Gerard de Melo

Token Distillation: Attention-Aware Input Embeddings for New Tokens

Query-Level Uncertainty in Large Language Models

Knowledgeless Language Models: Decoupling Linguistic Competence and Factual Knowledge

Cross Lingual Long-tailed Entity Alignment in Knowledge Graphs

Rethinking Graph-Based Document Classification: Learning Data-Driven Structures Beyond Heuristic Approaches

EvalRes: Evaluating VLMs' Sensitivity to Image Resolution and Relative Detail Size

Unifying Latent Uncertainty Signals in Large Language Models for Improved Factual Precision

Image Token Matters: Mitigating Hallucination in Discrete Tokenizer-based Large Vision-Language Models via Latent Editing

Mitigate the Gap: Improving Cross-Modal Alignment in CLIP

Vector Grimoire: Codebook-based Shape Generation under Raster Image Supervision

Vector Grimoire: Codebook-based Shape Generation under Raster Image Supervision