Research Feed Page 17

The paper introduces a method for frozen language models to detect insufficient information and abstain from answering in a single forward pass without re-training.

The paper introduces sMuon, an algorithm that enables the use of the Muon optimizer within low rank adaptation techniques to improve fine-tuning performance.

The paper introduces a two-stage architecture that separates the interpretation of individual evidence sources from the final aggregation of those results to improve system reliability and auditability.

The paper provides a theoretical framework to identify which information must be serialized into a handover record to maintain model performance when switching between sessions.

DIVE optimizes frozen language models by evolving a diverse population of skills to improve reasoning performance while reducing inference costs.

The researchers developed a sensing framework that enables surgical robots to estimate cable tension and contact location in real time using a parallelized computation model.

The paper provides a method to construct statistically consistent algorithms for the multi-label Jaccard loss by using MinHash representations and F1-to-Jaccard regret transfer.

The paper introduces a hybrid model using CNN spatial feature extraction and bidirectional LSTM temporal modeling to improve the accuracy of interpreting EEG signals for brain-computer interfaces.

The paper introduces PatientAct, a framework designed to simulate mental health clients using clinical theory to improve realism in LLM-based agent interactions.

The paper introduces a method that optimizes evolutionary search by valuing future lineage potential instead of immediate offspring performance.

The authors developed a physics-informed convolutional neural network to automate the estimation of relaxation times in fuel cell monitoring data by embedding electrochemical principles directly into the training loop.

The paper introduces the Sparse Orthogonal Regression Technique (SORT) to reconstruct functional relationships from noisy or irregular datasets by combining basis expansions with L1-regularized regression.

The authors propose using black-box adversarial attack tasks as a new benchmark for evaluating global optimization methods in high-dimensional spaces.

The paper investigates whether speech models identify actual Parkinson's disease indicators or simply rely on dataset-specific noise when transferred across different languages and conditions.

The paper uses information theory to prove that machine learning systems have hard performance limits dictated by data structure rather than algorithm choice.

This paper evaluates whether contrast-based metrics can objectively measure the legibility of reconstructed historical manuscripts using multispectral imaging.

Researchers evaluated whether using egocentric video data alongside traditional inertial sensors helps distinguish Parkinson's freezing of gait from voluntary movement in home environments.

The paper introduces a method for maintaining consistent beliefs among heterogeneous AI agents in 6G networks by translating compact belief updates through an edge server.

The paper introduces a multi-expert generative adversarial network architecture to synthesize high-fidelity sign language videos capturing complex hand and facial movements.

UltraIR is a foundation model that uses simulation-to-real transfer learning to improve the accuracy and scalability of infrared spectroscopy analysis.