11732 papers
The paper proposes Draft-Conditioned Constrained Decoding (DCCD), a training-free two-step inference method that decouples semantic planning from structural enforcement to significantly improve the accuracy and parameter efficiency of structured generation in large language models by mitigating the distortions caused by hard constraints.
This paper proposes "entropic-time inference," a novel paradigm that replaces linear token-based decoding with a self-organizing, entropy-driven architecture to dynamically allocate computational resources, optimize attention sparsification, and adapt sampling temperatures for more efficient and intelligent LLM generation.
This paper proposes CoIPO, a contrastive learning-based method that enhances the intrinsic robustness of large language models against prompt noise by minimizing the discrepancy between clean and noisy prompt outputs, demonstrating superior performance on the newly introduced NoisyPromptBench benchmark.
This paper introduces M-QUEST, a semantic framework and benchmark comprising 609 question-answer pairs across ten interpretive dimensions, designed to evaluate and advance the ability of large language models to perform commonsense reasoning and toxicity detection in internet memes.
The paper proposes IntPro, a proxy agent that enhances context-aware intent understanding by leveraging retrieval-conditioned inference from an individual intent history library, trained via supervised fine-tuning and multi-turn GRPO to effectively adapt to user-specific patterns across diverse scenarios.
This paper empirically evaluates the robustness of 13 Large Language Models against five structured Chain-of-Thought perturbation types, revealing that while model scaling significantly mitigates math errors, it offers limited protection against unit conversion errors and that vulnerability patterns vary heterogeneously across different corruption types.
This paper proposes a decision-safe framework for ranking large language models that utilizes a contextual Bradley-Terry-Luce model to construct statistically valid confidence sets for prompt-dependent rankings, thereby addressing the limitations of point estimates by quantifying uncertainty and distinguishing meaningful performance differences from noise.
The paper introduces NEURONA, a neuro-symbolic framework that integrates symbolic reasoning with fMRI grounding to decode interacting visual concepts, demonstrating that incorporating structural priors significantly enhances both decoding accuracy and generalization to unseen queries.
GreenPhase is an efficient, interpretable, and sustainable deep-learning model based on the Green Learning framework that achieves state-of-the-art earthquake detection and phase picking performance on the STEAD dataset while reducing computational costs by approximately 83% through its unique feed-forward, multi-resolution architecture that eliminates backpropagation.
This paper introduces SMMA, a fully automated deep learning framework that accurately measures geniohyoid muscle thickness during speech, enabling scalable analysis of speech motor control and objective assessment of related disorders by eliminating the need for time-consuming manual annotation.
This paper corrects mathematical errors in the original theoretical foundation of the UMAP algorithm, provides a self-contained derivation of the underlying functors and metric realization, and clarifies the correspondence between the theoretical framework and the practical UMAP implementation.
This paper proposes a "Learning Order Forest" method that employs a joint learning mechanism to iteratively construct tree-based distance structures for qualitative attributes, thereby effectively capturing local order relationships to achieve superior clustering performance on datasets with nominal values.
This paper introduces GLOT, a lightweight and efficient structure-aware pooling module that constructs and refines token-similarity graphs from frozen LLM outputs to achieve robust sentence representations with significantly fewer parameters and faster training times compared to existing methods.
This paper proposes a novel, implementable adaptive parameter selection strategy for kernel-based gradient descent that integrates bias-variance analysis with the splitting method and empirical effective dimension to achieve optimal generalization error bounds across diverse kernels, target functions, and error metrics.
This paper introduces heterogeneous time steps (HTS) to Equilibrium Propagation, demonstrating that assigning neuron-specific time constants drawn from biologically motivated distributions improves training stability and robustness while maintaining competitive performance.
This paper introduces the Surprisal-Rényi Free Energy (SRFE), a novel log-moment-based functional that bridges forward and reverse Kullback-Leibler divergences by revealing a mean-variance tradeoff and providing a variational characterization that controls large deviations in code-length, thereby clarifying the geometric and statistical structure underlying these distinct learning objectives.
This paper introduces a simple variant of the Squint algorithm for the classic expert problem and proves, via a straightforward modification of the original proof, that it achieves a regret bound similar to that of a recent variant of the NormalHedge algorithm.
This paper proposes a scalable, contrastive causal discovery model that leverages paired observational and single-regime soft interventional data to construct globally consistent causal structures, theoretically proving its ability to recover identifiable edges and outperform non-contrastive methods in both in-distribution and out-of-distribution scenarios.
This replication study of FairDICE, a multi-objective offline reinforcement learning algorithm, reveals that while its theoretical claims hold, a critical code error initially reduced it to standard behavior cloning and underspecified hyperparameters hindered reproducibility, though corrected experiments demonstrate its potential to scale to complex environments despite a reliance on online tuning.
This paper demonstrates that a significant portion of transformer MLP nonlinearity is redundant and context-dependent, showing that a lightweight gating mechanism can dynamically replace these computations with linear surrogates to reduce computational waste or, when applied strategically with full retraining, actively improve model performance by eliminating harmful nonlinearities.