Search and filter through extracted claims from AI researchers.
Showing 2781-2800 of 4537 claims
Large language models enable more expressive representation learning and improved generalization across diverse speakers, recording devices, and clinical environments for speech-based cognitive impairment detection
A lightweight continuous-assurance framework with dependency mapping, readiness contracts, and lifecycle governance can address reliability challenges in citizen-created organizational agents
AI agents created by non-engineering users through low-code/no-code tools create a reliability gap due to silent degradation from changing dependencies even without direct user modification
Non-invasive fine-tuning interventions for efficient reasoning can maintain behavior very similar to the original checkpoint while improving efficiency
Fine-tuning LLMs for efficient reasoning with non-invasive interventions that preserve original model behavior could become a standard technique in the field, similar to how quantization has become standard
Without constraints on the language model, surprisal theory makes no falsifiable predictions about human language processing.
The assumption that better corpus models improve predictions of human processing difficulty has been undermined by recent empirical work showing better corpus models can be worse predictors.
Surprisal theory is a tautology without additional constraints on the language model, because any pattern of difficulty is consistent with some language model under mild technical conditions.
Faithful explanations of time-series classifiers should identify subsequences that are both sufficient and necessary for predictions.
TimePNS improves time-series explanations by incorporating necessity signals through counterfactual interventions, not just sufficiency.
Existing sufficiency-oriented explanation methods can assign high importance to spurious subsequences that support predictions without being essential to the model's decision.
Most existing LLM systems for medical education focus on localized interactions rather than organizing entire clinical cases into decision-centered learning trajectories.
Task-specific fine-tuning substantially improves open-source LLMs on medical narrative generation tasks and narrows the gap with commercial models.
MedGame transforms static clinical cases into structured, executable storytelling games using a dual-engine design with Medical Narrative Designer and Story Director.
Metric saliency propagates through the image formation process itself, including multi-bounce light transport, capturing parameter dependencies that are semi-opaque to manual inspection.
Differentiable renderers can produce metric saliency maps that identify which scene elements most influence a scalar metric, analogous to gradient-based saliency in neural networks.
Molecular property prediction should use conformational ensembles rather than single representative conformations, since many molecules exist as ensembles in solution.
Pretraining EnsembleEGNN on conformational ensembles dramatically improves molecular property prediction compared to training from scratch (R²=0.477 vs 0.005).
Sycophancy should be understood as one expression of a broader judgment-updating process shaped by social influence, rather than as a one-dimensional failure mode.
LLM judgment revision is structured along three dimensions paralleling human social psychology: distance between incoming view and initial position, source attribution, and coalition structure.
Pipeline data may be stale or degraded.
Last synthesis: 2026-09-20. 9,017 pending.