STILL-3-TOOL-32B is a large language model that utilizes Python code to assist in the reasoning process, achieving an accuracy rate of 81.70% in the AIME 2024 evaluation.

Natural Language ProcessingTransformers

This is an audio enhancement model based on the ESPnet framework, specifically designed for the URGENT 2024 Challenge. It adopts the TF-GridNet v3 architecture and can handle single-channel audio with multiple sampling rates, effectively improving speech quality and clarity. It is suitable for speech enhancement tasks in noisy environments.

Audio ProcessingEspnetEnglish

Daredevil-8B is a super-fused model specifically designed to maximize MMLU scores. As of May 27, 2024, it is the highest-scoring Llama 3 8B model on MMLU.

Natural Language ProcessingTransformers

This model is a submission trained for the CLEF 2024 conference 'Touché Task: Ideology and Power Recognition in Parliamentary Debates', developed by the PolicyParsingPanthers team.

Natural Language ProcessingTransformers

The April 2024 edition of OLMo 7B is an upgraded version of the original OLMo 7B model. Based on the improved Dolma dataset and a phased training method, it has improved by 24 points in evaluations such as MMLU. This is an open and transparent language model that supports inference and fine-tuning tasks.

Natural Language ProcessingTransformersEnglish

MaGGIe is a binary mask-guided instance-aware alpha matting model for images and videos, accepted by CVPR 2024.

Computer VisionTransformers

As of April 8, 2024, Yamshadow Experiment 28-7B is the top-performing 7B parameter model on the Open LLM Leaderboard. Use with caution, as this may indicate overfitting to benchmarks.

Natural Language ProcessingTransformers

Q-Align is a multi-task visual assessment model focusing on Image Quality Assessment (IQA), Image Aesthetic Assessment (IAA), and Video Quality Assessment (VQA), published at ICML 2024.

MultimodalTransformers

LanguageBind is a multimodal model that extends video-language pretraining to N modalities through language-based semantic alignment, accepted by ICLR 2024.

MultimodalTransformers

LanguageBind is a multimodal pretraining framework that extends video-language pretraining to N modalities through language semantic alignment, accepted by ICLR 2024.

MultimodalTransformers