chartreuse-verte/prose-rewriter-1.7b-v1.6

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 3, 2026License:agpl-3.0Architecture:Transformer Open Weights Featherless Exclusive Cold

chartreuse-verte/prose-rewriter-1.7b-v1.6 is a 1.7 billion parameter language model based on Qwen/Qwen3-1.7B-Base, specifically designed for paragraph-level prose rewriting. It re-renders LLM-generated text to sound more human while preserving original semantics. This model excels at reducing AI-generated 'slop' and unsupported content, making it ideal for refining fictional prose. It operates with a context length of 32768 tokens and is optimized for narrative English text.

Loading preview...

Model Overview

chartreuse-verte/prose-rewriter-1.7b-v1.6 is a specialized 1.7 billion parameter model built upon Qwen/Qwen3-1.7B-Base, enhanced with a rank-32 LoRA adapter. Its core function is to rewrite prose, particularly text generated by other large language models, to achieve a more human-like style while strictly maintaining the original meaning. This version is a fidelity release, focusing on reducing unsupported content (hallucinations) without sacrificing coverage or stylistic improvements.

Key Capabilities and Features

  • Prose Rewriting: Transforms LLM-generated text into more natural, human-sounding prose at the paragraph level.
  • Fidelity Focus: Significantly reduces the rate of unsupported sentences (invention) from 12.4% in v1.5 to 9.0% in v1.6, maintaining semantic coverage.
  • Slop Reduction: Continues to effectively reduce 'slop' lexicon density and 'banned constructions' found in AI-generated text.
  • Contextual Rewriting: Supports three edit modes (match, inflate, compress) to guide the rewriting process based on the input paragraph's length relative to its ideal form.
  • Optimized for Fictional Prose: Trained on a diverse dataset including amateur and professionally edited fiction, making it suitable for narrative content.
  • Untied lm_head: Features an untied lm_head for improved performance and specific output head adaptation.

Ideal Use Cases

  • Refining AI-Generated Content: Perfect for post-processing text from other LLMs to enhance its naturalness and reduce AI-specific stylistic markers.
  • Creative Writing: Useful for authors or content creators working with fictional prose, dialogue, and narrative structures.
  • Content Quality Improvement: Can be integrated into workflows where human-like text quality is paramount, especially for English narrative content.

Limitations

  • Not an Instruct Model: Designed for a single, specific task with a fixed prompt format.
  • Fictional Prose Only: Not intended for technical documentation or non-narrative text.
  • Paragraph-Level Input: Best performance with single paragraphs; longer inputs should be split.
  • English Only: Optimized for English narrative register.
  • Input Length Sensitivity: Inputs below ~15 words may lead to padding and fabrication.

Finetunes
1 models