All papers

FETALFusion combines CNN and Mamba architectures to segment fetal heads accurately across different ultrasound acquisition methods.

Virtual Head Attention uses two linear operations to compress key-value cache overhead without increasing training compute or sacrificing output quality.

The paper proposes a shopping basket choice model using specialized negative sampling to capture product interactions from retail transactions.

BELT compresses key-value caches using attention-distribution alignment and learned global rank allocation across model heads.

The paper analyzes how vision-language models favor 2D image cues over internal 3D depth representations when judging relative distances.

The authors show that medical vision-language models frequently abandon correct clinical answers when prompted with fabricated medical documents.

The paper presents a saliency-guided patch replay method to mitigate catastrophic forgetting in spiking neural networks under tight memory budgets.

The paper proposes predictability-aware supervised fine-tuning to resolve training and inference order mismatches in diffusion language models.

SWEPatchTest provides a benchmark to evaluate whether agent-generated test suites correctly accept valid patches and reject invalid ones.

The study shows that training on-policy distillation exclusively on student errors degrades accuracy rather than improving it.

ChampionKV adaptively allocates key-value cache memory budgets across attention heads using observation-window concentration proxies.

The authors propose methods to train machine learning potentials that transfer across quantum-chemical methods to exceed single-method training accuracy.

The paper introduces a training-free token pruning and merging framework with new robustness metrics for vision transformers.

The paper proposes a neural operator and state-space framework for high-fidelity 3D turbulent flow reconstruction from coarse data.

The paper investigates transforming text embeddings into visual sketches to improve model explainability and evaluate their utility in text retrieval.

This paper presents an intermediate-representation method to audit whether specific private data was used in LoRA fine-tuning of language models.

The authors prove and demonstrate empirically that a two-layer attention Transformer can learn and implement in-context Q-learning algorithms.

The paper examines conceptual parallels between measurement science and machine learning uncertainty evaluation to propose a more unified assessment framework.

This paper proposes decoupling error proposing from error validation to filter invalid candidates under budget constraints for expert human review.

This paper shows chain-of-thought monitoring can fail even when reasoning is necessary, proposing a multi-level hierarchy to measure monitorability.

This paper presents a post-hoc multicalibration framework to correct systematic biases such as position and model family in LLM evaluations.

This paper formalizes graph-structured missing features where missingness spreads across edges, testing existing graph propagation methods under this setting.

This paper investigates theoretical limits and conditions for verifying reachability policies in unknown Markov decision processes using sample access.

This paper presents a generative compression framework using implicit randomized Hadamard dictionaries to speed up diffusion-based image and video coding.

This paper presents an interpretability protocol to monitor latent reasoning dynamics and stability across multiple looped language model families.

This paper proposes an inference-time intervention method to detect and correct modality misrouting that causes multimodal hallucinations.

This paper introduces a scheduler that decouples prefix matching from state readiness to eliminate execution bubbles in hybrid LLM serving.

This paper defines conditions under which learned multidimensional representation blocks preserve causal abstractions and high-level concept interventions.

This paper introduces an order-flow allocation framework and evaluation protocol to analyze how intervention rules influence adaptive trading dealers.

This paper studies how different framings of user claims induce sycophancy in large language models on factual questions.

This paper introduces a self-distillation method that searches over repeated layer execution orders to create stronger teacher guidance.

This paper establishes functional convergence bounds for wide neural networks approximating Gaussian processes on arbitrary domains.

This paper proposes predicting the final state first and using diffusion to interpolate long-horizon physical particle trajectories.

This paper proposes a recurrent architecture that adds rotational degrees of freedom to prevent oscillatory bottlenecks in looped reasoning.

This paper presents a neuro-symbolic graph transformer with first-order logic loss to recognize hierarchical human intentions from multimodal activity sequences.

QORH models multi-relation rules as query-specific hypergraphs with message passing to improve inductive link prediction on unseen knowledge graph entities.

RePara trains diffusion language models using group-relative reinforcement learning on short token blocks to improve math generation speed and accuracy.

The paper evaluates seven language models on estimating soil organic carbon concentrations, calculating carbon stock, and comparing agricultural paddock pairs.

SATResetCoP restarts connection-tableau search branches while preserving accumulated clauses in an incremental SAT solver to improve theorem proving efficiency.

This paper audits a specific scalar feature scaling method to test whether claimed curvature benefits hold for domain generalization benchmarks.

Runoff eliminates option-order bias in multiple-choice model predictions by scoring items independently before running balanced candidate comparisons.

The authors present a framework using level-wise multimodal memory to reconstruct coherent heading hierarchies across long, multi-page documents.

LiFUSE uses frequency-spatial self-supervised pretraining on non-contrast CT scans to improve representation learning for liver disease detection.

FALLNEST condenses large imbalanced wearable motion sensor datasets into small balanced synthetic sets for training fall detection models.

The authors analyze how different optimizers leave distinct implicit bias fingerprints that influence training data reconstruction from model checkpoints.

AnnalsChallenge provides 50 formalized mathematics theorem statements in Lean 4 drawn from papers published in the Annals of Mathematics.

LR-Guard distills explicit safety reasoning into compact latent representations to speed up guardrail classification while enabling on-demand trace reconstruction.

AffectWorld benchmarks world models on predicting emotion-driven human actions and subsequent environment changes across annotated video samples.

This paper examines using provenance reaction action graphs and machine learning to encode multi-step material synthesis pathways from experimental history.

ShoeFit presents a multi-view dataset and a dual-stream diffusion framework for consistent virtual footwear try-on synthesis.

The authors evaluate how 8-bit and 4-bit quantization methods affect model accuracy across Bavarian, Standard German, and English benchmarks.

The authors propose a probe-based method to measure and calibrate systematic scoring biases in black-box LLM evaluation judges.

The authors evaluate offline graph-modified embeddings for vector retrieval, improving multi-hop evidence recovery without requiring query-time graph access.

The authors trace internal representations across prompt optimization techniques to show how discrete prompts, prompt tuning, and prefixes alter model processing.

The authors reproduce and compare five language model factuality alignment methods on a standardized benchmark across identical base models and evaluation protocols.

The authors propose a resource-efficient two-stage approach for estimating 3D road user shapes and orientations from roadside cameras and radar.

The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities.

The authors present a graph-based aggregation method to defend federated language model fine-tuning against backdoor attacks by identifying update consistency patterns.

The authors evaluate language models on finance peer review tasks, finding predictive ability for publication outcomes through correctness rather than originality assessments.

The authors introduce a topology-aware molecular representation language that enables fragment-level representation learning and precise structural edits.