# OpenAI — Improving language model behavior by training on a curated dataset

- Company: OpenAI (openai.com)
- Announced: 2021-06-10T07:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/improving-language-model-behavior
- Record: https://forck.live/items/932-improving-language-model-behavior-by-training-on-a-curated-dataset
- Subject: GPT / ChatGPT / API

OpenAI research finds that fine-tuning on a small, curated dataset can improve language model behavior with respect to specific behavioral values.

## Evidence

Verbatim from https://openai.com/index/improving-language-model-behavior:

> Our latest research finds we can improve language model behavior with respect to specific behavioral values by fine-tuning on a small, curated dataset.

---

Record: https://forck.live/items/932-improving-language-model-behavior-by-training-on-a-curated-dataset
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md