All papers

Proposes a clinical world-model architecture using unified patient states and ordered action histories to anticipate treatment outcomes.

Routes gradients from imperfect monitor flags into sacrificial parameters to mitigate reward hacking during reinforcement learning.

Argues that competitive national races toward superintelligent artificial intelligence create risks of conflict and catastrophic loss of control.

Extracts and categorizes mathematical proof mechanisms and lemmas from an open release of synthetic mathematical manuscripts.

Measures post-training activation geometry shifts against run-to-run noise, identifying distinct structural signatures left by supervised fine-tuning.

Presents an open-source research workbench and portable packaging format to preserve relationships between computations, artifacts, and assessments.

Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction.

Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking.

Introduces a dynamic spatial audio reasoning benchmark and an architecture tracking acoustic sources across moving indoor scenes.

Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability.

Presents a compliant teleoperation control framework that reduces excessive contact forces during dexterous manipulation data collection.

Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition.

Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances.

Presents a multimodal benchmark testing whether models can explain observed chemical effects by linking structural features to reaction mechanisms.

Introduces a native sparse attention mechanism that combines sliding windows and keypoint selection for linear-time context extrapolation.

Uses reinforcement learning on large routing instances to help diffusion-based solvers generalize beyond their training instance sizes.

The paper demonstrates recipes to extract single-pass probability decisions directly from open instruction-tuned language models.

The paper demonstrates multi-fidelity training to optimize data generation compute and error scaling for neural PDE surrogates.

The paper presents a benchmark of synthetic paired comparisons to evaluate video reward models across specific visual and motion dimensions.

The paper describes a modular software backend architecture for extracting energy domain information and answering grounded questions from web sources.

The paper audits an enterprise confidence scoring mechanism for text-to-SQL models on the BIRD and Spider benchmarks.

The paper presents a reinforcement learning post-training method that scores intermediate generation trajectories in diffusion language models.

The paper proposes information-theoretic tests to measure learned equivariance in intermediate neural network representations without knowing explicit group transformations.

The paper presents a reasoning-conditioned framework that refines chain-of-thought explanations to improve propaganda detection without explanation annotations.

The paper introduces phase-conditioned affordance fields to improve equivariant visuomotor policy learning for robotic manipulation.

The paper introduces an adaptive network-slicing orchestration framework for medical IoT applications operating across edge-computing infrastructure.

The paper evaluates small language models for text-to-SQL tasks and tests lightweight iterative retry workflows to improve query accuracy.

The paper introduces decision models that balance fast single-pass predictions with adaptive reasoning for uncertain cases using reinforcement learning.

The paper proposes clustering observations by feature importance profiles to incorporate latent factor representations into deep learning models.

The paper develops an approach to map AI semantic blind spots and evaluates the quality of distinct human ideas.

The paper presents a statistical mechanics framework using modern Hopfield models to analyze inference reliability and phase transitions in Transformer attention.

The paper introduces a distribution-level evaluation metric for medical image segmentation to assess prediction sets without requiring paired ground-truth masks.

The paper presents a benchmark to evaluate whether formal specifications generated by language models accurately reflect natural-language requirements.

The paper proposes a hierarchical and attribute-aware supervision method to improve training efficiency and long-tail performance in generative recommendations.

The authors analyze compatibility across multi-schema latent spaces in motion models, showing shared representations require explicit paired data to align.

The authors propose an evolutionary framework that optimizes instance coverage and updates language model weights to discover complementary optimization heuristics.

The paper introduces a test-time wrapper that distorts exposed output probabilities to defend black-box machine learning models from query-based adversarial attacks.

The authors introduce a topology-aware transformer meta-trained across multiple electricity markets to forecast day-ahead electricity prices for unseen grids.

The authors provide a comprehensive textbook covering reinforcement learning methods, game-playing algorithms, and key historical milestones.

The paper develops a sequential betting approach to reliably detect data distribution shifts in unstructured text and image streams over time.

The authors propose a conditional flow-matching generative model to extend the field of view in diffusion brain MRI scans across scanners.

The paper presents a multi-station transformer model that jointly detects phase arrivals, locations, magnitudes, and associations for earthquake monitoring.

The authors introduce a time-frequency diffusion framework that improves the generation of rare extreme events in time-series data.

The authors propose a visual token compression framework that reduces computational costs by retaining only visual tokens with high reconstruction residuals.

The authors present a flight navigation policy for quadrotors that penalizes movement outside the camera view to avoid collisions in cluttered environments.

The paper develops a plug-and-play module for convolutional neural networks that provides optimal rotation and translation invariance based on invariant theory.

The paper proposes using test-time training memory to maintain consistent long-horizon visual scenes in real-time camera-controlled video world models.

The study evaluates scientific hypotheses generated by language models, concluding that they extend existing knowledge rather than starting new research directions.

The paper introduces a training-free technique and benchmark to help vision-language models correctly process and reject negated concepts in text prompts.

The authors design an FPGA implementation of spiking neural networks that performs both inference and training on-chip without backpropagation through time.

The paper evaluates seven machine learning and graph models on detecting misinformation within social media noise from conflict-related tweets.

The authors introduce a framework to track safety model judgments through output parsing to prevent malformed responses from causing bad decisions.

The paper presents a neurosymbolic framework combining query plan checks and candidate selection to improve text-to-SQL translation for financial governance.

The paper presents an in-context foundation model architecture for next-event prediction across diverse operational event logs.

The paper introduces a framework combining 3D Gaussian Splatting and physical simulation to evaluate robot policies using handheld fisheye cameras.

The paper introduces a dual-stage robust aggregation defense against data and model poisoning attacks in federated financial fraud detection.

The paper evaluates the energy efficiency, carbon intensity, and accuracy of ten post-training quantization methods across various language models and GPUs.

The paper presents a convex-analytic framework to select query directions for safety verification in Markov decision processes.

The paper investigates how indexed instruction wording affects the performance of lexical rankers for robot task retrieval in manipulation benchmarks.

The paper presents an efficient video super-resolution method using hybrid attention and reward-yield distribution matching distillation.