FreeBot Dataset

Live site: https://freequillbot.netlify.app

Author: Vatsal K. (alias VatsalDaPro)

Training and evaluation data for FreeBot — a free, self-hosted paraphrasing tool built as a personal project. The goal: replicate the core behavior of tools like QuillBot's paraphraser using small open models fine-tuned locally, rather than a commercial API.

Why this dataset exists

Zero-shot prompting of small instruct models (Qwen2.5-0.5B/1.5B, and others tested in an initial 6-model bake-off) only ever produced synonym-swapping — near-identical sentence structure with a few words substituted. It never did genuine structural restructuring. Fixing that required fine-tuning on real paraphrase pairs, which meant assembling a dataset that:

• Covers 4 distinct paraphrase modes: Natural, Formal, Simple, Shorter — each with genuinely different rewriting behavior, not just a different prompt on the same rewrite.
• Spans 3 length tiers: small (single sentence), medium (paragraph), large (multi-paragraph).
• Doesn't leak into the held-out evaluation set used to measure fine-tune quality.

How it was built

1. Dataset survey — surveyed a wide range of candidate public paraphrase datasets. Most were rejected for wrong shape (detector-style, not generation pairs), too-high word overlap (built for paraphrase detection rather than generation — useless for teaching restructuring), semantic drift, or eval-set leakage.
2. Real-corpus mining — for the sources that passed, applied a length-dependent "genuinely shorter/different" heuristic (character-count based, tiered by input length) plus spaCy NER-based entity-loss detection to catch dropped named entities.
3. LLM-assisted generation — used a local 4B-class quantized model to generate additional Formal/Simple pairs from real source sentences (not synthetic prompts), tiered by length, filtered inline for: overlap ratio, content-word retention (info-loss detection), refusal-phrase blocklist, and question-mark-preservation (task-non-compliance detection — catches the model answering a question instead of paraphrasing it).
4. Hand-written + user-curated pairs — supplemented underrepresented modes with manually written and manually reviewed examples.
5. Dedup + leak-check — every source deduped internally, against each other, and against the frozen held-out eval set (eval_set.json, 167 examples, never used in training).

Files

• training_set.json — the final assembled training set used for fine-tuning: 5,200 examples (Formal 2,000 / Simple 1,200 / Natural 1,000 / Shorter 1,000), each with mode, length_tier, source_text, paraphrase, source (provenance tag).
• eval_set.json — 167-example frozen held-out evaluation set. Never used as training data.
• finetune_eval_results.json — aggregate overlap-ratio results from evaluating the fine-tuned model against eval_set.json (see Results below).
• datasets/ — raw and intermediate per-source filtered pools (one file per source, plus mined/heuristic-checked intermediate files) that feed into training_set.json.
• Root-level dataset_*.json files — per-mode intermediate pools (hand-written, filtered, LLM-generated) before final assembly.

The fine-tuned model

The model trained on this dataset is published at Pixiano45/freebot_qwen0.5 (gated).

Qwen2.5-0.5B-Instruct + LoRA (rank 16, alpha 16, targeting all attention + MLP projections), trained via Unsloth for 3 epochs on the full 5,200-example set, mode-conditioned via a prompt prefix per example ("Paraphrase the following text so it sounds natural/formal/simply/concisely..."), formatted with the model's own chat template.

• 8.8M trainable params out of 502.8M total (LoRA).
• Training: 975 steps, batch size 4 × grad accum 4 (effective 16), lr 2e-4 with warmup+decay, bf16.
• Final training loss: ~1.40-1.44, converged from ~2.5 within the first 20 steps, no divergence or instability.

Results — held-out eval overlap ratio (lower = more genuine restructuring, not synonym-swapping)

Mode	Zero-shot baseline (avg)	Fine-tuned (avg / median)
Natural	~0.60-0.79	0.305 / 0.324
Formal	~0.60-0.79	0.409 / 0.414
Simple	~0.60-0.79	0.378 / 0.375
Shorter	~0.60-0.79	0.454 / 0.429

The fine-tune roughly halved the word-overlap ratio against the source text across every mode compared to zero-shot prompting of the same base model — confirming it learned genuine structural rewriting rather than word substitution. Manual review of the worst-case (highest-overlap) outputs per mode showed no hallucination or meaning drift; the highest-overlap cases were simply source sentences that were already maximally concise for their target mode (e.g. "Restart your router if the connection drops for more than a minute" has little room to restructure further without changing meaning).

Inference speed: ~45ms/token, ~1.2s average per output (unquantized bf16, single sequence, RTX 5060 Ti).

License / usage

Access to this dataset requires agreeing to the repository's usage terms before use. Contact vatsaldapro@gmail.com for permitted-use details before any reuse, redistribution, or derivative work.

---

Author: Vatsal K. (alias VatsalDaPro)