ kpc for the reddening of $E(B-V)=0.62$.

astro-ph.HE↗

arXiv2025

A demarcation criterion for hydrogen burning of millinovae

Millinovae are a new class of transient supersoft X-ray sources with no clear signature of mass ejection. They show similar triangle shapes of $V/I$ band light curves with thousand times fainter peaks than typical classical novae. Maccarone et al. regarded the prototype millinova, ASASSN-16oh, as a dwarf nova and interpreted the supersoft X-rays to originate from an accretion belt on a white dwarf (WD). Kato et al. proposed a nova model induced by a high-rate mass-accretion during a dwarf nova outburst; the X-rays originate from the photosphere of a hydrogen-burning hot WD whereas the $V/I$ band photons are from the irradiated accretion disk. Because each peak brightness differs largely from millinova to millinova, we suspect that not all the millinova candidates host a hydrogen burning WD. Based on the light curve analysis of the classical nova KT Eri that has a bright disk, we find that the disk is more than two magnitudes brighter when the disk is irradiated by the hydrogen burning WD than when not irradiated. We present the demarcation criterion for hydrogen burning to be $I_{\rm q} - I_{\rm max} > 2.2$, where $I_q$ and $I_{\rm max}$ are the $I$ magnitudes in quiescence and at maximum light, respectively. Among many candidates, this requirement is satisfied with the two millinovae in which soft X-rays were detected.

astro-ph.SR↗

arXiv2025

Revisiting In-context Learning Inference Circuit in Large Language Models

In-context Learning (ICL) is an emerging few-shot learning paradigm on Language Models (LMs) with inner mechanisms un-explored. There are already existing works describing the inner processing of ICL, while they struggle to capture all the inference phenomena in large language models. Therefore, this paper proposes a comprehensive circuit to model the inference dynamics and try to explain the observed phenomena of ICL. In detail, we divide ICL inference into 3 major operations: (1) Input Text Encode: LMs encode every input text (in the demonstrations and queries) into linear representation in the hidden states with sufficient information to solve ICL tasks. (2) Semantics Merge: LMs merge the encoded representations of demonstrations with their corresponding label tokens to produce joint representations of labels and demonstrations. (3) Feature Retrieval and Copy: LMs search the joint representations of demonstrations similar to the query representation on a task subspace, and copy the searched representations into the query. Then, language model heads capture these copied label representations to a certain extent and decode them into predicted labels. Through careful measurements, the proposed inference circuit successfully captures and unifies many fragmented phenomena observed during the ICL process, making it a comprehensive and practical explanation of the ICL inference process. Moreover, ablation analysis by disabling the proposed steps seriously damages the ICL performance, suggesting the proposed inference circuit is a dominating mechanism. Additionally, we confirm and list some bypass mechanisms that solve ICL tasks in parallel with the proposed circuit.

cs.CL↗

arXiv2025

Affinity and Diversity: A Unified Metric for Demonstration Selection via Internal Representations

The performance of In-Context Learning (ICL) is highly sensitive to the selected demonstrations. Existing approaches to demonstration selection optimize different objectives, yielding inconsistent results. To address this, we propose a unified metric--affinity and diversity--that leverages ICL model's internal representations. Our experiments show that both affinity and diversity strongly correlate with test accuracies, indicating their effectiveness for demonstration selection. Moreover, we show that our proposed metrics align well with various previous works to unify the inconsistency.

cs.CL↗

arXiv2025

Token-based Decision Criteria Are Suboptimal in In-context Learning

In-Context Learning (ICL) typically utilizes classification criteria from output probabilities of manually selected label tokens. However, we argue that such token-based classification criteria lead to suboptimal decision boundaries, despite delicate calibrations through translation and constrained rotation applied. To address this problem, we propose Hidden Calibration, which renounces token probabilities and uses the nearest centroid classifier on the LM's last hidden states. In detail, we assign the label of the nearest centroid previously estimated from a calibration set to the test sample as the predicted label. Our experiments on 6 models and 10 classification datasets indicate that Hidden Calibration consistently outperforms current token-based baselines by about 20%~50%, achieving a strong state-of-the-art in ICL. Our further analysis demonstrates that Hidden Calibration finds better classification criteria with less inter-class overlap, and LMs provide linearly separable intra-class clusters with the help of demonstrations, which supports Hidden Calibration and gives new insights into the principle of ICL. Our official code implementation can be found at https://github.com/hc495/Hidden_Calibration.

cs.CL↗

arXiv2024

Understanding Token Probability Encoding in Output Embeddings

In this paper, we investigate the output token probability information in the output embedding of language models. We find an approximate common log-linear encoding of output token probabilities within the output embedding vectors and empirically demonstrate that it is accurate and sparse. As a causality examination, we steer the encoding in output embedding to modify the output probability distribution accurately. Moreover, the sparsity we find in output probability encoding suggests that a large number of dimensions in the output embedding do not contribute to causal language modeling. Therefore, we attempt to delete the output-unrelated dimensions and find more than 30% of the dimensions can be deleted without significant movement in output distribution and sequence generation. Additionally, in the pre-training dynamics of language models, we find that the output embeddings capture the corpus token frequency information in early steps, even before an obvious convergence of parameters starts.

cs.CL↗

arXiv2024

A multiwavelength light curve analysis of the classical nova KT Eri: Optical contribution from a large irradiated accretion disk

KT Eri is a classical nova which went into outburst in 2009. Recent photometric analysis in quiescence indicates a relatively longer orbital period of 2.6 days, so that KT Eri could host a very bright accretion disk during the outburst like in the recurrent nova U Sco, the orbital period of which is 1.23 days. We reproduced the optical $V$ light curve as well as the supersoft X-ray light curve of KT Eri in outburst, assuming a large irradiated disk during a nova wind phase of the outburst while a normal size disk after the nova winds stop. This result is consistent with the temporal variation of wide-band $V$ brightness that varies almost with the intermediate-band Strömgren $y$ brightness, because the $V$ flux is dominated by continuum radiation, the origin of which is a photospheric emission from the very bright disk. We obtained the white dwarf mass to be $M_{\rm WD}= 1.3\pm0.02 ~M_\odot$, the hydrogen-burning turnoff epoch to be $\sim 240$ days after the outburst, the distance modulus in the $V$ band to be $(m-M)_V=13.4\pm 0.2$, and the distance to KT Eri to be $d=4.2\pm0.4$ kpc for the reddening of $E(B-V)= 0.08$. The peak absolute $V$ brightness is about $M_{V, \rm max}= -8.0$ and the corresponding recurrence time is $\sim 3,000$ yr from its ignition mass together with the mean mass-accretion rate of $\dot{M}_{\rm acc}\sim 1\times 10^{-9} ~M_\odot$ yr$^{-1}$ in quiescence. Thus, we suggest that KT Eri is not a recurrent nova.

astro-ph.SR↗

arXiv2024

X-Ray Flashes on Helium Novae

A helium nova occurs on a white dwarf (WD) accreting hydrogen-deficient matter from a helium star companion. When the mass of a helium envelope on the WD reaches a critical value, unstable helium burning ignites to trigger a nova outburst. A bright soft X-ray phase appears in an early outbursting phase of a helium nova before it optically rises toward maximum. Such an X-ray bright phase is called the X-ray flash. We present theoretical light curves of X-ray flashes for 1.0, 1.2, and 1.35 $M_\odot$ helium novae with mass accretion rates of $(1.6-7.5) \times 10^{-7}~M_\odot$ yr$^{-1}$. Long durations of the X-ray flashes (100 days to 10 years) and high X-ray luminosities ($\sim 10^{38}$ erg s$^{-1}$) indicate that X-ray flashes are detectable as a new type of X-ray transients or persistent X-ray sources. An X-ray flash is a precursor of optical brightening, so that the detection of X-ray flashes on helium novae enables us to plan arranged observation for optical pre-maximum phases that have been one of the frontiers of nova study. We found a candidate object of helium-burning X-ray flash from literature on extra-galactic X-ray surveys. This X-ray transient source is consistent with our X-ray flash model of a $1.35 ~M_\odot$ WD.

astro-ph.SR↗

arXiv2024

Physics of nova outbursts: Theoretical models of classical nova outbursts with optically thick winds on $1.2~M_\odot$ and $1.3~M_\odot$ white dwarfs

We present time-dependent nova outburst models with optically thick winds for a 1.2 and 1.35 $M_\odot$ white dwarfs (WDs) with a mass accretion rate of $5 \times 10^{-9}~M_\odot$ yr$^{-1}$ and for a 1.3 $M_\odot$ WD with $2 \times 10^{-9}~M_\odot$ yr$^{-1}$. The X-ray flash occurs 11 days before the optical peak of the 1.2 $M_\odot$ WD and 2.5 days before the peak of the 1.3 $M_\odot$ WD. The wind mass loss rate of the 1.2 $M_\odot$ WD (1.3 $M_\odot$ WD) reaches a peak of $6.4 \times 10^{-5}~M_\odot$ yr$^{-1}$ ($7.4 \times 10^{-5}~M_\odot$ yr$^{-1}$) at the epoch of the maximum photospheric expansion with the lowest photospheric temperature of $\log T_{\rm ph}$ (K)=4.33 (4.35). The nuclear energy generated during the outburst is lost in a form of radiation (61% for the 1.2 $M_\odot$ WD; 47% for the 1.3 $M_\odot$ WD), gravitational energy of ejecta (39%; 52%), and kinetic energy of the wind (0.28%; 0.29%). We found an empirical relation for fast novae between the time to optical maximum from the outburst $t_{\rm peak}$ and the expansion timescale $τ_{\rm exp}$ at $t=0$. With this relation, we are able to predict the time to optical maximum $t_{\rm peak}$ from the ignition model (at $t=0$) without following a time-consuming nova wind evolution.

astro-ph.SR↗