HypCBM reformulates concept activations as geometric containment in hyperbolic space to produce sparse, hierarchy-aware signals that match Euclidean models trained on 20 times more data.
hub
Wordnet: a lexical database for english
13 Pith papers cite this work. Polarity classification is still indexing.
13
Pith papers citing it
hub tools
citation-role summary
background 2
dataset 1
citation-polarity summary
representative citing papers
S²R² improves robustness of LoRA-tuned LLMs to prompt perturbations by penalizing semantic-segment drift while preserving clean performance and cross-dataset transfer.
DualGuard uses adaptive dual-stream watermark signals to detect and trace both paraphrase and spoofing attacks in LLM outputs while preserving text quality.
BERT stores relational knowledge extractable via cloze queries without fine-tuning and matches supervised baselines on open-domain QA tasks.
LLMs prompted with few-shot examples and rationales generate better reasoned distractors for MCQs than fine-tuned contrastive models across six benchmarks.
EMERGE is a benchmark dataset of 233K Wikipedia passages paired with 1.45 million Wikidata edit operations across seven yearly snapshots from 2019 to 2025 for evaluating knowledge graph updates from emerging text.
Authors release the multimodal WJoconde knowledge graph for French cultural heritage and a LLM-VLM pipeline that extracts and validates new triples from unstructured text and images to extend the graph.
Frozen DINOv2-L features with k-NN classification and PCA/ICA refinement achieve state-of-the-art few-shot performance on four benchmarks without any backpropagation or fine-tuning.
Gradient-boosted models with SHAP analysis find word familiarity as the dominant predictor of English vocabulary difficulty across Spanish, German, and Chinese L1 learners, with orthographic transfer adding value only for the first two groups.
Qwen-Image is a foundation model that reaches state-of-the-art results in image generation and editing by combining a large-scale text-focused data pipeline with curriculum learning and dual semantic-reconstructive encoding for editing consistency.
Gyan is a novel explainable non-transformer language model that achieves SOTA results on multiple datasets by mimicking human-like compositional context and world models.
Systematic experiments show that text decomposition methods and privacy budget allocation strategies produce significantly different privacy-utility trade-offs even under comparable total epsilon budgets.
A symbolic system extracts events from 450 property crime reports, with 54.1% high-confidence outputs, 93.7% mapped via PropBank-VerbNet-WordNet, and 100% human agreement on incident initiation, stolen items, and temporal cues.
citing papers explorer
-
Hyperbolic Concept Bottleneck Models
HypCBM reformulates concept activations as geometric containment in hyperbolic space to produce sparse, hierarchy-aware signals that match Euclidean models trained on 20 times more data.
-
Where Do Prompt Perturbations Break Generation? A Segment-Level View of Robustness in LoRA-Tuned Language Models
S²R² improves robustness of LoRA-tuned LLMs to prompt perturbations by penalizing semantic-segment drift while preserving clean performance and cross-dataset transfer.
-
DualGuard: Dual-stream Large Language Model Watermarking Defense against Paraphrase and Spoofing Attack
DualGuard uses adaptive dual-stream watermark signals to detect and trace both paraphrase and spoofing attacks in LLM outputs while preserving text quality.
-
Language Models as Knowledge Bases?
BERT stores relational knowledge extractable via cloze queries without fine-tuning and matches supervised baselines on open-domain QA tasks.
-
Beyond Fine-Tuning: In-Context Learning and Chain-of-Thought for Reasoned Distractor Generation
LLMs prompted with few-shot examples and rationales generate better reasoned distractors for MCQs than fine-tuned contrastive models across six benchmarks.
-
EMERGE: A Benchmark for Updating Knowledge Graphs with Emerging Textual Knowledge
EMERGE is a benchmark dataset of 233K Wikipedia passages paired with 1.45 million Wikidata edit operations across seven yearly snapshots from 2019 to 2025 for evaluating knowledge graph updates from emerging text.
-
Multimodal Cultural Heritage Knowledge Graph Extension with Language and Vision Models
Authors release the multimodal WJoconde knowledge graph for French cultural heritage and a LLM-VLM pipeline that extracts and validates new triples from unstructured text and images to extend the graph.
-
Rethinking the Good Enough Embedding for Easy Few-Shot Learning
Frozen DINOv2-L features with k-NN classification and PCA/ICA refinement achieve state-of-the-art few-shot performance on four benchmarks without any backpropagation or fine-tuning.
-
What makes a word hard to learn? Modeling L1 influence on English vocabulary difficulty
Gradient-boosted models with SHAP analysis find word familiarity as the dominant predictor of English vocabulary difficulty across Spanish, German, and Chinese L1 learners, with orthographic transfer adding value only for the first two groups.
-
Qwen-Image Technical Report
Qwen-Image is a foundation model that reaches state-of-the-art results in image generation and editing by combining a large-scale text-focused data pipeline with curriculum learning and dual semantic-reconstructive encoding for editing consistency.
-
Gyan: An Explainable Neuro-Symbolic Language Model
Gyan is a novel explainable non-transformer language model that achieves SOTA results on multiple datasets by mimicking human-like compositional context and world models.
-
A Systematic Exploration of Text Decomposition and Budget Distribution in Differentially Private Text Obfuscation
Systematic experiments show that text decomposition methods and privacy budget allocation strategies produce significantly different privacy-utility trade-offs even under comparable total epsilon budgets.
-
Ontology for Policing: Conceptual Knowledge Learning for Semantic Understanding and Reasoning in Law Enforcement Reports
A symbolic system extracts events from 450 property crime reports, with 54.1% high-confidence outputs, 93.7% mapped via PropBank-VerbNet-WordNet, and 100% human agreement on incident initiation, stolen items, and temporal cues.