arXiv research

This research discovers model architecture and training dataset characteristics through strategic input probing.

A key challenge in developing and deploying Machine Learning (ML) systems is understanding their performance across a wide range of inputs. To address this challenge, we created the What-If Tool, an open-source application that allows practitioners to probe, visualize, and analyze ML systems, with minimal coding. The W…

We study the statistical behavior of reasoning probes in a stylized model of iterative computation inspired by neural algorithmic reasoning. The underlying computation is given by a looped Boolean circuit whose graph is a perfect $ν$-ary tree ($ν\ge 2$), with outputs recursively fed back as inputs across computation ro…

This two-part work puts forth the idea of engaging power electronics to probe an electric grid to infer non-metered loads. Probing can be accomplished by commanding inverters to perturb their power injections and record the induced voltage response. Once a probing setup is deemed topologically observable by the tests o…

Probe-level models have led to improved performance in microarray studies but the various sources of probe-level contamination are still poorly understood. Data-driven analysis of probe performance can be used to quantify the uncertainty in individual probes and to highlight the relative contribution of different noise…

This paper enhances language models with knowledge awareness.

A probing scheme is considered with an accessible and controllable qubit, used to probe an out-of equilibrium system consisting of a second qubit interacting with an environment. Quantum spontaneous synchronization between the probe and the system emerges in this model and, by tuning the probe frequency, can occur both…

We identify spectral conditions for reliable neural probe interpretation.

Optimal probing framework for scalable network monitoring.

Paper introduces Manifold Probe for discovering representation manifolds in superposition.

OLPA optimizes online user-centric selection with probing, achieving near-optimal regret bounds.

This paper explores what causal structures can be distinguished by observational and interventional probing schemes.

BoC probe assesses neural network confidence coherence, revealing architecture-specific uncertainty.

Paper analyzes why deeper layers of ViTs perform worse on out-of-distribution tasks.

System guides freehand obstetric ultrasound probe movements.

The ability of modeling the other agents, such as understanding their intentions and skills, is essential to an agent's interactions with other agents. Conventional agent modeling relies on passive observation from demonstrations. In this work, we propose an interactive agent modeling scheme enabled by encouraging an a…

CwA optimizes search performance by jointly learning a balanced database partition and a neural probing function.

This paper focuses on the problem of estimating historical traffic volumes between sparsely-located traffic sensors, which transportation agencies need to accurately compute statewide performance measures. To this end, the paper examines applications of vehicle probe data, automatic traffic recorder counts, and neural …

Study examines flaws in probing LLMs' knowledge and introduces a new method.

Self-attention prefers sparse functions of input sequences, reducing sample complexity.

Neural networks classify OOD images by their nearest neighbor in training data.

Distribution grids currently lack comprehensive real-time metering. Nevertheless, grid operators require precise knowledge of loads and renewable generation to accomplish any feeder optimization task. At the same time, new grid technologies, such as solar photovoltaics and energy storage units are interfaced via invert…

Method reveals hidden token embeddings of large language models.

The paper explores linear representations in language models using counterfactuals.

Differentiable Masking reveals how neural models make decisions across layers.

The increasing scale and sophistication of cyberattacks has led to the adoption of machine learning based classification techniques, at the core of cybersecurity systems. These techniques promise scale and accuracy, which traditional rule or signature based methods cannot. However, classifiers operating in adversarial …

Improved unsupervised probing for ranking tasks using Contrast-Consistent Ranking.

This work incorporates topological features via persistence diagrams to classify point cloud data arising from materials science. Persistence diagrams are multisets summarizing the connectedness and holes of given data. A new distance on the space of persistence diagrams generates relevant input features for a classifi…

Cosmic shear is a primary cosmological probe for several present and upcoming surveys investigating dark matter and dark energy, such as Euclid or WFIRST. The probe requires an extremely accurate measurement of the shapes of millions of galaxies based on imaging data. Crucially, the shear measurement must address and c…

Researchers create holographic super-embeddings for M5 and M2 branes.

Noise Injection probes deep learning dynamics during training phases.

New method tightens federated probe-logit distillation rates under varying bandwidths.

We study supersymmetric probe M5-branes in the AdS_4 solution that arises from M5-branes wrapped on a hyperbolic 3-manifold M_3. This amounts to introducing internal defects within the framework of the 3d-3d correspondence. The BPS condition for a probe M5-brane extending along all of AdS_4 requires it to wrap a surfac…

We use Donaldson invariants of regular surfaces with p_g >0 to make quantitative statements about modulispaces of stable rank 2 sheaves. We give two examples: a quantitative existence theorem for stable bundles, and a computation of the rank of the canonical holomorphic two forms on the moduli space. The results are in…

Accumulation of standardized data collections is opening up novel opportunities for holistic characterization of genome function. The limited scalability of current preprocessing techniques has, however, formed a bottleneck for full utilization of contemporary microarray collections. While short oligonucleotide arrays …

PROBE algorithm efficiently solves sparse high-dimensional linear regression.

This work defines idealized SSL representations and improves existing methods.

Minimal token perturbations reveal how Transformer models process information.

Higher gauge theory via differential nonabelian cohomology

Data augmentation doesn't improve robustness, contrary to belief.

New definition reveals encoding explanations that retain predictive power.

Concept Hierarchies and Formal Concept Analysis are theoretically well grounded and largely experimented methods. They rely on line diagrams called Galois lattices for visualizing and analysing object-attribute sets. Galois lattices are visually seducing and conceptually rich for experts. However they present important…

Predicts trainability of deep neural networks using reconstruction entropy.

New method predicts aphasia severity with narrower uncertainty intervals.

Entropy data replaces classical charts for smooth manifolds.

Calibrating classifiers reduces grouping loss using sufficiency criteria.

Neural network models have a reputation for being black boxes. We propose to monitor the features at every layer of a model and measure how suitable they are for classification. We use linear classifiers, which we refer to as "probes", trained entirely independently of the model itself. This helps us better understand …

Researchers propose a new SSL risk decomposition method to evaluate and improve self-supervised learning models.