Taxonomy-Aligned-Conversational-Tutor/TACTutor

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 29, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

TACTutor is a 4.5 billion parameter conversational English tutor developed by Taxonomy-Aligned-Conversational-Tutor (TACT). Built upon Qwen3.5-4B-Base, it was post-trained with supervised fine-tuning and taxonomy-aligned Group Relative Policy Optimization. This model is specifically designed to generate pedagogically adaptive teacher responses for English learning conversations, making it suitable for research in dialogue systems for education.

Loading preview...

TACTutor: A Taxonomy-Aligned Conversational English Tutor

TACTutor is a 4.5 billion parameter model developed by Taxonomy-Aligned-Conversational-Tutor (TACT) for generating teacher responses in English learning dialogues. Starting from Qwen3.5-4B-Base, the model underwent post-training with supervised fine-tuning (SFT) and taxonomy-aligned Group Relative Policy Optimization (GRPO). It is distributed as a merged Hugging Face Transformers model.

Key Capabilities

  • Generates pedagogically adaptive teacher responses: Designed to provide feedback, guide revisions, offer clarification, and manage conversations in an English tutoring context.
  • Research artifact for adaptive dialogue: Intended for research into pedagogically adaptive dialogue systems, focusing on English language learning.
  • Evaluated on TACTBench: Achieved an overall TACT score of 0.832051 on the 78-item TACTBench diagnostic benchmark.

Intended Use Cases

  • Research on English tutoring dialogue: Ideal for exploring and developing systems that provide adaptive and supportive conversational guidance for English learners.
  • Development of educational AI tools: Can be integrated into research prototypes for English language learning applications, with careful review of outputs.

Limitations

  • The model is a research artifact and not a replacement for human teachers.
  • Outputs should be reviewed, especially in high-stakes educational settings.
  • May produce incorrect, overly explicit, or contextually inappropriate responses.
  • Behavior outside English-language tutoring has not been established.

Finetunes
1 models