Expansion goal
The essay extender starts from text supplied by the writer. Its goal is to develop underexplained ideas with reasoning, context, implications, examples framed as examples, and useful transitions while preserving the source thesis and point of view.
Prompt safeguards
The generation prompt instructs the selected model to preserve paragraph order, factual claims, and existing citations. It also instructs the model not to repeat sentences, pad the word count, invent evidence, invent citations, or add unsupported statistics.
Prompt instructions reduce risk but cannot guarantee compliance. Provider models can still misunderstand context or generate errors.
Input and output limits
A request must contain at least 20 and no more than 1,500 words or language-specific text units. The editor accepts no more than 12,000 characters. Users can request an approximate 1.5x, 2x, or 3x expansion.
The server caps a generation at 6,000 output tokens. Actual length depends on the source, model, language, and whether the model completes the request within that limit.
Models and processing
The editor lists only models enabled in the current arxvi configuration and shows the credit cost for the selected model. The configured providers currently supported by this workflow are OpenAI, Anthropic, and DeepSeek.
The submitted draft is sent to the selected provider only after the user starts generation. arxvi streams the provider response to the editor.
Human review standard
The writer should compare the result with the source, remove unsupported additions, verify facts and citations, check assignment rules, and revise the language before use. A useful expansion must add substance, remain traceable to the original argument, and leave final editorial control with the writer.
Corrections and updates
Product behavior and this methodology are reviewed when model configuration, prompts, limits, or data handling change. Report a mismatch or factual error to support@arxvi.com so it can be investigated and corrected.