arXiv subjects

Nikolay Nikolov

Publications and source records attributed to Nikolay Nikolov.

At least 19 recordsLinked to original sources

arXiv2026

AR-VLA: True Autoregressive Action Expert for Vision-Language-Action Models

We propose a standalone autoregressive (AR) Action Expert that generates actions as a continuous causal sequence while conditioning on refreshable vision-language prefixes. In contrast to existing Vision-Language-Action (VLA) models and diffusion policies that reset temporal context with each new observation and predict actions reactively, our Action Expert maintains its own history through a long-lived memory and is inherently context-aware. This structure addresses the frequency mismatch between fast control and slow reasoning, enabling efficient independent pretraining of kinematic syntax and modular integration with heavy perception backbones, naturally ensuring spatio-temporally consistent action generation across frames. To synchronize these asynchronous hybrid V-L-A modalities, we utilize a re-anchoring mechanism that mathematically accounts for perception staleness during both training and inference. Experiments on simulated and real-robot manipulation tasks demonstrate that the proposed method can effectively replace traditional chunk-based action heads for both specialist and generalist policies. AR-VLA exhibits superior history awareness and substantially smoother action trajectories while maintaining or exceeding the task success rates of state-of-the-art reactive VLAs. Overall, our work introduces a scalable, context-aware action generation schema that provides a robust structural foundation for training effective robotic policies. Code and Videos available at https://arvla.insait.ai

cs.RO↗

arXiv2026

Charge Migration and Residual Non-Linearity in NIRSpec BOTS Observations

We investigate the effect of charge migration and residual non-linearity on the JWST/NIRSpec G395H NRS1 and NRS2 detectors using Bright Object Time Series (BOTS) observations of the ultra-hot Jupiter WASP-121b. These full-orbit phase curve observations were taken over 37.8 hours (1.57 days), and provide an excellent testbed of the non-linearity behavior of NRS1 and NRS2 over long timescales. For both detectors, our analysis demonstrates charge losses at the center of the spectral trace and charge excesses at the trace edges. We find that the NRS1 detector displays ~3x larger deviations from linearity compared to NRS2. Given the large transit (~1.5%) and eclipse (~0.5%) signals for WASP-121b, we also investigate variations in the distribution of charge throughout the time-series observation. Our results show that charge distribution varies at different planetary orbital phases for NRS1, which manifests as a change in the morphology and shape of the spectral trace over the course of the time-series. The effect of charge distribution on the trace shape is not evident for NRS2.

astro-ph.IM↗

arXiv2025

The Profinite Rigidity of Torsion-Free Lamplighter Groups

We prove that the torsion-free lamplighter group $\Gamma = \mathbb{Z}^n \wr \mathbb{Z}$ of any rank $n \in \mathbb{N}$ is profinitely rigid in the absolute sense: the finite quotients of $\Gamma$ determine its isomorphism type uniquely among all finitely generated residually finite groups. The proof combines the theory of profinite rigidity for modules over Noetherian domains with an analysis of the algebraic properties of the lower central series of groups with the same profinite completion as $\Gamma$.

math.GR↗

arXiv2025

SPEAR-1: Scaling Beyond Robot Demonstrations via 3D Understanding

Robotic Foundation Models (RFMs) hold great promise as generalist, end-to-end systems for robot control. Yet their ability to generalize across new environments, tasks, and embodiments remains limited. We argue that a major bottleneck lies in their foundations: most RFMs are built by fine-tuning internet-pretrained Vision-Language Models (VLMs). However, these VLMs are trained on 2D image-language tasks and lack the 3D spatial reasoning inherently required for embodied control in the 3D world. Bridging this gap directly with large-scale robotic data is costly and difficult to scale. Instead, we propose to enrich easy-to-collect non-robotic image data with 3D annotations and enhance a pretrained VLM with 3D understanding capabilities. Following this strategy, we train SPEAR-VLM, a 3D-aware VLM that infers object coordinates in 3D space from a single 2D image. Building on SPEAR-VLM, we introduce our main contribution, $~\textbf{SPEAR-1}$: a robotic foundation model that integrates grounded 3D perception with language-instructed embodied control. Trained on $\sim$45M frames from 24 Open X-Embodiment datasets, SPEAR-1 outperforms or matches state-of-the-art models such as $\pi_0$-FAST and $\pi_{0.5}$, while it uses 20$\times$ fewer robot demonstrations. This carefully-engineered training strategy unlocks new VLM capabilities and as a consequence boosts the reliability of embodied control beyond what is achievable with only robotic data. We make our model weights and 3D-annotated datasets publicly available at https://spear.insait.ai.

cs.RO↗

arXiv2025

Testing the performance of cross-correlation techniques to search for molecular features in JWST NIRSpec G395H observations of transiting exoplanets

Cross-correlations techniques offer an alternative method to search for molecular species in JWST observations of exoplanet atmospheres. In a previous article, we applied cross-correlation functions for the first time to JWST NIRSpec/G395H observations of exoplanet atmospheres, resulting in a detection of CO in the transmission spectrum of WASP-39b and a tentative detection of CO isotopologues. Here we present an improved version of our cross-correlation technique and an investigation into how efficient the technique is when searching for other molecules in JWST NIRSpec/G395H data. Our search results in the detection of more molecules via cross-correlations in the atmosphere of WASP-39b, including $\rm H_{2}O$ and $\rm CO_{2}$, and confirms the CO detection. This result proves that cross-correlations are a robust and computationally cheap alternative method to search for molecular species in transmission spectra observed with JWST. We also searched for other molecules ($\rm CH_{4}$, $\rm NH_{3}$, $\rm SO_{2}$, $\rm N_{2}O$, $\rm H_{2}S$, $\rm PH_{3}$, $\rm O_{3}$ and $\rm C_{2}H_{2}$) that were not detected, for which we provide the definition of their cross-correlation baselines for future searches of those molecules in other targets. We find that that the cross-correlation search of each molecule is more efficient over limited wavelength regions of the spectrum, where the signal for that molecule dominates over other molecules, than over broad wavelength ranges. In general we also find that Gaussian normalization is the most efficient normalization mode for the generation of the molecular templates.

astro-ph.EP↗

arXiv2025

Generalist Robot Manipulation beyond Action Labeled Data

Recent advances in generalist robot manipulation leverage pre-trained Vision-Language Models (VLMs) and large-scale robot demonstrations to tackle diverse tasks in a zero-shot manner. A key challenge remains: scaling high-quality, action-labeled robot demonstration data, which existing methods rely on for robustness and generalization. To address this, we propose a method that benefits from videos without action labels - featuring humans and/or robots in action - enhancing open-vocabulary performance and enabling data-efficient learning of new tasks. Our method extracts dense, dynamic 3D point clouds at the hand or gripper location and uses a proposed 3D dynamics predictor for self-supervision. This predictor is then tuned to an action predictor using a smaller labeled dataset for action alignment. We show that our method not only learns from unlabeled human and robot demonstrations - improving downstream generalist robot policies - but also enables robots to learn new tasks without action labels (i.e., out-of-action generalization) in both real-world and simulated settings.

cs.RO↗

arXiv2025

ADApt: Edge Device Anomaly Detection and Microservice Replica Prediction

The increased usage of Internet of Things devices at the network edge and the proliferation of microservice-based applications create new orchestration challenges in Edge computing. These include detecting overutilized resources and scaling out overloaded microservices in response to surging requests. This work presents ADApt, an extension of the ADA-PIPE tool developed in the DataCloud project, by monitoring Edge devices, detecting the utilization-based anomalies of processor or memory, investigating the scalability in microservices, and adapting the application executions. To reduce the overutilization bottleneck, we first explore monitored devices executing microservices over various time slots, detecting overutilization-based processing events, and scoring them. Thereafter, based on the memory requirements, ADApt predicts the processing requirements of the microservices and estimates the number of replicas running on the overutilized devices. The prediction results show that the gradient boosting regression-based replica prediction reduces the MAE, MAPE, and RMSE compared to others. Moreover, ADApt can estimate the number of replicas close to the actual data and reduce the CPU utilization of the device by 14%-28%.

cs.DC↗

arXiv2025

Refractory and Volatile Species in the UV-to-IR Transmission Spectrum of Ultra-hot Jupiter WASP-178b with HST and JWST

The atmospheres of ultra-hot Jupiters are unique compared to other planets because of the presence of both refractory and volatile gaseous species, enabling a new lens to constrain a planet's composition, chemistry, and formation. WASP-178b is one such ultra-hot Jupiter that was recently found to exhibit enormous NUV absorption between 0.2 and 0.4 $\mu$m from some combination of Fe+, Mg, and SiO. Here, we present new infrared observations of WASP-178b with HST/WFC3 and JWST/NIRSpec/G395H, providing novel measurements of the volatile species H$_2$O and CO in WASP-178b's atmosphere. Atmospheric retrievals find a range of compositional interpretations depending on which dataset is retrieved, the type of chemistry assumed, and the temperature structure parametrization used due to the combined effects of thermal dissociation, the lack of volatile spectral features besides H$_2$O and CO, and the relative weakness of H$_2$O and CO themselves. Taken together with a new state-of-the-art characterization of the host star, our retrieval analyses suggests a solar to super-solar [O/H] and [Si/H], but sub-solar [C/H], perhaps suggesting rock-laden atmospheric enrichment near the H$_2$O iceline. To obtain meaningful abundance constraints for this planet, it was essential to combine the JWST IR data with short-wavelength HST observations, highlighting the ongoing synergy between the two facilities.

astro-ph.EP↗

arXiv2025

Jordan-Holder Theorem for profinite groups and applications

We generalize the notions of composition series and composition factors for profinite groups, and prove a profinite version of the Jordan-Holder Theorem. We apply this to prove a Galois Theorem for infinite prosolvable extensions. In addition, we investigate the connection between the abstract and topological composition factors of a nonstrongly complete profinite group.

math.GR↗

arXiv2024

Krull-Schmidt Theorem for small profinite groups

We prove that every small profinite group can be decomposed into a direct product of indecomposable profinite groups, and that such a decomposition is unique up to order and isomorphisms of the components. We also investigate the cancellation property of some free pro-$\mathcal{C}$ groups, and give a new criterion for a profinite group to be small.

math.GR↗

arXiv2024

ReVLA: Reverting Visual Domain Limitation of Robotic Foundation Models

Recent progress in large language models and access to large-scale robotic datasets has sparked a paradigm shift in robotics models transforming them into generalists able to adapt to various tasks, scenes, and robot modalities. A large step for the community are open Vision Language Action models which showcase strong performance in a wide variety of tasks. In this work, we study the visual generalization capabilities of three existing robotic foundation models, and propose a corresponding evaluation framework. Our study shows that the existing models do not exhibit robustness to visual out-of-domain scenarios. This is potentially caused by limited variations in the training data and/or catastrophic forgetting, leading to domain limitations in the vision foundation models. We further explore OpenVLA, which uses two pre-trained vision foundation models and is, therefore, expected to generalize to out-of-domain experiments. However, we showcase catastrophic forgetting by DINO-v2 in OpenVLA through its failure to fulfill the task of depth regression. To overcome the aforementioned issue of visual catastrophic forgetting, we propose a gradual backbone reversal approach founded on model merging. This enables OpenVLA -- which requires the adaptation of the visual backbones during initial training -- to regain its visual generalization ability. Regaining this capability enables our ReVLA model to improve over OpenVLA by a factor of 77\% and 66\% for grasping and lifting in visual OOD tasks. Comprehensive evaluations, episode rollouts and model weights are available on the ReVLA Page

cs.CV↗

arXiv2024

Debris Disks can Contaminate Mid-Infrared Exoplanet Spectra: Evidence for a Circumstellar Debris Disk around Exoplanet Host WASP-39

The signal from a transiting planet can be diluted by astrophysical contamination. In the case of circumstellar debris disks, this contamination could start in the mid-infrared and vary as a function of wavelength, which would then change the observed transmission spectrum for any planet in the system. The MIRI/LRS WASP-39b transmission spectrum shows an unexplained dip starting at $\sim$10 $\mu$m that could be caused by astrophysical contamination. The spectral energy distribution displays excess flux at similar levels to that which are needed to create the dip in the transmission spectrum. In this article, we show that this dip is consistent with the presence of a bright circumstellar debris disk, at a distance of $>$2 au. We discuss how a circumstellar debris disk like that could affect the atmosphere of WASP-39b. We also show that even faint debris disks can be a source of contamination in MIRI exoplanet spectra.

astro-ph.EP↗

arXiv2023

Groups of profinite type and profinite rigidity

We say that a group $G$ is of \textit{profinite type} if it can be realized as a Galois group of some field extension. Using Krull's theory, this is equivalent to the ability of $G$ to be equipped with a profinite topology. We also say that a group of profinite type is \textit{profinitely rigid} if it admits a unique profinite topology. In this paper we study when abelian groups and some group extensions are of profinite type or profinitely rigid. We also discuss the connection between the properties of profinite type and profinite rigidity to the injectivity and surjectivity of the cohomology comparison maps, which were studied by Sury and other authors.

math.GR↗

arXiv2023

Comparison of Microservice Call Rate Predictions for Replication in the Cloud

Today, many users deploy their microservice-based applications with various interconnections on a cluster of Cloud machines, subject to stochastic changes due to dynamic user requirements. To address this problem, we compare three machine learning (ML) models for predicting the microservice call rates based on the microservice times and aiming at estimating the scalability requirements. We apply the linear regression (LR), multilayer perception (MLP), and gradient boosting regression (GBR) models on the Alibaba microservice traces. The prediction results reveal that the LR model reaches a lower training time than the GBR and MLP models. However, the GBR reduces the mean absolute error and the mean absolute percentage error compared to LR and MLP models. Moreover, the prediction results show that the required number of replicas for each microservice by the gradient boosting model is close to the actual test data without any prediction.

cs.DC↗

arXiv2023

On the finite images of finitely generated perfect groups

Let $d \geq 2$ be an integer. We conjecture that there is a finitely generated perfect group whose homomorphic images include all finite $d$-generated perfect groups. We prove a special case of this conjecture for the finite perfect groups with a subnormal series of bounded length and factors which are abelian or semisimple.

math.GR↗

arXiv2023

Demushkin groups of uncountable rank

We extend the theory of countably generated Demushkin groups to Demushkin groups of arbitrary rank. We investigate their algebraic properties and invariants, count their isomorphism classes and study their realization as absolute Galois group. At the end, we compute their profinite completion and conclude with some results on profinite completion of absolute Galois groups.

math.NT↗

arXiv2023

Scaling Data Science Solutions with Semantics and Machine Learning: Bosch Case

Industry 4.0 and Internet of Things (IoT) technologies unlock unprecedented amount of data from factory production, posing big data challenges in volume and variety. In that context, distributed computing solutions such as cloud systems are leveraged to parallelise the data processing and reduce computation time. As the cloud systems become increasingly popular, there is increased demand that more users that were originally not cloud experts (such as data scientists, domain experts) deploy their solutions on the cloud systems. However, it is non-trivial to address both the high demand for cloud system users and the excessive time required to train them. To this end, we propose SemCloud, a semantics-enhanced cloud system, that couples cloud system with semantic technologies and machine learning. SemCloud relies on domain ontologies and mappings for data integration, and parallelises the semantic data integration and data analysis on distributed computing nodes. Furthermore, SemCloud adopts adaptive Datalog rules and machine learning for automated resource configuration, allowing non-cloud experts to use the cloud system. The system has been evaluated in industrial use case with millions of data, thousands of repeated runs, and domain users, showing promising results.

cs.AI↗

arXiv2022

Spectroscopic Time-series Performance of JWST/NIRSpec from Commissioning Observations

We report on JWST commissioning observations of the transiting exoplanet HAT-P-14 b, obtained using the Bright Object Time Series (BOTS) mode of the NIRSpec instrument with the G395H/F290LP grating/filter combination ($3-5\mu$m). While the data were used primarily to verify that the NIRSpec BOTS mode is working as expected, and to enable it for general scientific use, they yield a precise transmission spectrum which we find is featureless down to the precision level of the instrument, consistent with expectations given HAT-P-14~b's small scale-height and hence expected atmospheric features. The exquisite quality and stability of the \emph{JWST/NIRSpec} transit spectrum -- almost devoid of any systematic effects -- allowed us to obtain median uncertainties of 50-60 ppm in this wavelength range at a resolution of $R=100$ in a single exposure, which is in excellent agreement with pre-flight expectations and close to the (or at the) photon-noise limit for a $J = 9.094$, F-type star like HAT-P-14. These observations showcase the ability of NIRSpec/BOTS to perform cutting-edge transiting exoplanet atmospheric science, setting the stage for observations and discoveries to be made in Cycle 1 and beyond.

astro-ph.EP↗