BR-RM is an innovative two-round reasoning reward model that solves the 'judgment diffusion' problem in traditional reward models through adaptive branching and branch-based reflection mechanisms, achieving industry-leading performance in multiple reward modeling benchmark tests.

Natural Language ProcessingTransformersEnglish

This is a quantized version of the judgment model with 32B parameters, suitable for tasks such as model evaluation, reward modeling, and text sorting.

Natural Language ProcessingTransformersEnglish

The static quantization version of CompassJudger-2-32B-Instruct, suitable for tasks such as model judgment, evaluation, reward modeling, and text sorting.

Natural Language ProcessingTransformersEnglish

This is a model based on the Cross Encoder architecture, specifically designed for text pair classification tasks. It is fine-tuned on the Quora question pair dataset and is suitable for semantic similarity judgment and semantic search scenarios.

Natural Language ProcessingSentence-transformersEnglish

RM-R1 is a training framework for reasoning reward models (ReasRM), which evaluates candidate answers by generating scoring criteria or reasoning traces, providing explainable judgments.

Natural Language ProcessingTransformersEnglish

QAT version of Amoral-Gemma-3, focused on generating analytically neutral responses, avoiding value judgments, suitable for sensitive and controversial topics.

Natural Language ProcessingGgufEnglish

A Quantization-Aware Training (QAT) version based on Google Gemma-3-4B, focused on generating analytically neutral responses, avoiding value judgments, and suitable for sensitive and controversial topics.

Natural Language ProcessingTransformersEnglish

rank1 is an information retrieval reranking model trained on Qwen2.5-0.5B, improving relevance judgment accuracy through generated reasoning chains.

Natural Language ProcessingTransformersEnglish

rank1-3b is an information retrieval re-ranking model trained on Qwen2.5-3B, which performs relevance judgments by generating reasoning chains

Natural Language ProcessingTransformersEnglish

This model is a fine-tuned version of DistilBERT-base-uncased on the CoLA (Corpus of Linguistic Acceptability) dataset, designed for grammatical acceptability judgment tasks.

Natural Language ProcessingTransformers

rank1 is a 14-billion-parameter reasoning re-ranking model that improves the performance of information retrieval tasks by generating explicit reasoning chains before making relevance judgments.

Natural Language ProcessingTransformersEnglish

rank1-7b is a 7-billion-parameter reranking model trained on Qwen2.5-7B, performing relevance judgments through generated reasoning chains

Natural Language ProcessingTransformersEnglish

DeepHermes 3 Preview is the latest version of Nous Research's flagship Hermes series large language models, combining reasoning and standard response modes with improved annotation, judgment, and function calling capabilities.

Natural Language ProcessingTransformersEnglish

Root Judge is a powerful medium-sized large language model designed for reliable and customizable LLM system evaluation. Fine-tuned based on Llama-3.3-70B-Instruct, it excels in pairwise preference judgment and multi-round instruction following tasks with source references.

Natural Language ProcessingSafetensorsEnglish

IntJudge-8B is a judgment model for evaluating open-ended multimodal generation methods, built on Qwen2-VL-7B-Instruct, with an 82.42% consistency rate with human judgment.

MultimodalTransformersEnglish

A high-performance reward model built on the Gemma-2-27B architecture, trained using the purified Skywork-Reward-Preference-80K-v0.2 dataset, excelling in preference judgment in complex scenarios.

Natural Language ProcessingTransformers

Vi-Qwen2-7B-RAG is a large language model fine-tuned based on Qwen2-7B-Instruct, specifically designed for Vietnamese RAG tasks. This model is trained on a Vietnamese dataset, supports a context of up to 8192 tokens, and has capabilities such as multi-context processing, negative information filtering, information integration, and positive-negative judgment.

Natural Language ProcessingTransformersOther

The Skywork Critic series of models are advanced judgment models that excel in paired preference evaluation. They can compare and evaluate a pair of input contents and provide detailed judgments.

Natural Language ProcessingPytorch

A safety protection tool fine-tuned based on internlm2-7b-chat, providing human-readable judgment explanations and fine-grained harm scoring

Natural Language ProcessingTransformersEnglish

A Flan-T5-xl fine-tuned model for targeted sentiment analysis, supporting sentiment polarity judgment (positive/negative/neutral) in English text

Natural Language ProcessingTransformersEnglish