6628 papers
This paper introduces Best-of-Tails (BoT), an adaptive inference-time alignment framework that dynamically balances optimistic and pessimistic selection strategies by characterizing reward distribution tail heaviness via the Hill estimator and using Tsallis divergence as a tunable regularizer to optimize performance across diverse reasoning and preference tasks.
NEST is a novel device placement framework that unifies network, compute, and memory awareness through structured dynamic programming to jointly optimize hybrid parallelism strategies, achieving up to 2.43 times higher throughput and improved scalability compared to state-of-the-art baselines.
This paper introduces the first formal framework for cooperative multi-agent reinforcement learning with submodular rewards, proposing algorithms that achieve a 1/2-approximation guarantee and sublinear regret bounds while overcoming the exponential complexity of joint policy optimization.
The paper introduces CREDO, a method that combines interpretable credal envelopes with conformal calibration to generate distribution-free prediction intervals that explicitly account for epistemic uncertainty while maintaining rigorous coverage guarantees.
This paper introduces a novel framework for 3D gravity and magnetic joint inversion that reframes the problem as a rectified flow on the Noddyverse dataset, utilizing a Ginzburg-Landau regularizer and guidance methodology to capture solution distributions and improve ore identification while also releasing a trained VAE for 3D densities.
This paper establishes the exact minimax regret rate for contextual bilateral trade under heavy-tailed valuations with infinite variance by extending self-bounding properties to real-valued settings and combining them with truncated-mean estimation to derive an epoch-based algorithm that interpolates between classical nonparametric and trivial linear rates.
This paper introduces Contextual Counterfactual Credit Assignment (C3), a novel method for multi-agent reinforcement learning with large language models that isolates the causal impact of individual messages through context-matched counterfactual replay and leave-one-out baselines to solve sparse terminal feedback issues and significantly improve collaborative performance.
This paper introduces IGLU, a novel parametric activation function derived from a scale mixture of GELU gates that utilizes a Cauchy CDF to provide heavy-tailed gradient properties and robustness against vanishing gradients, alongside a computationally efficient rational approximation (IGLU-Approx) that achieves competitive or superior performance across vision and language tasks compared to standard baselines like ReLU and GELU.
SymLang is an open-source framework that integrates symmetry-constrained grammars, language-model-guided program synthesis, and Bayesian model selection to robustly discover accurate, interpretable governing equations from noisy and partial observations, significantly outperforming existing baselines in structural recovery and physical consistency.
This paper introduces "stochastic attention," a training-free method that interprets attention as gradient descent on a Hopfield energy function and utilizes Langevin dynamics to enable a single temperature-controlled mechanism for both exact retrieval and diverse, novel generation without requiring learned score networks or architectural changes.
This paper introduces a Physics-Informed Neural Operator (PINO) surrogate model that accelerates the retention analysis of Ferroelectric Vertical NAND devices by over 10,000 times compared to traditional TCAD simulations while maintaining physical accuracy, thereby enabling efficient optimization of device designs against charge detrapping and ferroelectric depolarization.
This paper introduces Single-pass Possibilistic Clustering (SPC), a novel streaming algorithm that leverages damped window footprints, covariance union, and a tunable fuzzifier parameter to effectively model non-spherical clusters while outperforming existing methods in purity and normalized mutual information.
This paper proposes a novel data augmentation paradigm that leverages Large Language Models to generate diverse, industry-resembling CAD programs by conditioning them on reference surfaces and modeling procedures, thereby addressing the scarcity of complex, spline-based geometric data in existing training sets.
This paper establishes that while enforcing fairness in attribute-aware machine learning systems consistently improves outcomes for disadvantaged groups, fairness constraints in attribute-blind settings can lead to "leveling down" where both groups suffer, depending on the underlying data distribution and the presence of masked candidates.
XGenBoost introduces two XGBoost-based generative models—a DDIM variant for small datasets and a hierarchical autoregressive model for large-scale data—that leverage tree-ensemble inductive biases to outperform existing methods in mixed-type tabular synthesis with lower training costs.
The paper introduces DySECT, a dynamic self-evolving toolkit that creates a symbiotic closed-loop system where an LLM continuously extracts structured triples to expand a knowledge base, which in turn refines the LLM's extraction capabilities through prompt tuning, few-shot sampling, or fine-tuning.
This paper proposes CN-CBF, a composite neural control barrier function method that combines multiple Hamilton-Jacobi-trained neural CBFs with a residual architecture to enable safe, non-conservative robot navigation in dynamic environments, achieving up to 18% higher success rates than baselines in both simulation and hardware experiments.
The paper introduces NerVE, a lightweight eigenspectral framework that reveals how FFN nonlinearities and optimizer geometry govern information flow and latent dimension utilization in LLMs, providing stable spectral signatures that predict generalization and guide architectural design beyond trial-and-error.
The paper introduces Swimba, a Switch Mamba model that enhances State Space Model capacity through parameter-space Mixture-of-Experts routing, achieving improved performance while maintaining a single state trajectory to preserve computational efficiency.
This paper presents a physics-consistent neural network framework for solving Cosserat elasticity problems in microstructured media, which enforces kinematic constraints during training and utilizes derived stability conditions like quasiconvexity and Legendre-Hadamard inequalities to validate the energetic stability of the learned equilibrium solutions.