Home › Datasets › Rewriting & Editing

General Assistant Pack

How to Fine-Tune an LLM for Rewriting & Editing

Rewriting text for tone, clarity, length, and style; grammar and structural editing of existing content.

Fine-tune a model on this dataset Preview Dataset

$19 for the dataset. Training is priced separately by model, starting at $47.

What can this kind of fine-tuning teach an LLM?

A general-purpose model may already know some of this subject. Fine-tuning serves a different purpose: it repeatedly demonstrates the kinds of requests, response patterns, terminology, decisions and workflows you want the model to handle more consistently.

For Rewriting & Editing, the dataset focuses on rewriting text for tone, clarity, length, and style; grammar and structural editing of existing content. Fine-tuning shapes recurring model behavior; it does not turn the model into a live database or guarantee correctness on every new case.

What training data should cover

Useful supervised fine-tuning data should resemble the work you expect the model to perform. This Rebiha dataset's domain definition is organized around 18 coverage areas:

How Rebiha prepared the training material

This page contains 854 instruction-response pairs from the existing Rebiha dataset. Its domain definition was designed around the coverage areas above so the training material spans multiple realistic tasks rather than one narrow prompt pattern.

Rebiha's broader dataset workflow uses structured quality review to help identify problems such as contradictions, unsupported claims, unsupported assumptions, calculation errors and other weak training examples. You can inspect real examples on this page before deciding whether the dataset fits your model and use case.

What is in this dataset

854 instruction-response pairs covering rewriting text for tone, clarity, length, and style; grammar and structural editing of existing content. Each pair is a realistic instruction with a full written response, in the JSONL format used for supervised fine-tuning. No synthetic filler, no duplicated rows.

Example pairs

Real rows from the dataset, shown in full.

Rewrite the following passage to be more concise without losing any important information: 'Due to the fact that the weather conditions were not favorable, the event organizers made the decision to postpone the outdoor concert until a later date, which will be announced to the public in the near future at such time as the weather forecast becomes more reliable and predictable.'

Because the weather was poor, organizers postponed the outdoor concert. A new date will be announced once forecasts are reliable. What changed: Removed redundant phrases ('due to the fact that' → 'because', 'made the decision to postpone' → 'postponed'), cut filler ('at such time as', 'in the near future'), and tightened the final clause. Original length: 39 words. New length: 20 words—roughly half, with no loss of meaning.

Please edit this paragraph for grammar and punctuation errors. Keep the original wording as much as possible, only fixing mistakes: 'The team's success, which included winning three consecutive championships, were largely due to their coach's strategy and the players commitment. However, there was disagreements among the staff about training methods, which leaded to tension in the locker room. Each player had there own opinion, but the coach's authority, ultimately prevailed.'

The team's success, which included winning three consecutive championships, was largely due to their coach's strategy and the players' commitment. However, there were disagreements among the staff about training methods, which led to tension in the locker room. Each player had his or her own opinion, but the coach's authority ultimately prevailed. What changed: - 'were' → 'was' (subject 'success' is singular) - 'players commitment' → 'players' commitment' (missing apostrophe for possessive plural) - 'there was disagreements' → 'there were disagreements' (subject-verb agreement) - 'leaded' → 'led' (incorrect past tense) - 'there own opinion' → 'his or her own opinion' (fixing homophone 'there' → possessive; also corrected pronoun agreement since 'each player' is singular) - Removed the comma after 'authority' because 'ultimately prevailed' is not a nonrestrictive clause—it's part of the main predicate.

Rewrite the following casual, informal text into a formal business tone suitable for an official company memo: 'Hey team, so we got the numbers back from the last quarter, and honestly, they kinda suck. Sales are down like 15%, which is not great. We need to get our act together and figure out what's going wrong. Let's set up a call for next week to hash it out. Bring ideas, don't just complain. Thanks!'

To all team members, Following the release of last quarter's performance data, we must address a concerning decline: sales have decreased by approximately 15% compared to the previous period. This shortfall requires immediate attention and a coordinated response. Accordingly, a strategy meeting has been scheduled for next week to analyze contributing factors and develop corrective actions. All attendees are expected to come prepared with constructive proposals rather than focusing solely on challenges. Your cooperation is appreciated. What changed: Replaced colloquial phrases ('kinda suck', 'get our act together', 'hash it out') with neutral, professional language ('concerning decline', 'coordinated response', 'strategy meeting'). Removed emotional exclamation ('not great!' → 'concerning decline'), softened direct criticism ('don't just complain' → 'rather than focusing solely on challenges'), and added formal salutation/closing. Retained all key information: the 15% drop, the need for a meeting, and the expectation of actionable input.

Who this is for

Builders who want their fine-tuned model to reliably perform this specific capability, as part of a broader assistant or agent.

Which models this works with

The dataset is plain JSONL, so it works with any instruction-tuned open model. On Rebiha you can train it directly on:

Training uses LoRA, or QLoRA on a 4-bit base for larger models. Your base model's weights are never modified — training produces a separate adapter.

How it works

  1. Pick this dataset and a base model.
  2. Rebiha runs the job on a managed GPU. You set up nothing.
  3. Download your model: a ready-to-run GGUF, plus developer assets (adapter, tokenizer, configs).

Fine-tuning or RAG?

Use RAG when the model needs facts that change — your prices, your documents, today's inventory. Use fine-tuning when you want consistent behavior and structure without prompting for it every time. This dataset teaches the shape of this capability, not facts to look up.

How managed fine-tuning works

  1. Inspect the dataset examples and choose a compatible base model.
  2. Rebiha runs the training job on managed GPU infrastructure using LoRA or QLoRA as appropriate.
  3. Download portable model assets for your own workflow, including the available ready-to-run and developer outputs for the selected training path.

Fine-tuning or RAG?

Use fine-tuning when the objective is recurring behavior: response structure, domain patterns, terminology, workflow, or how the model approaches a class of requests. Use RAG when answers depend on changing or proprietary facts that should be retrieved at request time. The two approaches can also work together.

Know what goes in. Own what comes out.

You can inspect real examples before training, choose the model you want to adapt, and keep portable model assets when training is complete.

Related datasets

Start fine-tuning