10378 papers
By employing an interventional approach that modifies Transformer architecture, this paper demonstrates that enforcing spherical topology and uniform attention routing eliminates the delayed generalization phenomenon known as grokking in modular addition tasks, provided these architectural priors align with the task's intrinsic symmetries.
This paper introduces ASR-TRA, a novel test-time reinforcement learning framework that leverages audio-text semantic rewards and causal intervention to overcome confirmation bias in existing adaptation methods, thereby significantly improving ASR robustness and accuracy in noisy and accented environments without ground-truth labels.
SlideSparse is the first system to enable Sparse Tensor Core acceleration for accuracy-preserving structured sparsity on commodity GPUs by employing a sliding window decomposition and activation lifting technique, achieving near-theoretical speedups for LLMs without hardware modifications.
This paper introduces Recursive Inference Machines (RIMs), a neural reasoning framework that unifies neural backbones with classical recursive inference mechanisms to enhance the performance of models like Tiny Recursive Models on complex reasoning benchmarks such as ARC-AGI and Sudoku Extreme, as well as tabular data classification tasks.
This paper proposes a privacy-preserving, two-stage federated learning framework that clusters distributed wind turbines based on long-term behavioral statistics using Double Roulette Selection and recursive Auto-split refinement to train localized LSTM models, achieving competitive short-term forecasting accuracy that outperforms geographic partitioning while maintaining data locality.
This paper proposes a robust auditing framework for automatic speech recognition systems that moves beyond traditional Word Error Rate by introducing the Sample Difficulty Index and semantic metrics to quantify and mitigate the "diversity tax" disproportionately affecting marginalized speakers.
This paper introduces "Whisperer," a sample-efficient visual prompting framework that bootstraps frozen OCR models by using a four-stage behavioral cloning curriculum to learn diffusion-based preprocessors that enhance degraded text inputs, achieving an 8% absolute reduction in Character Error Rate without modifying the downstream model's weights.
This paper demonstrates that distribution shift is the primary cause of performance degradation in deeper layers of Vision Transformers and reveals that optimal out-of-distribution probing requires selecting between feedforward network activations and normalized self-attention outputs depending on the severity of the shift.
This paper proposes Bayesian Supervised Causal Clustering (BSCC), a novel framework that identifies homogeneous patient subgroups by simultaneously clustering individuals based on their covariate profiles and treatment effects, and validates its practical utility through simulations and real-world data from the International Stroke Trial.
This paper establishes a formal geometric framework linking AI debate and RLAIF by demonstrating that the value of debate scales with knowledge divergence between models, transitioning from negligible benefit to essential oversight as representations diverge, while identifying specific regimes where debate unlocks inaccessible outcomes or risks coordination failure.
The paper proposes Latent Policy Steering (LPS), a robust offline reinforcement learning method that achieves state-of-the-art performance by using a differentiable one-step MeanFlow policy to backpropagate original-action-space Q-gradients directly to a latent actor, thereby eliminating the need for proxy latent critics and sensitive hyperparameter tuning while ensuring policies remain within dataset support.
WavSLM is a single-stream speech language model that achieves competitive speech generation and consistency without text supervision by quantizing and distilling WavLM representations into a single codebook for autoregressive next-chunk prediction.
This paper proposes using asymmetric Shapley values as a superior metric for quantifying the importance of high-dimensional genomic features in clinical prediction models, addressing limitations of traditional approaches by accounting for collinearity and known causal directions, and provides efficient algorithms validated through a colorectal cancer progression study.
This paper introduces GALACTIC, a unified framework that bridges local and global counterfactual explainability for unsupervised time-series clustering by generating minimal perturbations to cross cluster boundaries and employing a provably efficient submodular optimization algorithm to derive concise, non-redundant global summaries of these transitions.
The paper proposes FairFinGAN, a WGAN-based framework that integrates fairness constraints via a classifier to generate synthetic financial data that effectively mitigates bias against protected attributes while maintaining high utility for downstream predictive tasks.
This paper demonstrates that predictive inference is governed by four distinct, pairwise non-nested admissibility geometries—Blackwell risk dominance, anytime-valid supermartingales, marginal coverage, and Cesàro approachability—each offering a unique certificate of optimality and proving that admissibility is irreducibly relative to the chosen criterion rather than a universal property.
This paper establishes a comprehensive statistical theory for globally optimal empirical risk minimization decision trees by deriving sharp oracle inequalities and minimax optimal rates over a novel piecewise sparse heterogeneous anisotropic Besov space, thereby providing rigorous theoretical guarantees for their performance in high-dimensional regression and classification under both sub-Gaussian and heavy-tailed noise settings.
This paper proposes Geometric-Aware Quantization (GAQ), a framework that enables efficient, low-bit inference for SO(3)-equivariant Graph Neural Networks by decoupling magnitude and direction to rigorously preserve continuous symmetry, thereby achieving significant speedups and memory reductions on molecular simulation benchmarks without compromising physical consistency.
This paper proposes InfoFlow KV, an information-flow-aware method that uses attention-norm signals and global positional reordering to selectively recompute key-value caches, thereby improving the efficiency and accuracy of retrieval-augmented generation for long-context tasks.
This paper introduces TS-BOSS, a scalable, score-based algorithm for learning causal structures in multivariate time series that extends the Best Order Score Search framework with dynamic Bayesian networks and grow-shrink trees, demonstrating superior performance in high auto-correlation regimes compared to standard constraint-based methods.