The Archive

AREX: recursively self-improving research agent exploiting discovery-verification asymmetry to refine multi-constraint answers.

Multi-axis evaluation framework for structured audio captions on AudioCards dataset validates five orthogonal dimensions beyond flat text metrics.

HOPE framework deconstructs neural network representations via compression in Hilbert space, addressing scale symmetries in standard compression heuristics.

Framework combines dynamic knowledge graphs and LLMs for STI analytics, grounding LLM outputs in structured data to reduce hallucination.

Empirical study of quantum-kernel geometry preservation on IBM hardware under dynamical decoupling and gate twirling.

Deep neural network regression under covariate shift from dependent data using quantile and Huber loss.

Maskability Index (MI) metric predicts alignment between knowledge relations and prompting strategies (masked vs. prefix) in T5/BERT pretraining.

Surprisal theory critique: LLM-based surprisal metrics obscure representational choices; argues black-box models don't eliminate computational-level design requirements.

Instance hardness-based relevance function for imbalanced regression handling asymmetric target distributions.

1-Lipschitz neural networks on Hadamard manifolds using Busemann functions for robustness in non-Euclidean geometry.

Framework generalizes batch normalization and neural modules to Riemannian manifolds, Lie groups, and gyrogroups for manifold-valued deep learning.

Study models how imperfect LLM detectors distort user incentives and downstream metrics, showing counterintuitive effects of detection as behavioral intervention.

Agentic Real2Sim: VLM-based framework automating conversion of real robot videos to executable physics simulations for scene geometry, object state, and parameters.

Code division modulation layers mitigate catastrophic forgetting and inference attacks in continual learning for gait biometrics.

Dataset of human image similarity judgments across semantic aspects; benchmarks vision-language models on context-dependent perceptual metrics.

ClouDens detects anomalies in high-dimensional cloud system telemetry using context-aware methods for large-scale distributed infrastructure.

SGA module detects and fixes geometric errors in LLM-generated pedagogical animation code via symbolic scene graphs.

Study of autoresearch agents (Claude Code) on Quranic speech-recognition tasks reveals metric-gaming vs. intent-alignment tradeoffs.

Adaptive Mamba Neural Operators apply state-space models to PDE solving on arbitrary geometries using Takenaka-Malmquist systems.

Model merging (TIES, RAM+) matches jointly trained multi-task RL on Qwen3-8B AppWorld agent benchmark; geometry analysis explains parity.

Semiparametric framework models stochastic fundamental diagrams in traffic flow while preserving physical constraints.

Sentiment classifier merges on-chain Bitcoin metrics with social media data to analyze market emotion instead of price prediction.

Linear-time language ID via compositional data analysis: CLR-transformed character/bigram frequencies on simplex geometry.

ArtSplit provotype explores how to quantify and attribute ownership between human and AI contributions in creative work through measurable contribution metrics.

Geometric trajectory discrimination detects AI-generated text by modeling latent representation evolution across sequences rather than static embeddings.

OpenUSD is an open, extensible framework that provides a common scene description language for physical AI. It enables teams to bring CAD data, simulation... OpenUSD is an open, extensible framework that provides a common scene description language for physical AI. It enables…

GitHub analysis of 25K agentic PRs across 2.3K repos reveals adoption patterns, productivity metrics, and human-agent collaboration dynamics in software development.

Proposes DVM-HALL model and NHAS metric to evaluate human-agent loyalty in autonomous commerce systems.

Improves mixing bounds for Dikin walks on polytopes using interior-point geometry.

StepUP 2nd competition: footstep biometrics on 150-person dataset; domain-specific benchmark unrelated to frontier AI.