Lam-2 is the second-generation small language model (SLM) developed based on the custom architecture Aricate V4. It performs excellently in question and answer tasks and has outstanding language coherence and creativity. Compared with the previous generation products, Lam-2 has significantly improved in grammatical correctness and text generation quality.

Natural Language ProcessingSafetensorsFrench

Tiny GptOssForCausalLM is a minimized causal language model designed specifically for unit tests of the TRL (Transformer Reinforcement Learning) library. This model is extremely small in scale and is mainly used to verify the correctness of the operation of reinforcement learning-related functions in the TRL library under a specific model architecture.

Natural Language ProcessingTransformers

YOLOv13 is an accurate and lightweight object detector. It proposes an adaptive correlation enhancement mechanism based on hypergraphs, a full-process aggregation and distribution paradigm, and replaces ordinary large-kernel convolutions with depthwise separable convolutions, effectively improving the detection performance in complex scenarios and achieving state-of-the-art performance in the MS COCO benchmark test.

Document image orientation classification module, used to distinguish the orientation of document images and correct them through post-processing to improve the accuracy of OCR processing.

Llama-3.1-70B-Instruct-RM-RB2 is one of the 7 reward models released by the Allen Institute for Artificial Intelligence, used to develop benchmarks and conduct correlation analysis with downstream PPO/Best-of-N performance.

Natural Language ProcessingTransformersEnglish

This is an image-to-text conversion model capable of processing both image and text inputs to generate corresponding text outputs.

MultimodalTransformers

ESM++ is a faithful implementation of ESMC, supporting batch operations and fully compatible with the Huggingface ecosystem without relying on the official ESM Python package. The large version corresponds to ESMC's 600-million parameter model.

Scientific ComputingTransformers

A Spanish grammar correction model based on the BART architecture, trained on the COWS-L2H dataset and 80,984 synthetic data entries, optimized for single-sentence correction

Natural Language ProcessingTransformersMultiple Languages

SuperGlue is an image feature matching model based on graph neural networks, which can jointly find correspondences and reject non-matchable points. It is suitable for pose estimation and image matching in challenging indoor and outdoor environments.

Computer VisionTransformers

Omni-Judge is an open-source mathematical evaluation model designed to assess the correctness of model-generated solutions based on questions and standard answers.

Natural Language ProcessingTransformers

Reflection Llama-3.1 70B is an open-source large language model trained with 'reflection tuning' technology, capable of autonomously detecting reasoning errors and correcting its approach.

Natural Language ProcessingTransformers

Image orientation correction model based on SwinV2 architecture for automatic detection and correction of image orientation

Computer VisionTransformers

CogVideoX-2B is a video generation model with 2 billion parameters, capable of generating corresponding video content based on text input. This model achieves a good balance between cost and compatibility, supports multiple optimization techniques, and is suitable for various video generation scenarios.

MultimodalDiffusersEnglish

CrystalChat-7B-Web2Code is a multimodal large language model based on the CrystalChat-7B large language model, specifically designed for generating HTML code from web page images. This model draws on the training method of LLaVa-v1.5, features full open source and high transparency, and supports generating corresponding HTML code from hand-drawn images or existing web pages.

MultimodalTransformersEnglish

A Korean-English bilingual enhanced model based on Llama3, optimized for Korean language capabilities through vocabulary expansion and knowledge correlation

Natural Language ProcessingTransformersMultiple Languages

This model is an experimental sequence-to-sequence architecture specifically designed for Italian, aiming to correct approximately 93% of errors produced by low-quality Optical Character Recognition (OCR) systems on Italian texts.

Natural Language ProcessingTransformersOther

Spivavtor-Large is an instruction fine-tuned Ukrainian text editing model focused on tasks such as text rewriting, simplification, grammar correction, and coherence optimization.

Natural Language ProcessingTransformersOther

This is a fine-tuned version of the Microsoft Phi-2 model for the natural language to SQL (NL2SQL) task on the b-mc2/sql-create-context dataset. This model can generate corresponding SQL queries based on natural language questions and database schemas, and it outperforms the defog/sqlcoder-7b-2 model in both inference time and accuracy.

Natural Language ProcessingTransformersMultiple Languages

A model-agnostic plug-and-play module suitable for open-source and API-based models, enhancing AI safety through residual correction strategies

Natural Language ProcessingTransformersEnglish

A fine-tuned T5-small model for text correction, specializing in fixing spelling errors and texts with missing spaces, particularly optimized for short English conversational sentences.

Natural Language ProcessingTransformersEnglish