Skip to content

Latest

arXiv cs.LG Papers

MetaOthello: A Controlled Study of Multiple World Models in Transformers

arXiv:2602.23164v2 Announce Type: replace Abstract: Foundation models must handle multiple generative processes, yet mechanistic interpretability largely studies capabilities in isolation; it remains unclear how a single…

arXiv cs.LG Papers

Warp RL: Reshaping Base Policy Distributions for Dynamics Adaptation

arXiv:2606.31043v2 Announce Type: replace Abstract: Residual reinforcement learning adapts a pretrained robot policy by learning an additive correction to its actions. While effective when adaptation amounts…

arXiv cs.LG Papers

StateFlow: Dual-State Recurrent Modeling for Long-Horizon Time Series Forecasting

arXiv:2607.00197v1 Announce Type: new Abstract: Long-horizon multivariate time series forecasting (LTSF) remains challenging due to non-stationarity, regime shifts, and error accumulation. The Variability-Aware Recursive Neural Network…

arXiv cs.LG Papers

Theoria: Rewrite-Acceptability Verification over Informal Reasoning States

arXiv:2607.01223v1 Announce Type: cross Abstract: When should an AI system's answer be trusted? Formal proof assistants offer certainty but cannot reach most of the problem distribution;…

arXiv cs.LG Papers

Tail-Shape Estimation in LLM Evaluation Is Fragile: A Protocol for Diagnosing False Positives

arXiv:2606.16511v2 Announce Type: replace Abstract: Recent work motivates moving large language model (LLM) evaluation from mean-based to tail-aware metrics, including conditional value-at-risk and tail-index estimates of…

arXiv cs.LG Papers

Verifiable Rewards for Calibrated Probabilistic Forecasting

arXiv:2607.00164v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards can in principle train calibrated probabilistic forecasters, since a proper scoring rule such as the Brier…

arXiv cs.LG Papers

Device Passport: Enabling Spatio-Temporal Pretrained Models to Generalize Across Input Layouts

arXiv:2607.00249v1 Announce Type: new Abstract: New device layouts pose a challenging modeling problem due to the lack of large datasets for each specific layout. Biosignal foundation…

arXiv cs.LG Papers

Scaling Up Thermodynamic AI Models

arXiv:2607.00170v1 Announce Type: new Abstract: Thermodynamic computing devices based on the Ising model show great promise for low-power AI inference and edge computing, but scalable methods…

arXiv cs.LG Papers

Optimization on the Oblique Manifold for Sparse Simplex Constraints via Multiplicative Updates

arXiv:2503.24075v4 Announce Type: replace-cross Abstract: Low-rank optimization problems with sparse simplex constraints involve variables that must satisfy nonnegativity, sparsity, and sum-to-1 conditions, making their optimization particularly…

arXiv cs.LG Papers

EVOTS: Evolutionary Transformer Search for Time Series Forecasting

arXiv:2607.00154v1 Announce Type: new Abstract: Evolutionary neural architecture design for multivariate time-series forecasting remains underexplored, with most approaches relying on fixed Transformer architectures despite substantial variation…