All papers

Analyzes convergence rates of a multi-objective optimization method pairing norm-constrained linear minimization oracles with entropic task-weight updates.

Establishes stochastic first-order and zeroth-order algorithms achieving optimal rates for constrained nonsmooth nonconvex optimization.

Introduces randomized momentum transport to track stochastic gradients using probabilistically sampled transport corrections alongside cheap predictors.

Proposes a two-phase variance reduction method that adapts to average, standard, and noiseless smoothness settings without predefined hyperparameter constants.

Develops a sequential conditioning framework enabling Gaussian process linear partial differential equation solvers to enforce linear inequality constraints.

Evaluates combinatorial fusion analysis to build diverse ensembles combining classical models and variational quantum circuits for classification tasks.

Repurposes pretrained tabular foundation models into supervised dataset encoders using dataset-level readouts and contrastive post-training.

Presents Shared Canonical Component Removal to quantify predictive information retained by raw machine learning representations beyond predefined expert features.

Investigates training mechanisms in continuous diffusion language models, proposing design modifications for self-conditioning and embedding geometry.

Examines how chronological versus shuffled pre-training knowledge sequencing influences the sample efficiency of subsequent model knowledge updates.

Evaluates full sandbox snapshot exploration for coding agents, finding benefits primarily when recovery requires restoring runtime dependencies outside Git repositories.

Applies dynamic budget allocation algorithms at inference time to select scaffolds and epitopes during de novo nanobody generative design.

Derives finite-sample convergence guarantees for fully discrete conditional diffusion models with finitely many denoising steps under smooth density assumptions.

Introduces PISO, an optimization framework that combines partially available gradients with zeroth-order residual tracking to reduce variance from random probing.

Analyzes the dimensional requirements of exact recurrent memory states for cumulative learning under convex residual loss functions.

Establishes finite-time error bounds for single-trajectory asynchronous Q-learning that incorporate the hitting time of the underlying Markov decision process.

Introduces Prompteina, an all-atom generative model and dataset for generating de novo protein structures conditioned on natural language prompts.

Develops a multi-scale biophysical spiking neural network model to simulate sleep-wake regulation and discrepancies between objective sleep and subjective restedness.

Applies reinforcement learning to manage bundle scheduling and flow control at relay nodes in deep-space communication networks.

Analyzes round complexity in parallel coordinate samplers, proving a separation between schedule-averaged and output-law Kullback-Leibler divergences.

Proves upper bounds on the topological expressivity of neural networks measured by the Betti numbers of their output spaces.

The paper proposes a complex-valued neural network to fuse audio and visual similarity streams for open-vocabulary audio-visual event localization.

The paper proposes MIDAPN, a unified spatiotemporal network using identity-aware graphs and temporal learning across numerical and narrative-flow time series.

The paper presents DVLA-RL++, combining complementary semantic purification and reinforcement learning gating to reduce context contamination in few-shot learning.

The paper proposes data-space iteration, a few-step generative method removing flow discretization by directly refining noise predictions in data space.

Introduces an end-to-end diffusion model operating directly in radar space for few-step probabilistic precipitation nowcasting.

Presents a dense correspondence framework combining generative and semantic representations to handle non-rigid image editing and generation.

Designs a vision transformer using a single recurrent block with depth-programmed experts to reduce stored parameters without accuracy loss.

Presents an embodied coding agent framework using active physical and visual probing to refine aerial manipulation policies.

Evaluates explainability methods across CNN, ViT, and Mamba models using eye-tracking gaze data as ground truth.

Proposes a hybrid linear and sparse attention framework to accelerate feed-forward 3D foundation models across many views.

Synthesizes stereo manipulation data from monocular robot demonstrations using estimated depth and video diffusion models.

Develops a kernelized parameter sensitivity atlas to amortize input-specific gradient attribution across vision, forecasting, and biology models.

Introduces an asynchronous prefix denoising framework for extreme latent compression in diffusion models.

Introduces a federated vision-language model tuning framework combining global shared adapters with client-specific personalized projection adapters.

Studies how text-to-image anchor properties correlate with downstream video generation quality in image-to-video diffusion pipelines.

Audits joint-embedding predictive learning in medical CT scans, showing a bias that degrades performance on unpredictable localized pathology.

Introduces Contrastive Vision On-policy Distillation to prevent vision-language models from relying on answer shortcuts rather than visual evidence.

Investigates decoding register tokens alongside patch tokens in vision transformers to preserve global scene appearance details.

Introduces a streaming causal transformer framework for generating human motion continuously from spatial audio inputs.

Aligns robot policy representations with retrieved pretraining examples to maintain generalization during low-data fine-tuning.

Proposes a memory-efficient cross-attention method for distilling large paired audio-visual datasets into compact synthetic sets.

Evaluates shortcut learning and out-of-distribution failure in geospatial foundation models on cross-disaster damage assessment tasks.

Presents a unified plug-and-play posterior sampling framework combining diffusion and flow models via exact Gaussian conditioning.

Uses a frozen world model to continually steer generalist robot policies onto new tasks without catastrophic forgetting.

Evaluates public LLM watermark detectors and introduces a split-key mechanism combining public detection with private verification.

Presents a hypothesis testing framework using shadow models to audit whether a language model was distilled from a proprietary teacher.

Establishes information bounds and query complexity frontiers for verifying answers when transferring hints from related source tasks.

The paper presents a two-server federated learning protocol using secret sharing and public verification to defend against poisoning attacks in industrial networks.

The paper introduces a hypergraph autoencoder framework to filter malicious model updates and preserve Byzantine resilience in federated large language models.

The paper evaluates label-only backdoor defenses and demonstrates that inspecting representations in unpoisoned feature spaces improves poison detection accuracy.

The paper evaluates the theoretical and empirical limits of using query monitoring to detect distributed model extraction campaigns among ordinary users.

The paper models structured sentiment analysis as dependency graph parsing implemented through sequence labeling across five languages.

The paper proposes a four-tensor attention architecture that jointly normalizes attention over semantic and temporal fibers for scene prediction.

The paper introduces a dual-path parametric memory method that combines pooled history representations with sequential adapter updates.

The paper proposes a sparse factor framework to map multimodal behavioral data to interpretable concepts for depression assessment.

The paper demonstrates that entropy-based byte patching misses predictable calculation points and proposes a boundary-dependence signal to fix it.

The paper develops an LLM-based pipeline to automatically detect outcome switching and presentation spin in clinical trial reports.

The paper introduces a method to unlearn target knowledge while recovering collateral capability loss in non-target tasks.

The paper proposes an event-centric memory framework that converts dialogue history into query-relevant graph working memory for conversational agents.