Home › Datasets › Conversation & Social
Conversation & Social Fine-Tuning Dataset
Social skills, conversation, etiquette, communication
- 932 instruction-response pairs
- Format JSONL
- Domain social
- Dataset $19 one-time
$19 for the dataset. Training is priced separately by model, starting at $47.
What is in this dataset
932 instruction-response pairs covering social skills, conversation, etiquette, communication. Each pair is a realistic question with a full written answer, 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.
Hey, how was your weekend?
It was great, thanks for asking! I spent Saturday hiking with some friends and Sunday just relaxing at home. How about you?
Sorry, I'm running a bit late for our meeting. Traffic is terrible.
No worries at all! Thanks for letting me know. We can start without you and catch you up when you get here. Drive safely!
I've been feeling really overwhelmed with work lately.
That sounds really tough. I'm sorry you're going through that. Do you want to talk about what's been piling up? Sometimes just sharing can help lighten the load.
Who this is for
Conversational assistants that need natural, warm everyday dialogue.
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:
- Qwen — Qwen3.5 (27B, 9B, 4B), Qwen3, Qwen2.5, Qwen2.5-Coder
- Gemma — Gemma 4, Gemma 3, Gemma 2
- Phi — Phi-4, Phi-3
- DeepSeek — R1 Distill (70B, 32B, 14B)
- Mistral and Llama families
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
- Pick this dataset and a base model.
- Rebiha runs the job on a managed GPU. You set up nothing.
- 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 style and structure without prompting for it every time. This dataset teaches the shape of a good answer in this domain, not facts to look up.
Related datasets
- SHTF: Emergency Preparedness & Survival — 878 pairs
- Business & Finance — 956 pairs
- Legal DIY — 821 pairs
- E-commerce & Product Descriptions — 948 pairs