arXiv subjects

Alexey Gavrilov

Publications and source records attributed to Alexey Gavrilov.

5 recordsLinked to original sources

arXiv2026

AquaJEPA: An Action-Conditioned Multimodal JEPA Family for Underwater Robot Dynamics

Underwater robots rely on complementary sensors whose reliability changes abruptly with water visibility and vehicle motion. We introduce AquaJEPA, a sensor-configurable family of action-conditioned joint-embedding predictive models spanning full multimodal, camera-only, sonar-only, and sensor-dropout configurations. Its members share a latent objective and receding-horizon control interface that predict future representations and physical dynamics from camera, forward-looking sonar, proprioception, and thruster commands. Trained from scratch on one hour of action-labelled data, the family is evaluated in Stonefish on 120 fresh paired scenarios spanning unseen layouts, visibility changes, dynamics shifts, and scheduled DVL loss. AquaJEPA-base achieves the strongest aggregate closed-loop performance, improving success over state-only by 12.5 percentage points and reducing final error by 0.189 m; both paired 95% intervals exclude zero. In a separate three-seed evaluation, it reduces paired final error relative to AquaJEPA-S by 0.118 m, with the same direction for every seed. AquaJEPA-robust more than halves prediction error during camera and camera-DVL blackouts. These results show that full multimodal prediction improves over state-only control and the sonar-only family member in this benchmark, while sensor-dropout training provides robustness under sensor loss.

cs.RO↗

arXiv2026

One Future, Every Robot: Label-Efficient Collective-State Prediction with Decentralized JEPA

Decentralized robots often need a common view of what their team is becoming, even though each robot sees different evidence and cannot rely on a central estimate or output-level consensus. We ask whether compatible collective-state predictions can emerge under this constraint. Collective-State JEPA (CS-JEPA) trains every robot to predict the same fixed-width latent future from its own history and bounded neighbor messages, with no agreement loss; predictions and plans are never pooled at deployment. In a fresh independent replication, agreement improves for every seed and every evaluated split. Accuracy improves at the same time, ruling out the uninformative solution in which all robots merely collapse to one prediction: relative to capacity-matched raw-future reconstruction, collective-state error falls by 28.4 percent in distribution and by 64.4 to 75.6 percent under topology and swarm-size shift. Translation-free and crossed-pretraining controls preserve this joint result, while action-conditioned and rigid-body evaluations show that the receiver-local representation supports independent decisions. A shared latent future can therefore align decentralized predictions without consensus training while preserving useful, label-efficient information.

cs.RO↗

arXiv2026

Where Quality Breaks in Compressed Short-Text Generation: Staged Bottleneck Localization

Compressed short-text generators can fail in two different places: the codec may discard information before generation starts, or the latent generator may produce weak codes. Without separating these failure modes, researchers can spend compute improving the wrong component. We study this problem in a controlled 64-to-16 TinyStories case study built from a hierarchical VQ-VAE-2 codec and a masked discrete diffusion generator (MDLM). We use a staged validation protocol that separates codec reconstruction fidelity, latent generation quality, and auxiliary latent diagnostics under one shared external GPT-2 scorer, while reporting complementary semantic metrics for the geometry study. In the tested configuration, codec reconstruction alone raises median external perplexity from 15.17 to 27.36 (+80.4%) and p95 from 25.10 to 98.91 (+294.1%), showing that the dominant quality loss appears before latent generation begins. Under the same scorer, code-space MDLM remains materially stronger than token-space diffusion, reducing mean, median, and p95 by 32.9%, 30.9%, and 36.6%, respectively. Geometry-aware regularization improves local latent proxies but does not improve decoded-text metrics in the available runs. The contribution is methodological rather than algorithmic: the paper presents a reusable staged diagnosis for one concrete pipeline and shows that, in this setting, codec fidelity rather than latent denoising sets the practical quality ceiling.

cs.CL↗

arXiv2022

Reduced neural activity during volatile anesthesia compared to TIVA: evidence from a novel EEG signal processing analysis

Post-operative cognitive decline is a well-known phenomenon and of crucial importance especially in the elderly. General anesthesia can be accomplished by inhalation-based (volatile) or total intravenous anesthesia (TIVA). While their effects on post-operative symptoms have been investigated, little is known about their influence on brain functionalities during the surgery itself. To assess differences 17 patients were divided to receive either volatile anesthesia (n=9), or TIVA (n=8). The level of anesthesia was kept to be equal in both groups. A single bipolar EEG electrode (Neurosteer system) was placed on the participants foreheads. It presented real-time activity and collected their data during the surgery. The dependent variables included frequency bands (delta, theta, alpha, and beta), and three features (VC9, ST4, and A0) previously extracted with the device and provided by Neurosteer. All surgeries were uneventful, and all patients showed bispectral index (BIS) score less than 60. Feature activity under volatile anesthesia (in comparison to TIVA) was significantly lower for the delta, theta and alpha frequency bands and for the three features. Further analysis showed that the largest difference between anesthesia types was for feature A0. The EEG frequency bands and novel brain activity features provide evidence that volatile anesthesia further reduces components of brain activity in comparison to TIVA anesthesia. Specifically, A0, which previously showed a correlation with cognitive decline severity and cognitive load, exhibited the most prominent difference between anesthesia types. Together, this study suggests that measuring brain activity during anesthesia using sensitive features, enables revealing that different anesthesia types may affect brain activity differently, which could affect the recovery from anesthesia, and consequently reduce post-operative cognitive decline

q-bio.NC↗

arXiv2020

A novel heterogeneous structure formed by a single multiblock copolymer chain

We studied structures formed by a single $(AB)_k$ multiblock copolymer chain, in which interaction between A-type beads was purely repulsive, and B-type beads tended to aggregate. We studied how attraction between A-type beads and B-type beads affected the structure of the chain. We discovered formation of an equilibrium globular structure, which had unique heterogeneous checkerboard-like distribution of contact density. Unlike the structures usually formed by a single $(AB)_k$ multiblock copolymer chain, this structure had contact enrichment at the boundaries of A and B blocks. This structure was formed by a multiblock copolymer chain, in which B-type beads could form maximum two reversible bonds either with A-type or B-type beads, A-type beads could form maximum one reversible bond with a B-type bead, and interactions between A-type beads were purely repulsive. Multiblock copolymer chains with this type of intrachain interactions can model structure of chromatin in various organisms.

cond-mat.soft↗