arXiv subjects

Yifu Cai

Publications and source records attributed to Yifu Cai.

At least 19 recordsLinked to original sources

arXiv2026

Phantom crossing from the Standard Model and General Relativity

Suggestions of a late-time phantom crossing from DESI baryon acoustic oscillation measurements, combined with cosmic microwave background and Type-Ia supernova observations, have renewed interest in non-standard dark energy models. In this Letter, we propose a new realization of a low redshift phantom crossing using only well understood ingredients from fermion condensation and general relativity. Our construction relies on the interplay between effective quintom-like dark energy and backreaction from non-linear structure formation. The full cosmological evolution is consistent with $\Lambda$CDM until low redshift, where the backreaction starts to become prominent and a phase transition occurs. At this point, the total dark energy equation of state first crosses $w_{\text{DE}}=-1$ from above. For benchmark values of the backreaction energy density, a second crossing from below occurs at a lower redshift, as suggested by recent observation. Fitting our theoretical model with CPL parametrization, we find the result is consistent with the combined DESI+CMB+SNIa data analysis with DESY5, PantheonPlus, and Union3 datasets. Specifically, for backreaction density parameter $\Omega_{\rm BR}(z=0) = 0.0572$, the CPL parametrization gives the phantom crossing redshift $z_* \simeq 0.35$, $w_0 \simeq -0.76 $ and $w_a \simeq -0.93$.

astro-ph.CO↗

arXiv2026

A short review on Quintom dark energy theory

In this paper, we provide a short review on the Quintom dark energy theory. Firstly, we discuss the No-Go theorem associated with dynamical dark energy, then present some examples of models in which the equation of state (EoS) evolves with time and can cross $w=-1$ . Secondly, we discuss the bouncing universe and emergent universe with Quintom matter. Finally, we discuss the possibility of studying the nature of dark energy by measuring the Cosmic Microwave Background (CMB) polarization rotation angle.

astro-ph.CO↗

arXiv2026

DESI and Dynamical Dark Energy from Extended Pre-geometric Gravity

We consider the simplest quadratic extension of MacDowell-Mansouri pre-geometric gravity preserving the topological pre-volume form symmetry. After symmetry breaking, it becomes $(\mathrm{Lovelock})^2$ gravity, dual to a Galileon-like Horndeski scalar-tensor theory. The gravitational Higgs mechanism forces the Gauss-Bonnet coupling to be inversely proportional to the bare cosmological constant. The quadratic correction renders the gravitational $\theta$-angle dynamical in the form of a gravi-axion, whose effective mass sets the dark energy scale, thus naturally realizing a dynamical dark energy. The model fits DESI's BAO+FS data exceptionally well ($\chi^2_{\rm red} = 1.394$), deviating from $\Lambda\mathrm{CDM}$ by only a few percent in the gravitational slip parameter $\gamma(z)$ with stable tensor perturbations. This analysis establishes a concrete, testable bridge between pre-geometric gravity and cosmic acceleration.

gr-qc↗

arXiv2026

TimeSeriesExamAgent: Creating Time Series Reasoning Benchmarks at Scale

Large Language Models (LLMs) have shown promising performance in time series modeling tasks, but do they truly understand time series data? While multiple benchmarks have been proposed to answer this fundamental question, most are manually curated and focus on narrow domains or specific skill sets. To address this limitation, we propose scalable methods for creating comprehensive time series reasoning benchmarks that combine the flexibility of templates with the creativity of LLM agents. We first develop TimeSeriesExam, a multiple-choice benchmark using synthetic time series to evaluate LLMs across five core reasoning categories: pattern recognitionnoise understandingsimilarity analysisanomaly detection, and causality. Then, with TimeSeriesExamAgent, we scale our approach by automatically generating benchmarks from real-world datasets spanning healthcare, finance and weather domains. Through multi-dimensional quality evaluation, we demonstrate that our automatically generated benchmarks achieve diversity comparable to manually curated alternatives. However, our experiments reveal that LLM performance remains limited in both abstract time series reasoning and domain-specific applications, highlighting ongoing challenges in enabling effective time series understanding in these models. TimeSeriesExamAgent is available at https://github.com/magwiazda/TimeSeriesExamAgent.

cs.AI↗

arXiv2026

Feynman: Knowledge-Infused Diagramming Agent for Scalable Visual Designs

Visual design is an essential application of state-of-the-art multi-modal AI systems. Improving these systems requires high-quality vision-language data at scale. Despite the abundance of internet image and text data, knowledge-rich and well-aligned image-text pairs are rare. In this paper, we present a scalable diagram generation pipeline built with our agent, Feynman. To create diagrams, Feynman first enumerates domain-specific knowledge components (''ideas'') and performs code planning based on the ideas. Given the plan, Feynman translates ideas into simple declarative programs and iterates to receives feedback and visually refine diagrams. Finally, the declarative programs are rendered by the Penrose diagramming system. The optimization-based rendering of Penrose preserves the visual semantics while injecting fresh randomness into the layout, thereby producing diagrams with visual consistency and diversity. As a result, Feynman can author diagrams along with grounded captions with very little cost and time. Using Feynman, we synthesized a dataset with more than 100k well-aligned diagram-caption pairs. We also curate a visual-language benchmark, Diagramma, from freshly generated data. Diagramma can be used for evaluating the visual reasoning capabilities of vision-language models. We plan to release the dataset, benchmark, and the full agent pipeline as an open-source project.

cs.LG↗

arXiv2026

A sound horizon independent measurement of $H_0$ from BOSS, DESI and DES Y3

We present a sound horizon independent measurement of the Hubble parameter using a multiprobe large-scale structure analysis. Removing the dependency on the sound horizon with a rescaling procedure at the matter power spectrum level, we analyse the BOSS full-shape power spectrum and bispectrum (for the first time) using the effective field theory of large-scale structure up to one loop. We combine this analysis with the auto- and cross-angular power spectra from the DESI Legacy Imaging Survey DR9, the $3 \times 2$pt analysis from DES Y3, and the CMB gravitational lensing power spectrum from Planck PR3. Our baseline analysis, that does not rely on supernovae data, yields $h = 0.702^{+0.022}_{-0.024}$, $\Omega_m = 0.310 \pm 0.013$, and $\sigma_8 = 0.799 \pm 0.020$, corresponding to $3-4 \%$ precision measurements. When adding supernovae data from Pantheon+, we obtain a $2.6 \%$ measurement of $h$, with $h = 0.686 \pm 0.018$. We further note that our EFTBOSS analysis indicates a slight deviation of the BAO scale parameter (at $1.8 \sigma$) from its $\Lambda$CDM value, caused by the small scales of the bispectrum. We finally use the sound horizon-free EFTBOSS analysis as a diagnosis for the presence of new physics, finding that our results are consistent with the recent hints of evolving dark energy.

astro-ph.CO↗

arXiv2025

(Lovelock)$^2$ inflation: explaining the ACT data and equivalence to Higgs-Gauss-Bonnet inflation

We revisit the Starobinsky model of inflation in light of recent data from the Atacama Cosmology Telescope (ACT), which indicates a potential preference for a slightly larger scalar spectral index $n_s$ than predicted by the standard $R^2$ scenario. We demonstrate that a natural one-parameter generalization to a quadratic model $\sim L+L^2$ in the Lovelock invariant $L=R+\frac{\alpha}{4}{\cal G}$ ($\cal G$ is the Gauss--Bonnet term), can effectively resolve this minor tension. Scalar-tensor formulation of this theory yields an Einstein-frame Starobinsky-type scalar potential augmented by Gauss--Bonnet and derivative couplings, which modify the inflationary slow-roll dynamics. We show that a non-zero coupling $\alpha$ for the Gauss-Bonnet term can shift $(n_s, r)$ along a trajectory that brings the predictions into better agreement with the ACT likelihood. We also find that $L+L^2$ gravity, in its scalar-tensor formulation, is equivalent to Higgs inflation coupled to the Gauss--Bonnet term, and belongs to the Horndeski/galileon class of modified gravities. This work establishes the quadratic $f(L)$ gravity as a compelling and physically motivated extension that preserves the successes of Starobinsky inflation while improving its fit to modern precision cosmological data.

gr-qc↗

arXiv2025

A Focused Review of Quintom Cosmology: From Quintom Dark Energy to Quintom Bounce

The recently released data of DESI DR2 favors a dynamical dark energy theory, with the equation of state crossing the cosmological constant boundary $w=-1$. In this paper, we briefly review quintom cosmology, especially the quintom bounce. We will give three examples of a quintom bounce and one example of a cyclic universe with quintom matter.

astro-ph.CO↗

arXiv2025

The Quintom theory of dark energy after DESI DR2

Observations from DESI DR2 are challenging the $\Lambda$CDM paradigm by suggesting that the equation-of-state parameter of dark energy evolves across $w = -1$, a phenomenon known as the Quintom scenario. Inspired by this development, we present a staged review of Quintom cosmology including its theoretical foundations, observational supports, and implications as well as possible extensions. We first trace the historical progression from Einstein's static cosmological constant to modern dynamical dark energy, summarizing recent cosmological constraints that favor an evolving $w(z)$ along time. A key focus is the theoretical no-go theorem for dark energy showing that no single canonical field or perfect fluid model can smoothly cross the $w = -1$ boundary. We then survey viable Quintom constructions, including two-field models, single-scalar fields with higher derivatives, modified gravity frameworks, interacting dark energy, and an effective field theory approach that unifies these mechanisms. Possible interactions of Quintom fields with ordinary matter and the potential roles in yielding non-singular universe solutions are discussed.

astro-ph.CO↗

arXiv2025

Holographic superconductor with dark sector probed by entanglement entropy in higher dimensional AdS spacetime

We investigate the holographic entanglement entropy (HEE) with dark matter in a higher-dimensional AdS black hole spacetime including full back reaction, revealing its role as a diagnostic tool for critical phenomena in strongly coupled systems. By analyzing the HEE, we uncover distinct signatures of the metal/superconductor phase transition, demonstrating that the critical temperature is dynamically tuned by both the dark matter coupling strength and the chemical potential ratio between visible and dark matter sectors. Notably, near the criticality, the HEE exhibits a novel scaling behavior: it grows linearly with the dark matter coupling but displays a nonlinear, accelerated enhancement as the chemical potential ratio between the Maxwell and dark matter sectors increases.

hep-th↗

arXiv2025

TimeSeriesGym: A Scalable Benchmark for (Time Series) Machine Learning Engineering Agents

We introduce TimeSeriesGym, a scalable benchmarking framework for evaluating Artificial Intelligence (AI) agents on time series machine learning engineering challenges. Existing benchmarks lack scalability, focus narrowly on model building in well-defined settings, and evaluate only a limited set of research artifacts (e.g., CSV submission files). To make AI agent benchmarking more relevant to the practice of machine learning engineering, our framework scales along two critical dimensions. First, recognizing that effective ML engineering requires a range of diverse skills, TimeSeriesGym incorporates challenges from diverse sources spanning multiple domains and tasks. We design challenges to evaluate both isolated capabilities (including data handling, understanding research repositories, and code translation) and their combinations, and rather than addressing each challenge independently, we develop tools that support designing multiple challenges at scale. Second, we implement evaluation mechanisms for multiple research artifacts, including submission files, code, and models, using both precise numeric measures and more flexible LLM-based evaluation approaches. This dual strategy balances objective assessment with contextual judgment. Although our initial focus is on time series applications, our framework can be readily extended to other data modalities, broadly enhancing the comprehensiveness and practical utility of agentic AI evaluation. We open-source our benchmarking framework to facilitate future research on the ML engineering capabilities of AI agents.

cs.LG↗

arXiv2024

Sound Speed Resonance of Gravitational Waves in Gauss-Bonnet-coupled inflation

We demonstrate the occurrence of Sound Speed Resonances (SSR) in Gauss-Bonnet-coupled inflation across a wide range of coupling functions and parameters. After inflation, the damped oscillations of the inflaton around its potential minimum induces damped oscillations of the sound speed of tensor modes, leading to resonant amplification of the latter. Once the inflaton stabilizes around the minimum, the tensor sound speed reduces to unity (speed of light). In the context of multi-field inflation, the sound speed oscillations can be followed by a second phase of inflation, resulting in a distinctive stochastic background of Gravitational Waves (GWs). We show that these GW signals can be probed by upcoming experiments such as $\textbf{SKA}$, $\textbf{DECIGO}$, and $\textbf{BBO}$ depending on the duration of the second inflationary phase.

gr-qc↗

arXiv2024

TimeSeriesExam: A time series understanding exam

Large Language Models (LLMs) have recently demonstrated a remarkable ability to model time series data. These capabilities can be partly explained if LLMs understand basic time series concepts. However, our knowledge of what these models understand about time series data remains relatively limited. To address this gap, we introduce TimeSeriesExam, a configurable and scalable multiple-choice question exam designed to assess LLMs across five core time series understanding categories: pattern recognition, noise understanding, similarity analysis, anomaly detection, and causality analysis. TimeSeriesExam comprises of over 700 questions, procedurally generated using 104 carefully curated templates and iteratively refined to balance difficulty and their ability to discriminate good from bad models. We test 7 state-of-the-art LLMs on the TimeSeriesExam and provide the first comprehensive evaluation of their time series understanding abilities. Our results suggest that closed-source models such as GPT-4 and Gemini understand simple time series concepts significantly better than their open-source counterparts, while all models struggle with complex concepts such as causality analysis. We believe that the ability to programatically generate questions is fundamental to assessing and improving LLM's ability to understand and reason about time series data.

cs.AI↗

arXiv2024

MOMENT: A Family of Open Time-series Foundation Models

We introduce MOMENT, a family of open-source foundation models for general-purpose time series analysis. Pre-training large models on time series data is challenging due to (1) the absence of a large and cohesive public time series repository, and (2) diverse time series characteristics which make multi-dataset training onerous. Additionally, (3) experimental benchmarks to evaluate these models, especially in scenarios with limited resources, time, and supervision, are still in their nascent stages. To address these challenges, we compile a large and diverse collection of public time series, called the Time series Pile, and systematically tackle time series-specific challenges to unlock large-scale multi-dataset pre-training. Finally, we build on recent work to design a benchmark to evaluate time series foundation models on diverse tasks and datasets in limited supervision settings. Experiments on this benchmark demonstrate the effectiveness of our pre-trained models with minimal data and task-specific fine-tuning. Finally, we present several interesting empirical observations about large pre-trained time series models. Pre-trained models (AutonLab/MOMENT-1-large) and Time Series Pile (AutonLab/Timeseries-PILE) are available on Huggingface.

cs.LG↗

arXiv2024

Quasinormal Modes in Noncommutative Schwarzschild black holes

We investigate the quasinormal modes of a massless scalar field in a Schwarzschild black hole, which is deformed due to noncommutative corrections. We introduce the deformed Schwarzschild black hole solution, which depends on the noncommutative parameter $Θ$. We then extract the master equation as a Schrödinger-like equation, giving the explicit expression of the effective potential which is modified due to the noncommutative corrections. After that, we solve the master equation numerically. The significance of these results is twofold. Firstly, our results can be related to the detection of gravitational waves by the near future gravitational wave detectors, such as LISA, which will have a significantly increased accuracy. In particular, these observed gravitational waves produced by binary strong gravitational systems have oscillating modes which can provide valuable information. Secondly, our results can serve as an additional tool to test the predictions of GR, as well as to examine the possible detection of this kind of gravitational corrections.

gr-qc↗

arXiv2022

Enhance Primordial Black Hole Abundance through the Non-linear Processes around Bounce Point

The non-singular bouncing cosmology is an alternative paradigm to inflation, wherein the background energy density vanishes at the bounce point, in the context of Einstein gravity. Therefore, the non-linear effects in the evolution of density fluctuations ($δρ$) may be strong in the bounce phase, which potentially provides a mechanism to enhance the abundance of primordial black holes (PBHs). This article presents a comprehensive illustration for PBH enhancement due to the bounce phase. To calculate the non-linear evolution of $δρ$, the Raychaudhuri equation is numerically solved here. Since the non-linear processes may lead to a non-Gaussian probability distribution function for $δρ$ after the bounce point, the PBH abundance is calculated in a modified Press-Schechter formalism. In this case, the criterion of PBH formation is complicated, due to complicated non-linear evolutionary behavior of $δρ$ during the bounce phase. Our results indicate that the bounce phase indeed has potential to enhance the PBH abundance sufficiently. Furthermore, the PBH abundance is applied to constrain the parameters of bounce phase, providing a complementary to the surveys of cosmic microwave background and large scale structure.

astro-ph.CO↗

arXiv2022

BOSS full-shape analysis from the EFTofLSS with exact time dependence

We re-analyze the full shape of BOSS galaxy two-point function from the Effective-Field Theory of Large-Scale Structure at the one loop within $Λ$CDM with massive neutrinos using a big bang nucleosynthesis (BBN) prior, removing the Einstein-de Sitter (EdS) approximation in the time dependence of the loop, and, properly accounting for the redshift selection over the BOSS samples instead of assuming an effective redshift. We constrain, at $68\%$-confidence level (CL), the present-day matter fraction to $Ω_m=0.322 \pm 0.018$, the Hubble constant to $H_0=69.1\pm 0.14$ (km/s)/Mpc, the $\log$-amplitude of the primordial spectrum to $\ln (10^{10} A_s) = 2.97 \pm 0.25$, the spectral tilt to $n_s = 0.938 \pm 0.082$, and bound the total neutrino mass to $<1.1$ at $95\%$-CL. We find no significant shift in the posteriors of the cosmological parameters due to the EdS approximation, but a marginal difference in $\ln (10^{10} A_s)$ due to the effective redshift approximation of about $0.4σ$, where $σ$ is the $68\%$-confidence interval. Regarding the EdS approximation, we check that the same conclusion holds on simulations of volume like DESI in $Λ$CDM and $w$CDM, with a BBN prior. In contrast, for an approximate, effective redshift, to be assumed, we advocate systematic assessments on redshift selection for ongoing and future large-volume surveys.

astro-ph.CO↗

arXiv2022

BOSS Correlation Function Analysis from the Effective Field Theory of Large-Scale Structure

After calibrating the predictions of the Effective Field Theory of Large-Scale Structure against several sets of simulations, as well as implementing a new method to assert the scale cut of the theory without the use of any simulation, we analyze the Full Shape of the BOSS Correlation Function. Imposing a prior from Big Bang Nucleosynthesis on the baryon density, we are able to measure all the parameters in $Λ$CDM + massive neutrinos in normal hierarchy, except for the total neutrino mass, which is just bounded. When combining the BOSS Full Shape with the Baryon Acoustic Oscillation measurements from BOSS, 6DF/MGS and eBOSS, we determine the present day Hubble constant, $H_0$, the present matter fraction, $Ω_m$, the amplitude of the primordial power spectrum, $A_s$, and the tilt of the primordial power spectrum, $n_s$, to $1.4 \%, 4.5 \%, 23.5\%$ and $7.6\%$ precision, respectively, at $68 \%$-confidence level, finding $H_0=68.19 \pm 0.99$ (km/s)/Mpc, $Ω_m=0.309\pm 0.014$, $\ln (10^{10}A_{s })=3.12^{+0.21}_{-0.26}$ and $n_s=0.963^{+0.062}_{-0.085}$, and we bound the total neutrino mass to $0.87 \, \textrm{eV}$ at $95 \%$-confidence level. These constraints are fully consistent with Planck results and the ones obtained from BOSS power spectrum analysis. In particular, we find no tension in $H_0$ or $σ_8$ with Planck measurements, finding consistency at $1.2σ$ and $0.6σ$, respectively.

astro-ph.CO↗