Xtra-Computing/XtraGPT-14B

TEXT GENERATIONConcurrency Cost:1Model Size:14.7BQuant:FP8Ctx Length:32kPublished:Mar 9, 2025License:mg0-2.0Architecture:Transformer0.0K Cold

XtraGPT-14B by Xtra-Computing is a 14.7 billion parameter large language model, based on microsoft/phi-4, specifically fine-tuned for context-aware and controllable academic paper revision. It processes full paper context and follows criteria-guided instructions across 20 academic writing criteria, making it ideal for human-AI collaborative editing workflows. With a 32768 token context length, it excels at maintaining consistency and executing precise revisions within complex academic documents.

Loading preview...

XtraGPT: Context-Aware Academic Paper Revision

XtraGPT-14B, developed by Xtra-Computing, is a 14.7 billion parameter model specifically designed for human-AI collaborative academic paper revision. Unlike general-purpose LLMs, XtraGPT is fine-tuned to deeply understand the full context of a research paper and execute precise, criteria-guided revision instructions. It is built upon the microsoft/phi-4 architecture and supports a 32768 token context length.

Key Capabilities

  • Context-Aware Revision: Processes the entire paper to ensure revisions are consistent with the global narrative and context.
  • Controllable Editing: Follows specific user instructions, adhering to 20 academic writing criteria across 6 paper sections (e.g., Abstract, Introduction).
  • Human-AI Collaboration: Supports an iterative workflow where authors maintain creative control while leveraging AI for detailed improvements.
  • High-Quality Training: Trained on 140,000 instruction-revision pairs derived from top-tier conference papers (ICLR).

Good for

  • Academic authors seeking AI assistance for detailed paper revisions.
  • Researchers needing to ensure consistency and adherence to specific writing criteria.
  • Applications requiring precise, context-aware text manipulation within long-form documents.
  • Integrating AI into an iterative human-in-the-loop editing process for scientific writing.