arXiv subjects

Saurabh Jha

Publications and source records attributed to Saurabh Jha.

At least 37 records · Page 2Linked to original sources

arXiv2025

The Most Luminous Known Fast Blue Optical Transient AT 2024wpp: Unprecedented Evolution and Properties in the Ultraviolet to the Near-Infrared

We present an extensive photometric and spectroscopic ultraviolet-optical-infrared campaign on the luminous fast blue optical transient (LFBOT) AT 2024wpp over the first ~100 d. AT 2024wpp is the most luminous LFBOT discovered to date, with $L_{\rm{pk}}\approx(2-4)\times10^{45}$ erg s$^{-1}$ (5-10 times that of the prototypical AT 2018cow). This extreme luminosity enabled the acquisition of the most detailed LFBOT UV light curve thus far. In the first ~45 d, AT 2024wpp radiated $>10^{51}$ erg, surpassing AT 2018cow by an order of magnitude and requiring a power source beyond the radioactive $^{56}$Ni decay of traditional supernovae. Like AT 2018cow, the UV-optical spectrum of AT 2024wpp is dominated by a persistently blue thermal continuum throughout our monitoring, with blackbody parameters at peak of T>30,000 K and $R_{\rm{BB}}/t\approx0.2-0.3c$. A temperature of $\gtrsim$20,000 K is maintained thereafter without evidence for cooling. We interpret the featureless spectra as a consequence of continuous energy injection from a central source of high-energy emission which maintains high ejecta ionization. After 35 d, faint (equivalent width <10 Å) H and He spectral features with kinematically separate velocity components centered at 0 km s$^{-1}$ and -6400 km s$^{-1}$ emerge, implying spherical symmetry deviations. A near-infrared excess of emission above the optical blackbody emerges between 20-30 d with a power-law spectrum $F_{\rmν,NIR}\proptoν^{-0.3}$ at 30 d. We interpret this distinct emission component as either reprocessing of early UV emission in a dust echo or free-free emission in an extended medium above the optical photosphere. LFBOT asphericity and multiple outflow components (including mildly relativistic ejecta) together with the large radiated energy are naturally realized by super-Eddington accretion disks around neutron stars or black holes and their outflows.

astro-ph.HE↗

arXiv2025

CogniSQL-R1-Zero: Lightweight Reinforced Reasoning for Efficient SQL Generation

Translating natural language into SQL (Text-to-SQL) remains a core challenge at the intersection of language understanding and structured data access. Although large language models (LLMs) have improved fluency, generating correct and executable SQL, especially for complex queries, continues to be challenging. We introduce CogniSQL-R1-Zero, a reinforcement learning (RL) framework and model that produces accurate SQL using a lightweight reward signal based on execution correctness and format-tag compliance. By avoiding intermediate supervision, hybrid pipelines and complex reward shaping, our method encourages stable learning and stronger alignment with the ultimate task objective-producing executable programs. CogniSQL-R1-Zero achieves state-of-the-art execution accuracy on Text2SQL benchmark; BIRD bench, outperforming prior supervised and instruction-tuned baselines including SFT CodeS-7B, DeepSeek-Coder 236B, and Mistral 123B-despite being trained on a significantly smaller 7B backbone. This result underscores the scalability and efficiency of our RL-based approach when trained on just four NVIDIA A100 GPUs (40 GB VRAM each). To support further research in efficient and interpretable Text-to-SQL modeling, we release two curated datasets: (i) a collection of 5,024 reasoning traces with varying context lengths, and (ii) a positive-sampled corpus of 36,356 corpus of weakly supervised queries, each annotated with six semantically diverse reasoning paths. Together, these contributions advance scalable, execution-aligned Text-to-SQL generation.

cs.AI↗

arXiv2025

Rapid follow-up observations of infant supernovae with the Gran Telescopio Canarias

The first few hours of a supernova (SN) contain significant information about the progenitor system. The most modern wide-field surveys that scan the sky repeatedly every few days can discover all kinds of transients in those early epochs. At such times, some progenitor footprints may be visible, elucidating critical explosion parameters and helping to distinguish between leading explosion models. A dedicated spectroscopic classification programme using the optical spectrograph OSIRIS mounted on the Gran Telescopio Canarias was set up to try to obtain observations of supernovae (SNe) at those early epochs. With the time awarded, we obtained spectra of 10 SN candidates, which we present here. Half of them were thermonuclear SNe, while the other half were core-collapse SNe. Most (70\%) were observed within the first six days of the estimated explosion, with two being captured within the first 48\,hr. We present a characterization of the spectra, together with other public ancillary photometry from the Zwicky Transient Facility (ZTF) and the Asteroid Terrestrial-impact Last Alert System (ATLAS). This project shows the need for an accompanying rapid-response spectroscopic programme for existing and future deep photometric wide-field surveys located at the right longitude to be able to trigger observations in a few hours after the discovery of the SN candidate.

astro-ph.SR↗

arXiv2025

INDIGO: Page Migration for Hardware Memory Disaggregation Across a Network

Hardware memory disaggregation (HMD) is an emerging technology that enables access to remote memory, thereby creating expansive memory pools and reducing memory underutilization in datacenters. However, a significant challenge arises when accessing remote memory over a network: increased contention that can lead to severe application performance degradation. To reduce the performance penalty of using remote memory, the operating system uses page migration to promote frequently accessed pages closer to the processor. However, previously proposed page migration mechanisms do not achieve the best performance in HMD systems because of obliviousness to variable page transfer costs that occur due to network contention. To address these limitations, we present INDIGO: a network-aware page migration framework that uses novel page telemetry and a learning-based approach for network adaptation. We implemented INDIGO in the Linux kernel and evaluated it with common cloud and HPC applications on a real disaggregated memory system prototype. Our evaluation shows that INDIGO offers up to 50-70% improvement in application performance compared to other state-of-the-art page migration policies and reduces network traffic up to 2x.

cs.DC↗

arXiv2025

Causal AI-based Root Cause Identification: Research to Practice at Scale

Modern applications are built as large, distributed systems spanning numerous modules, teams, and data centers. Despite robust engineering and recovery strategies, failures and performance issues remain inevitable, risking significant disruptions and affecting end users. Rapid and accurate root cause identification is therefore vital to ensure system reliability and maintain key service metrics. We have developed a novel causality-based Root Cause Identification (RCI) algorithm that emphasizes causation over correlation. This algorithm has been integrated into IBM Instana-bridging research to practice at scale-and is now in production use by enterprise customers. By leveraging "causal AI," Instana stands apart from typical Application Performance Management (APM) tools, pinpointing issues in near real-time. This paper highlights Instana's advanced failure diagnosis capabilities, discussing both the theoretical underpinnings and practical implementations of the RCI algorithm. Real-world examples illustrate how our causality-based approach enhances reliability and performance in today's complex system landscapes.

cs.LG↗

arXiv2025

Queue management for slo-oriented large language model serving

Large language model (LLM) serving is becoming an increasingly critical workload for cloud providers. Existing LLM serving systems focus on interactive requests, such as chatbots and coding assistants, with tight latency SLO requirements. However, when such systems execute batch requests that have relaxed SLOs along with interactive requests, it leads to poor multiplexing and inefficient resource utilization. To address these challenges, we propose QLM, a queue management system for LLM serving. QLM maintains batch and interactive requests across different models and SLOs in a request queue. Optimal ordering of the request queue is critical to maintain SLOs while ensuring high resource utilization. To generate this optimal ordering, QLM uses a Request Waiting Time (RWT) Estimator that estimates the waiting times for requests in the request queue. These estimates are used by a global scheduler to orchestrate LLM Serving Operations (LSOs) such as request pulling, request eviction, load balancing, and model swapping. Evaluation on heterogeneous GPU devices and models with real-world LLM serving dataset shows that QLM improves SLO attainment by 40-90% and throughput by 20-400% while maintaining or improving device utilization compared to other state-of-the-art LLM serving systems. QLM's evaluation is based on the production requirements of a cloud provider. QLM is publicly available at https://www.github.com/QLM-project/QLM.

cs.DC↗

arXiv2025

ITBench: Evaluating AI Agents across Diverse Real-World IT Automation Tasks

Realizing the vision of using AI agents to automate critical IT tasks depends on the ability to measure and understand effectiveness of proposed solutions. We introduce ITBench, a framework that offers a systematic methodology for benchmarking AI agents to address real-world IT automation tasks. Our initial release targets three key areas: Site Reliability Engineering (SRE), Compliance and Security Operations (CISO), and Financial Operations (FinOps). The design enables AI researchers to understand the challenges and opportunities of AI agents for IT automation with push-button workflows and interpretable metrics. ITBench includes an initial set of 94 real-world scenarios, which can be easily extended by community contributions. Our results show that agents powered by state-of-the-art models resolve only 13.8% of SRE scenarios, 25.2% of CISO scenarios, and 0% of FinOps scenarios. We expect ITBench to be a key enabler of AI-driven IT automation that is correct, safe, and fast.

cs.AI↗

arXiv2025

Hierarchical Autoscaling for Large Language Model Serving with Chiron

Large language model (LLM) serving is becoming an increasingly important workload for cloud providers. Based on performance SLO requirements, LLM inference requests can be divided into (a) interactive requests that have tight SLOs in the order of seconds, and (b) batch requests that have relaxed SLO in the order of minutes to hours. These SLOs can degrade based on the arrival rates, multiplexing, and configuration parameters, thus necessitating the use of resource autoscaling on serving instances and their batch sizes. However, previous autoscalers for LLM serving do not consider request SLOs leading to unnecessary scaling and resource under-utilization. To address these limitations, we introduce Chiron, an autoscaler that uses the idea of hierarchical backpressure estimated using queue size, utilization, and SLOs. Our experiments show that Chiron achieves up to 90% higher SLO attainment and improves GPU efficiency by up to 70% compared to existing solutions.

cs.DC↗

arXiv2024

Efficient Interactive LLM Serving with Proxy Model-based Sequence Length Prediction

Large language models (LLMs) have been driving a new wave of interactive AI applications across numerous domains. However, efficiently serving LLM inference requests is challenging due to their unpredictable execution times originating from the autoregressive nature of generative models. Existing LLM serving systems exploit first-come-first-serve (FCFS) scheduling, suffering from head-of-line blocking issues. To address the non-deterministic nature of LLMs and enable efficient interactive LLM serving, we present a speculative shortest-job-first (SSJF) scheduler that uses a light proxy model to predict LLM output sequence lengths. Our open-source SSJF implementation does not require changes to memory management or batching strategies. Evaluations on real-world datasets and production workload traces show that SSJF reduces average job completion times by 30.5-39.6% and increases throughput by 2.2-3.6x compared to FCFS schedulers, across no batching, dynamic batching, and continuous batching settings.

cs.DC↗

arXiv2024

JWST Validates HST Distance Measurements: Selection of Supernova Subsample Explains Differences in JWST Estimates of Local H0

JWST provides new opportunities to cross-check the HST Cepheid/SNeIa distance ladder, which yields the most precise local measure of H0. We analyze early JWST subsamples (~1/4 of the HST sample) from the SH0ES and CCHP groups, calibrated by a single anchor (N4258). We find HST Cepheid distances agree well (~1 sigma) with all 8 combinations of methods, samples, and telescopes: JWST Cepheids, TRGB, and JAGB by either group, plus HST TRGB and Miras. The comparisons explicitly include the measurement uncertainty of each method in N4258, an oft-neglected but dominant term. Mean differences are ~0.03 mag, far smaller than the 0.18 mag "Hubble tension." Combining all measures produces the strongest constraint yet on the linearity of HST Cepheid distances, 0.994+-0.010, ruling out distance-dependent bias or offset as the source of the tension at ~7 sigma. Yet, measurements of H0 from current JWST subsamples produce large sampling differences whose size and direction we can directly estimate from the full HST set. We show that Delta(H0)~2.5 km/s/Mpc between the CCHP JWST program and the full HST sample is entirely consistent with differences in sample selection. Combining all JWST samples produces a new, distance-limited set of 16 SNeIa at D<25 Mpc and more closely resembles the full sample thanks to "reversion to the mean" of larger samples. Using JWST Cepheids, JAGB, and TRGB, we find 73.4+-2.1, 72.2+-2.2, and 72.1+-2.2 km/s/Mpc, respectively. Explicitly accounting for SNe in common, the combined-sample three-method result from JWST is H0=72.6+-2.0, similar to H0=72.8 expected from HST Cepheids in the same galaxies. The small JWST sample trivially lowers the Hubble tension significance due to small-sample statistics and is not yet competitive with the HST set (42 SNeIa and 4 anchors), which yields 73.2+-0.9. Still, the joint JWST sample provides important crosschecks which the HST data passes.

astro-ph.CO↗

arXiv2023

Evolution of the Mass-Metallicity Relation from Redshift $z\approx8$ to the Local Universe

A tight positive correlation between the stellar mass and the gas-phase metallicity of galaxies has been observed at low redshifts. The redshift evolution of this correlation can strongly constrain theories of galaxy evolution. The advent of JWST allows probing the mass-metallicity relation at redshifts far beyond what was previously accessible. Here we report the discovery of two emission-line galaxies at redshifts 8.15 and 8.16 in JWST NIRCam imaging and NIRSpec spectroscopy of targets gravitationally lensed by the cluster RXJ2129.4$+$0005. We measure their metallicities and stellar masses along with nine additional galaxies at $7.2 < z_{\rm spec} < 9.5$ to report the first quantitative statistical inference of the mass-metallicity relation at $z\approx8$. We measure $\sim 0.9$ dex evolution in the normalization of the mass-metallicity relation from $z \approx 8$ to the local Universe; at fixed stellar mass, galaxies are 8 times less metal enriched at $z \approx 8$ compared to the present day. Our inferred normalization is in agreement with the predictions of the FIRE simulations. Our inferred slope of the mass-metallicity relation is similar to or slightly shallower than that predicted by FIRE or observed at lower redshifts. We compare the $z \approx 8$ galaxies to extremely low metallicity analog candidates in the local Universe, finding that they are generally distinct from extreme emission-line galaxies or "green peas" but are similar in strong emission-line ratios and metallicities to "blueberry galaxies". Despite this similarity, at fixed stellar mass, the $z \approx 8$ galaxies have systematically lower metallicities compared to blueberry galaxies.

astro-ph.GA↗

arXiv2023

Constraints on the Hubble constant from Supernova Refsdal's reappearance

The gravitationally lensed Supernova Refsdal appeared in multiple images, produced through gravitational lensing by a massive foreground galaxy cluster. After the supernova appeared in 2014, lens models of the galaxy cluster predicted an additional image of the supernova would appear in 2015, which was subsequently observed. We use the time delays between the images to perform a blinded measurement of the expansion rate of the Universe, quantified by the Hubble constant (H0). Using eight cluster lens models, we infer H0 = 64.8 +4.4-4.3 km / s / Mpc, where Mpc is the megaparsec. Using the two models most consistent with the observations, we find H0 = 66.6 +4.1-3.3 km / s / Mpc. The observations are best reproduced by models that assign dark-matter halos to individual galaxies and the overall cluster.

astro-ph.CO↗

arXiv2023

A Magnified Compact Galaxy at Redshift 9.51 with Strong Nebular Emission Lines

Ultraviolet light from early galaxies is thought to have ionized gas in the intergalactic medium. However, there are few observational constraints on this epoch because of the faintness of those galaxies and the redshift of their optical light into the infrared. We report the observation, in JWST imaging, of a distant galaxy that is magnified by gravitational lensing. JWST spectroscopy of the galaxy, at rest-frame optical wavelengths, detects strong nebular emission lines that are attributable to oxygen and hydrogen. The measured redshift is z = 9.51 +- 0.01, corresponding to 510 million years after the Big Bang. The galaxy has a radius of 16.2+4.6-7.2 parsecs, which is substantially more compact than galaxies with equivalent luminosity at z = 6 to 8, leading to a high star formation rate surface density.

astro-ph.GA↗

arXiv2023

The Magnificent Five Images of Supernova Refsdal: Time Delay and Magnification Measurements

In late 2014, four images of Supernova (SN) "Refsdal," the first known example of a strongly lensed SN with multiple resolved images, were detected in the MACS J1149 galaxy-cluster field. Following the images' discovery, the SN was predicted to reappear within hundreds of days at a new position ~8 arcseconds away in the field. The observed reappearance in late 2015 makes it possible to carry out Refsdal's (1964) original proposal to use a multiply imaged SN to measure the Hubble constant H0, since the time delay between appearances should vary inversely with H0. Moreover, the position, brightness, and timing of the reappearance enable a novel test of the blind predictions of galaxy-cluster models, which are typically constrained only by the positions of multiply imaged galaxies. We have developed a new photometry pipeline that uses DOLPHOT to measure the fluxes of the five images of SN Refsdal from difference images. We apply four separate techniques to perform a blind measurement of the relative time delays and magnification ratios (mu_i/mu_1) between the last image SX and the earlier images S1-S4. We measure the relative time delay of SX-S1 to be 376.0+5.6-5.5 days and the relative magnification to be 0.30+0.05-0.03. This corresponds to a 1.5% precision on the time delay and 17% precision for the magnification ratios, and includes uncertainties due to millilensing and microlensing. In an accompanying paper, we place initial and blind constraints on the value of the Hubble constant.

astro-ph.CO↗

arXiv2023

A search for transients in the Reionization Lensing Cluster Survey (RELICS): Three new supernovae

The Reionization Cluster Survey (RELICS) imaged 41 galaxy clusters with the Hubble Space Telescope (HST), in order to detect lensed and high-redshift galaxies. Each cluster was imaged to about 26.5 AB mag in three optical and four near-infrared bands, taken in two distinct visits separated by varying time intervals. We make use of the multiple near-infrared epochs to search for transient sources in the cluster fields, with the primary motivation of building statistics for bright caustic crossing events in gravitational arcs. Over the whole sample, we do not find any significant ($\gtrsim5 σ$) caustic crossing events, in line with expectations from semi-analytic calculations but in contrast to what may be naively expected from previous detections of some bright events, or from deeper transient surveys that do find high rates of such events. Nevertheless, we find six prominent supernova (SN) candidates over the 41 fields: three of them were previously reported and three are new ones reported here for the first time. Out of the six candidates, four are likely core-collapse (CC) SNe -- three in cluster galaxies, and among which only one was known before, and one slightly behind the cluster at $z\sim0.6-0.7$. The other two are likely Ia -- both of them previously known, one probably in a cluster galaxy, and one behind it at $z\simeq2$. Our study supplies empirical bounds for the rate of caustic crossing events in galaxy cluster fields to typical HST magnitudes, and lays the groundwork for a future SN rate study.

astro-ph.GA↗

arXiv2022

SN 2016dsg: A Thermonuclear Explosion Involving A Thick Helium Shell

A thermonuclear explosion triggered by a helium-shell detonation on a carbon-oxygen white dwarf core has been predicted to have strong UV line blanketing at early times due to the iron-group elements produced during helium-shell burning. We present the photometric and spectroscopic observations of SN 2016dsg, a sub-luminous peculiar Type I SN consistent with a thermonuclear explosion involving a thick He shell. With a redshift of 0.04, the $i$-band peak absolute magnitude is derived to be around -17.5. The object is located far away from its host, an early-type galaxy, suggesting it originated from an old stellar population. The spectra collected after the peak are unusually red, show strong UV line blanketing and weak O I $λ$7773 absorption lines, and do not evolve significantly over 30 days. An absorption line around 9700-10500 Åis detected in the near-infrared spectrum and is likely from the unburnt helium in the ejecta. The spectroscopic evolution is consistent with the thermonuclear explosion models for a sub-Chandrasekhar mass white dwarf with a thick helium shell, while the photometric evolution is not well described by existing models.

astro-ph.SR↗

arXiv2022

Watch Out for the Safety-Threatening Actors: Proactively Mitigating Safety Hazards

Despite the successful demonstration of autonomous vehicles (AVs), such as self-driving cars, ensuring AV safety remains a challenging task. Although some actors influence an AV's driving decisions more than others, current approaches pay equal attention to each actor on the road. An actor's influence on the AV's decision can be characterized in terms of its ability to decrease the number of safe navigational choices for the AV. In this work, we propose a safety threat indicator (STI) using counterfactual reasoning to estimate the importance of each actor on the road with respect to its influence on the AV's safety. We use this indicator to (i) characterize the existing real-world datasets to identify rare hazardous scenarios as well as the poor performance of existing controllers in such scenarios; and (ii) design an RL based safety mitigation controller to proactively mitigate the safety hazards those actors pose to the AV. Our approach reduces the accident rate for the state-of-the-art AV agent(s) in rare hazardous scenarios by more than 70%.

cs.RO↗

arXiv2021

Progenitor and Close-In Circumstellar Medium of Type II Supernova 2020fqv from High-Cadence Photometry and Ultra-Rapid UV Spectroscopy

We present observations of SN 2020fqv, a Virgo-cluster Type II core-collapse supernova (CCSN) with a high temporal resolution light curve from the Transiting Exoplanet Survey Satellite (TESS) covering the time of explosion; ultraviolet (UV) spectroscopy from the Hubble Space Telescope (HST) starting 3.3 days post-explosion; ground-based spectroscopic observations starting 1.1~days post-explosion; along with extensive photometric observations. Massive stars have complicated mass-loss histories leading up to their death as CCSNe, creating circumstellar medium (CSM) with which the SNe interact. Observations during the first few days post-explosion can provide important information about the mass-loss rate during the late stages of stellar evolution. Model fits to the quasi-bolometric light curve of SN 2020fqv reveal ~0.23 $M_{\odot}$ of CSM confined within ~1450 $R_{\odot}$ ($10^{14}$ cm) from its progenitor star. Early spectra (<4 days post-explosion), both from HST and ground-based observatories, show emission features from high-ionization metal species from the outer, optically thin part of this CSM. We find that the CSM is consistent with an eruption caused by the injection of $\sim$$5\times 10^{46}$ erg into the stellar envelope $\sim$300 days pre-explosion, potentially from a nuclear burning instability at the onset of oxygen burning. Light-curve fitting, nebular spectroscopy, and pre-explosion \textit{HST} imaging consistently point to a red supergiant (RSG) progenitor with $M_{\rm ZAMS}$$\approx$$13.5$--$15 \, M_{\odot}$, typical for SN~II progenitor stars. This finding demonstrates that a typical RSG, like the progenitor of SN 2020fqv, has a complicated mass-loss history immediately before core collapse.

astro-ph.SR↗