Forge-AIFORGE-AI

Last updated: 2026-09-04

FINE-TUNING · 2026

FINEFORGE

Fine-tuning datasets, forged from video & papers.

The expensive step of a fine-tune is not training: it is gathering the examples. Almost every tool in the space assumes you already have the documents. FineForge takes care of the step before — finding them and deciding which ones are worth it.

In production, with real users: fineforgeai.com

What it does

You describe the domain in plain language. FineForge searches YouTube, arXiv and OpenAlex, a model scores every candidate from 0 to 100 against what you asked for, you approve or discard them one by one, and what survives comes out as training-ready JSONL plus one PDF per source.

StageWhat happensWhat you control
SearchYouTube, arXiv and OpenAlex from a single promptThe prompt and the sources
ScoringEvery candidate gets a 0–100 relevance scoreThe threshold
CurationYou approve or reject each one by handEverything
ExportJSONL ready to train, PDF summaries per sourceThe format

Who it is for

Teams that fine-tune models on a specific domain and are spending days digging through videos and papers by hand. Solo researchers with the same problem.

What it is not

Facts you can check

StatusIn production, with real users
SourcesYouTube, arXiv, OpenAlex
OutputJSONL plus PDF summaries
URLfineforgeai.com
Built byForge-AI, 2026