Projects
Applications:
- DIETA-Machine-Translation: DIETA is a compact, decoder-only Transformer (≈0.5B params) trained specifically for high-quality Italian↔English MT, released with five model checkpoints, three training datasets, and one evaluation set.
- MAdaKron: Adapter module to efficiently fine-tune Pre-trained Language Models.
- QueryExpansionLLMs: data and a sample code implementation for the paper titled “Comparatively Assessing Large Language Models for Query Expansion in Information Retrieval via Zero-Shot and Chain-of-Thought Prompting”.
- blending-diffusion-models: the implementation of the papers “How to Blend Concepts in Diffusion Models” presented at KMG@ECCV 2024.
- SGP23_AttPos4ShapeMatching: Attention and positional encoding for shape matching.
- visual-wsd-baseline: Visual Word Sense Disambiguation (Visual-WSD): Benchmark and Evaluation Script.
- Wiki crosslingual: intermediate-task training for multilingual LMs by predicting Wikipedia hyperlinks.
- Multi encoder-decoder OpenNMT-py: a multilingual implementation of a inner-attention encoder-decoder NMT model trainable with a language rotating schedule.
- supWSD: a supervised Word Sense Disambiguation system.
- BabelMorph: a multilingual morphological library for retrieving word inflections for nouns, verbs and adjectives based on Wiktionary.
- Babelfy: a Unified Approach to Word Sense Disambiguation and Entity Linking.
Resources:
- GLUE3D: GLUE3D is a Q&A benchmark for evaluation of 3D-LLMs object understanding capabilities.
- mu-shroom: Mu-SHROOM: The Multilingual Shared-task on Hallucinations and Related Observable Overgeneration Mistakes.
- ReasoningLLMs: data for the paper titled “Reasoning Capabilities and Invariability of Large Language Models”.
- Visual-WSD: SemEval-2023 Task 1: Visual Word Sense Disambiguation.
- XL-WSD: A Multilingual Benchmark for the Word Sense Disambiguation task.
- XL-WiC: A Multilingual Benchmark for Evaluating Semantic Contextualization.
- MuCoW: a multilingual contrastive Word Sense Disambiguation test suite for Machine Translation.
- WSD evaluation framework: an English benchmark for the Word Sense Disambiguation task.
- SEW: More than 200 million annotations of over 4 million different concepts and named entities.
- EuroSense: almost 123 million sense annotations for over 155 thousand distinct concepts and entities in 21 languages.
- SenseDefs: Almost 250 milion sense-annotations of over 35 million definitions for 256 languages.
- Two chapters of the Bible disambiguated: Sense-annotated corpus of 594 manual annotations (English and Spanish).
- Babelified Wikipedia: English Wikipedia disambiguated with Babelfy.