HyPE detects harmful prompts as outliers in hyperbolic space and HyPS sanitizes them using explainable attribution, outperforming prior defenses in accuracy and robustness across datasets and adversarial scenarios.
Fine -Grained VLM Fine -tuning via Latent Hierarchical Adapter Learning[J]
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 3verdicts
UNVERDICTED 3roles
background 1polarities
background 1representative citing papers
DiffPrune reformulates visual token pruning as continuous control of token information using an Information Throttler with importance-conditioned variance-preserving noise, enabling fully differentiable learning of scores that are hard-thresholded at inference.
HyNeuralMap applies the hyperbolic Lorentz model to embed visual semantics and neural responses into a shared hierarchical space, outperforming Euclidean baselines on semantic prediction and cross-modal retrieval.
citing papers explorer
-
Harnessing Hyperbolic Geometry for Harmful Prompt Detection and Sanitization
HyPE detects harmful prompts as outliers in hyperbolic space and HyPS sanitizes them using explainable attribution, outperforming prior defenses in accuracy and robustness across datasets and adversarial scenarios.
-
Beyond Surrogate Gradients: Fully Differentiable Token Pruning for Vision-Language Models
DiffPrune reformulates visual token pruning as continuous control of token information using an Information Throttler with importance-conditioned variance-preserving noise, enabling fully differentiable learning of scores that are hard-thresholded at inference.
-
HyNeuralMap: Hyperbolic Mapping of Visual Semantics to Neural Hierarchies
HyNeuralMap applies the hyperbolic Lorentz model to embed visual semantics and neural responses into a shared hierarchical space, outperforming Euclidean baselines on semantic prediction and cross-modal retrieval.