This is a video action recognition model fine-tuned on the WLASL dataset based on the VideoMAE-Base architecture. After 200 epochs of training, it achieved a top-1 accuracy of 52.96% and a top-5 accuracy of 79.88% on the evaluation set, specifically designed for sign language action recognition tasks.

Computer VisionTransformers

Apriel-1.5-15b-Thinker is a 15-billion parameter multimodal reasoning model developed by ServiceNow. It has the ability to reason with text and images, and its performance can rival models 10 times larger. It scored 52 points on the manual analysis index and performed excellently in enterprise domain benchmark tests.

MultimodalTransformers

Second-generation Italian named entity recognition model supporting 52 entity categories, widely applied in privacy protection, legal, and financial fields

Natural Language ProcessingSafetensorsMultiple Languages

SynthPose is a keypoint detection model based on the VitPose huge backbone network, fine-tuned with synthetic data to predict 52 human keypoints, suitable for kinematic analysis.

Computer VisionTransformers

SynthPose is a 2D human pose estimation model based on VitPose Base, fine-tuned with synthetic data, capable of predicting 52 anatomical keypoints

Computer VisionTransformers

Yi-Coder-1.5B is an open-source code language model with 1.5 billion parameters, supporting 52 programming languages and featuring 128K tokens of long-context understanding capability.

Natural Language ProcessingTransformers

MegaBeam-Mistral-7B-512k is a large language model with a long context trained based on Mistral-7B Instruct-v0.2. It supports 524,288 context tokens and performs excellently in multiple long-context benchmark tests, capable of efficiently processing ultra-long text inputs.

Natural Language ProcessingTransformers

Tele-FLM is an open-source multilingual large language model with 52 billion parameters, built on a decoder-only Transformer architecture and trained on approximately 2 trillion tokens. It has a stable and efficient pre-training paradigm and strong fact judgment ability, performing excellently at the same scale and sometimes even outperforming larger models.

Natural Language ProcessingTransformers

Italy's first Named Entity Recognition model capable of identifying 52 categories, based on BERT architecture, currently with 79% accuracy

Natural Language ProcessingTransformersOther

A bird image classifier fine-tuned on EfficientNet-B2, capable of recognizing 525 bird species with up to 99% accuracy

Computer VisionTransformers

Korean foundational model based on the BERT architecture, trained using Korean text data from the AI Hub web corpus (approximately 52 million texts)

Natural Language ProcessingTransformersMultiple Languages

This model is an English-French translation model fine-tuned on the kde4 dataset based on Helsinki-NLP/opus-mt-en-fr, achieving a BLEU score of 52.94 on the evaluation set.

Natural Language ProcessingTransformers

An Arabic automatic speech recognition model fine-tuned based on facebook/wav2vec2-large-xlsr-53, achieving a WER of 52% on the Common Voice Arabic dataset.

Audio ProcessingTransformersArabic

This is a speech recognition model fine-tuned on Breton language dataset based on facebook/wav2vec2-xls-r-300m, achieving a 52.3% Word Error Rate (WER) on the Common Voice 8 test set.

Audio ProcessingTransformersOther