thu-coai/Crispers-7B-v1

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Apr 24, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Crispers-7B-v1 by thu-coai is a 7.6 billion parameter conversational language model specifically designed for Cognitive Restructuring (CR) in multi-turn supportive dialogues. It integrates identification and restructuring stages of negative thoughts, sentence-level supportive conversation strategies, and a multi-channel loop mechanism for iterative CR. This model excels at providing emotional support and professional guidance, aiming to enhance users' emotional well-being through psychotherapeutic interactions.

Loading preview...

Overview

Crispers-7B-v1 is a 7.6 billion parameter conversational large language model developed by thu-coai, specifically engineered for Cognitive Restructuring (CR) in multi-turn supportive dialogues. It is part of a novel framework called CRDial, which addresses the limitations of existing CR implementations by focusing on a more aligned psychotherapeutic process. The model is trained on Crisp, a large-scale, high-quality bilingual dialogue dataset distilled from LLMs, enabling it to provide emotional support and professional guidance.

Key Capabilities

  • Cognitive Restructuring: Facilitates the identification and restructuring of negative thoughts through structured multi-turn dialogues.
  • Supportive Conversation Strategies: Integrates sentence-level supportive conversation techniques to foster emotional connection and well-being.
  • Iterative CR Workflow: Employs a multi-channel loop mechanism for an iterative and effective CR process.
  • Bilingual Support: Trained on a bilingual dataset, suggesting capabilities in multiple languages (e.g., English and Chinese as shown in examples).
  • Psychotherapeutic Interaction: Designed to mimic human-LLM interactive psychotherapy, offering a caring and compassionate persona.

Good for

  • Mental Health Support Applications: Ideal for developing AI-driven tools that assist individuals in managing negative thoughts and improving emotional well-being.
  • Conversational AI for Therapy: Suitable for research and development in AI-powered psychotherapeutic interventions.
  • Emotional Support Chatbots: Can be used to create chatbots that provide empathetic and professional guidance in multi-turn interactions.

Finetunes
1 models