Reverse auction for AI data

1. The buyer posts a specification, such as a dataset size, domain and quality bar.
2. Qualified sellers submit prices.
3. Bids fall until sellers stop undercutting.
4. The buyer awards the contract to the lowest qualifying bid.

Fits a buyer that knows what it needs and wants several vendors to compete, such as sourcing expert data from labs.

Example. A lab posts a 20,000-example coding-data requirement with a quality rubric and takes bids from five vendors.

• RLHF and preference data
• Supervised fine-tuning data
• Expert demonstration data
• Reinforcement learning environments
• Code training data
• Legal training data
• Speech and audio data
• Multimodal training data
• Evaluation sets

First-price sealed-bid auction

A first-price sealed-bid auction is a one-round auction in which buyers submit private bids and the highest bidder pays the amount they bid.

Posted-price sale

A posted-price sale is a fixed-price offer in which the seller names a price and any buyer can accept it.

rubric

A rubric is a written set of criteria that tells reviewers how to score or compare a model output.

reserve price

A reserve price is the minimum price at which a seller will complete an auction.

AI data auction

An AI data auction is a competitive sale in which buyers bid, or sellers undercut, to set the price and terms for training or evaluation data used by AI models.