@Techmeme@techhub.socialA deep dive into on-device and data center inference for robots, including a primer on robot models, deployments, supply chains, the "network wall", and more (SemiAnalysis)
https://newsletter.semianalysis.com/p/a-brain-too-big-to-carry-on-device
@burger_jaap@mastodon.social
@arXiv_qbioNC_bot@mastoxiv.pageToward a mechanistic understanding of inference in visual cortex and diffusion models
Zeyu Yun, Alexander Belsten, Dasheng Bi, Zahra Kadkhodaie, Yubei Chen, Bruno A. Olshausen
https://arxiv.org/abs/2607.15693 https://arxiv.org/pdf/2607.15693 https://arxiv.org/html/2607.15693
arXiv:2607.15693v1 Announce Type: new
Abstract: We describe a model of perceptual inference in primary visual cortex (V1) equivalent to a minimal diffusion model whose function can be readily understood from its parameters. The model is based on sparse coding with a non-factorial prior over latent variables in the form of an unconstrained, pairwise interaction matrix, extending standard sparse coding inference to a general recurrent dynamical system. We efficiently train these recurrent dynamics using a denoising score-matching objective and implicit differentiation. After training on natural images, the learned interaction matrix mirrors the structure of horizontal connections in superficial layers of V1 that link neurons of similar orientation tuning. This model exhibits exceptionally good denoising performance, restoring image features such as extended contours amid extreme visual ambiguity, nearly matching the behavior of standard, black-box diffusion architectures in generalization regime. Owing to the model's simplicity, the network's Jacobian can be decomposed directly in terms of the interaction matrix between latent variables, revealing mechanistically how the recurrent dynamics assign high probability over a continuous family of natural structural deformations. Intriguingly, within this circuit, a large fraction of latent variables learn to disconnect from visual input altogether, essentially forming a hierarchical representation that appears to enforce global consistency among image features. Together, the model and results bridge two distinct domains: for neuroscience, it generates concrete, testable hypotheses regarding functional connectivity in recurrent neural circuits during perceptual inference tasks; for machine learning, it elucidates the internal mechanisms learned by diffusion models that allow them to generate infinitely many novel images from a finite training set.
toXiv_bot_toot
@Techmeme@techhub.socialParis-based Arlequin AI, which is developing proprietary models based on topological neural networks, raised a €28M Series A co-led by redalpine and OTB (Tamara Djurickovic/Tech.eu)
https://tech.eu/2026/09/10/arlequin-ai-lands-eur28m-to-sca…
@arXiv_csGR_bot@mastoxiv.pageEVOLVE: Efficient Learned Volume Compression with Variable-Rate Encoding on a Cross-Domain Database
Kaiyuan Tang, Maizhe Yang, Chaoli Wang
https://arxiv.org/abs/2607.18187 https://arxiv.org/pdf/2607.18187 https://arxiv.org/html/2607.18187
arXiv:2607.18187v1 Announce Type: new
Abstract: Large-scale scientific simulations generate volumetric data at rates that far outpace advances in storage and network bandwidth, making effective lossy compression increasingly critical. However, conventional compressors often struggle to preserve fine structural details at high compression ratios (CRs), and implicit neural representations (INRs) require costly per-volume optimization and produce models with fixed CRs. To respond, we present EVOLVE, an autoencoder (AE)-based volume-compression framework that targets high CRs for offline compression, with three key contributions. First, we construct a large-scale cross-domain database of 6,376 volumes from 21 scientific simulations, curated via perceptual hashing to ensure diversity, enabling the optimized model to extract features that generalize across volumes within the covered scientific simulation domains. Second, we reexamine the design space of AE-based compressors and incorporate several macro- and micro-designs into a vanilla AE to develop EVOLVE, which substantially improves the expressive power and compression capability. Third, we develop a learnable gain mechanism with a three-stage training strategy to enable variable-rate encoding, allowing a single model to support continuous CR adjustment at inference time. Experiments on multiple unseen scientific simulation datasets demonstrate that EVOLVE achieves substantially higher CRs than conventional compressors at comparable reconstruction quality, while delivering compression speeds that are orders of magnitude faster than INR-based methods, highlighting its promise as a strong alternative for compressing scientific data. The code, model weights, and results are available on our project page at https://evolve-vis.github.io.
toXiv_bot_toot
@seeingwithsound@mas.toWhat about privacy guarantees? Network connectivity loss? Availability after bankruptcy? Network-adaptive cloud processing for visual neuroprostheses https://arxiv.org/abs/2602.13216 Network-adaptive cloud preprocessing for visual neuroprostheses
@Luc@dresden.networkHübsches Projekt zur Reflexion künstlicher Intelligenz, Deepdive in Prompt Injection:
#artificialIntelligence
@cdarwin@c.imOn April 23, 2026, the White House released a memo warning that Chinese entities were running “industrial-scale” distillation campaigns against American frontier AI models,
leveraging “tens of thousands of proxy accounts” to evade detection.
In February 2026, Anthropic similarly reported on Chinese labs’ coordinated distillation attacks using
“a single proxy network managed more than 20,000 fraudulent accounts”.
Both cases see “proxy”
— the middlemen between …
@arXiv_physicsgeoph_bot@mastoxiv.pageSeismic full waveform inversion via a physics-guided Fourier representation neural network
Gui Chen, Yang Liu, Haoran Zhang, Mi Zhang
https://arxiv.org/abs/2606.30126 https://arxiv.org/pdf/2606.30126 https://arxiv.org/html/2606.30126
arXiv:2606.30126v1 Announce Type: new
Abstract: Accurate subsurface velocity models are essential for seismic imaging, yet conventional full waveform inversion (FWI) often suffers from cycle skipping, noise sensitivity, and reliance on good initial models. We develop a physics-guided Fourier representation neural network (PGFRNN) for unsupervised acoustic FWI and simultaneous-source FWI (SSFWI), which embeds Fourier-transformed seismic data into a latent space and iteratively updates the velocity model using a softplus-approximated log-cosh (SALC) loss and a physics-guided optimizer. Numerical tests on the Overthrust model demonstrate that PGFRNN outperforms conventional L2- and SALC-loss-based FWI methods, achieving higher inversion accuracy and robustness to noise and challenging initial models.
toXiv_bot_toot
@arXiv_physicscompph_bot@mastoxiv.pageLSR-Net: Long-Short-Range Operator Learning for Pattern Dynamics on Manifolds
Qian Serena Hou, Zecheng Gan
https://arxiv.org/abs/2607.00750 https://arxiv.org/pdf/2607.00750 https://arxiv.org/html/2607.00750
arXiv:2607.00750v1 Announce Type: new
Abstract: We propose the Long-Short-Range Neural Network (LSR-Net), an extensible operator-learning framework for predicting pattern dynamics on planar domains, spherical surfaces, and general manifolds. The method decomposes the forward evolution operator into a long-range component, represented by a compact Fourier multiplier constructed via the Sum-of-Exponentials (SOE) approximation, and a short-range component adapted to the underlying geometry and its intrinsic symmetries. For general manifolds represented by irregularly sampled point clouds, the long-range component is implemented by Gaussian gridding onto an auxiliary regular grid, where the Fourier multiplier is efficiently applied in k-space using FFT and the result is interpolated back to the original sample points. We evaluate LSR-Net on several benchmark systems, including the Allen-Cahn, Cahn-Hilliard, Schnakenberg, and Turing systems, over planar domains, spherical surfaces, and a blob-shaped manifold. Numerical results demonstrate that LSR-Net consistently achieves higher accuracy and improved stability compared with baseline operator-learning models. In particular, for Allen-Cahn dynamics on the sphere, the RMSE is reduced by approximately three orders of magnitude compared with the Spherical Fourier Neural Operator (SFNO). Rotation and reflection equivariance tests further confirm that the learned operator is consistent with these geometric transformations. These results indicate that LSR-Net provides an effective and robust approach for learning pattern dynamics on complex geometries.
toXiv_bot_toot
@arXiv_csAR_bot@mastoxiv.pageUni-SFU: Algorithm-HW Co-Design for Universal SFUs via Mixed-Degree Piecewise Approximation
Miao Sun, Yucheng Huang, Mingcong Cao, Jaehyun Park, Partha Pratim Pande, Umit Y. Ogras
https://arxiv.org/abs/2608.11577 https://arxiv.org/pdf/2608.11577 https://arxiv.org/html/2608.11577
arXiv:2608.11577v1 Announce Type: new
Abstract: Nonlinear activation functions are essential to modern deep neural networks (DNNs), but their hardware evaluation places significant pressure on the special-function units (SFUs) of GPUs and custom accelerators. Therefore, piecewise polynomial approximations are commonly used within allowed error bounds to improve computational efficiency. However, existing techniques often approximate each activation function in isolation using fixed-degree polynomials and uniform segments, leading to hardware redundancy and sub-optimal precision. To address these limitations, we present Uni-SFU, an algorithm-hardware co-design framework that jointly optimizes approximation accuracy and silicon area for a diverse set of activation functions. Uni-SFU leverages a joint search across all target functions to assign mixed-degree polynomials to nonuniform segments, guided by an RTL-derived area cost model. This approach identifies a unified hardware configuration to implement the target activation functions under given accuracy constraints. Validated across over 700 neural network variants and three Natural Language Processing (NLP) models, Uni-SFU achieves a superior Mean Squared Error (MSE) below 8.22x10^-8, limiting top-1 accuracy degradation to within 1.02% compared to floating-point baselines. The proposed design occupies only 6,800 um2 in GF 22nm CMOS technology, achieving a superior trade-off between silicon area and system-level accuracy compared to SOTA counterparts.
toXiv_bot_toot
@arXiv_physicsgeoph_bot@mastoxiv.pageTwo kinds of robustness are not the same: disentangling fault tolerance and low-SNR robustness in multi-domain event detection on real data
Isao Kurosawa
https://arxiv.org/abs/2606.29339 https://arxiv.org/pdf/2606.29339 https://arxiv.org/html/2606.29339
arXiv:2606.29339v1 Announce Type: new
Abstract: Reliable event detection underpins induced-seismicity monitoring for Carbon dioxide Capture and Storage (CCS) and geothermal operations, distributed acoustic sensing (DAS), and industrial condition monitoring. In each setting a detector must stay reliable both when sensors fail and when the signal is buried in noise. These two failure modes are routinely conflated, and architectural complexity is often credited with robustness it may not deserve. We assemble a unified binary event-detection benchmark from three physically distinct real sources -- Hi-net seismic waveforms, Utah FORGE 2024 borehole DAS, and MAFAULDA industrial vibration -- each mapped to a common 8-channel, 256-sample representation, and evaluate a fault-tolerant detector (CEPHALON) trained with per-sample sensor-dropout against standard detectors (a 1D convolutional network, a temporal convolutional network, and a compact Transformer) trained with an identical recipe. On clean data every model is near-perfect (AUC ~ 0.99). Under progressive sensor loss, simple models with sensor-dropout are already robust and CEPHALON holds no advantage. Under additive noise, however, CEPHALON degrades far more gracefully: at -2.5 dB its overall AUC is 0.939 versus 0.532-0.572 for the convolutional baselines. Same-architecture ablations isolate the cause: disabling internal redundancy at inference reduces the low-SNR advantage only modestly, whereas removing sensor-dropout training collapses it (0.899 to 0.603 at -5 dB). The training recipe is therefore the dominant cause and parallel redundancy only secondary. We release a complete, numbered, reproducible pipeline so that every figure can be regenerated.
toXiv_bot_toot
@metacurity@infosec.exchangeMore than 20 models of a Chinese-made router available throughout the world ship with a backdoor that could allow access and potential connections to other devices on the network, researchers with cybersecurity firm VulnCheck.
https://www.reuters.com/world/asia-pacific/chinese-made-zbtlink-routers-have-backdoor-researchers-say-2026-08-05/
@burger_jaap@mastodon.socialFurthermore, by restricting the definition of ‘controllable consumer devices’ to those for which the network operator has a direct control interface (the German model), the authors overlook other control models that are already being applied in practice, including on a large scale.
@arXiv_physicsmedph_bot@mastoxiv.pagePRIME-SVR: Physics-infoRmed Implicit Multi-Echo Slice-to-Volume Reconstruction for Fetal T2 mapping
Busra Bulut, Maik Dannecker, Thomas Sanchez, Sara Neves Silva, Steven Jia, Jean-Baptiste Ledoux, Leo Pomar, Joanna Sichitiu, Yvan Gomez, Meriam Koob, Vincent Dunet, Maria Deprez, Guillaume Auzias, Francois Rousseau, Jana Hutter, Daniel Rueckert, Meritxell Bach Cuadra
https://arxiv.org/abs/2607.20136 https://arxiv.org/pdf/2607.20136 https://arxiv.org/html/2607.20136
arXiv:2607.20136v1 Announce Type: new
Abstract: Slice-to-volume reconstruction (SVR) is the standard method for obtaining high-resolution (HR) 3D fetal brain volumes from motion-corrupted 2D MRI slice stacks acquired in multiple orientations. Existing SVR methods are optimized and validated only for clinical-range echo times (TEs), limiting their use at non-clinical TEs and making them incompatible with quantitative T2 mapping, a protocol- and center-independent biomarker of fetal brain maturation requiring HR reconstructions across multiple TEs. We present PRIME-SVR, the first implicit neural representation (INR) framework for joint HR reconstruction from multi-echo MRI. A single fully connected network models a continuous function from spatial coordinates to signal intensities across TEs, while a second network estimates slice-specific acquisition degradations. Cross-TE coherence is enforced via a Bloch equation-derived regularization penalizing deviations from expected T2 decay, with adaptive weighting that strengthens coupling for degraded stacks. The method is fully self-supervised. We validate PRIME-SVR on 39 in vivo fetal acquisitions (13 subjects x 3 TEs) from two centers, two vendors, and two field strengths (1.5 T and 0.55 T). Compared to state-of-the-art SVR, PRIME-SVR improves reconstruction sharpness by 47%, anatomical accuracy by 30%, and cross-TE structural consistency by 14%. It enables reconstruction at late TEs previously inaccessible to SVR, yielding the first 0.8 mm isotropic T2 maps at 0.55 T and the first T2 maps derived from INR-based SVR. PRIME-SVR also accelerates quantitative imaging by reducing the data needed for multi-TE reconstruction, cutting acquisition from 15 to 10 minutes while keeping T2 accuracy within 1.7% in white and deep gray matter, or to 5 minutes with a mean T2 error of 2.3% for high-quality acquisitions.
toXiv_bot_toot
@arXiv_csPL_bot@mastoxiv.pageCrosslisted article(s) found for cs.PL. https://arxiv.org/list/cs.PL/new
[1/1]:
- The EDGE Language: Extended General Einsums for Graph Algorithms
Odemuyiwa, Porumbescu, Nayak, Pellauer, Emer, Owens
https://arxiv.org/abs/2404.11591 https://mastoxiv.page/@arXiv_csDS_bot/112291034292734261
- Composable Verification Pipelines for Multi-Agent Systems
Julian Alfredo Mendez, Andreas Br\"annstr\"om
https://arxiv.org/abs/2607.16266 https://mastoxiv.page/@arXiv_csLO_bot/116956805448253784
- AoA: Theorem Proving Agent over Abstract Syntax Tree of Redesigned Language
Qiyuan Xu, Joshua Ong Jun Leang, Renxi Wang, Wenda Li, Haonan Li, Luke Ong, Conrad Watt
https://arxiv.org/abs/2607.16372 https://mastoxiv.page/@arXiv_csSE_bot/116956840082703202
- Exact Network Surgery: Functional Invariance and Gradient Plasticity in Reactive Computational Gr...
Abdallah Khemais (ISITCOM, University of Sousse)
https://arxiv.org/abs/2607.16568 https://mastoxiv.page/@arXiv_csAI_bot/116957045910225353
- Topology in Synthetic Domain Theory and its Formalisation in Agda
Runze Xue
https://arxiv.org/abs/2607.17292 https://mastoxiv.page/@arXiv_csLO_bot/116956840838635477
- Portable models as a replacement for industrial heuristics in compiler optimizations
Fot Nikolai, Vinarsky Alexander
https://arxiv.org/abs/2607.17389 https://mastoxiv.page/@arXiv_csSE_bot/116957129471226499
- Proceedings 42nd International Conference on Logic Programming
Wolfgang Faber, Laura Giordano, Ricardo Rocha, V\'itor Santos Costa
https://arxiv.org/abs/2607.17707 https://mastoxiv.page/@arXiv_csLO_bot/116956852438717401
toXiv_bot_toot
@arXiv_qbioNC_bot@mastoxiv.pageHow the fly holds a single goal: normalization, not selection, in Drosophila FC2
Gioele Nanni, Christopher Lee
https://arxiv.org/abs/2607.18969 https://arxiv.org/pdf/2607.18969 https://arxiv.org/html/2607.18969
arXiv:2607.18969v1 Announce Type: new
Abstract: A walking fly steers toward a goal direction, held as a bump of activity across the FC2 neurons of the fan-shaped body. These neurons also inhibit one another over distance, more strongly the farther apart they are, a feedback proposed to keep the fly on a single goal. We asked, from the connectome, what circuit produces this inhibition, and whether it lets FC2 actively choose one goal among competitors (a winner-take-all) or simply keeps a goal set elsewhere as one clean bump. Tracing the wiring in a single FlyWire brain, we find the inhibition is almost entirely global: four FB5A cells inhibit every FC2 neuron roughly equally, with a smaller, distance-dependent contribution from hDelta interneurons and a negligible direct component. A ring-attractor winner-take-all (the kind the compass uses) requires local recurrent excitation that the FC2 wiring lacks, so this geometry cannot build one; and across a range of dynamical models, including a spiking network, no version of the circuit locks onto a winner at the connectome-scaled reference coupling. FC2 therefore normalizes an externally set goal rather than selecting it, with FB5A likely acting as the global normalizer, much as the APL neuron does in the mushroom body. We are explicit about two open points: a different mechanism, mutual inhibition between two competing goals (which hDelta supplies), could in principle select at very strong coupling, and we bound rather than exclude it; and FB5A's inhibitory identity is a low-confidence prediction of the connectome's transmitter classifier, not yet measured, and likely not GABAergic. We then ask where the goal is actually set: the connectome nominates an upstream hDelta network and rules out the leading proposed alternative, whose neurons supply under 0.2% of FC2's input. Finally, we propose a direct experiment, silencing FB5A while imaging FC2, that would test the account.
toXiv_bot_toot
