Predictive AI Evaluation Competition 2026
Stanford AIMS AI Measurement Science
Overview
Build a program that predicts the probability that an unobserved AI response is correct.
LLM coding assistants are permitted during development; disclosure in the technical report is requested.
Web search during evaluation is prohibited.
Notice dates
- Notice closing date
- 2026-11-06
Calendar dates from the saved notice. They do not establish a deadline time, time zone or current acceptance status.
Eligibility
Each person may join only one team; cross-institution teams are encouraged.
Age and country restrictions were not verified in the body text.
Teams may have up to 6 members.
Each person may join only one team; cross-institutional teams are encouraged.
Age and country restrictions were not confirmed in the page content.
Cautions and verification notes
The final code and technical report must be provided, and the selected Codabench submission must be identified.
The Sydney presentation session on 12/11 is not the competition submission deadline. Onsite participation fees and prizes are unconfirmed.
Official overview and body checked; the organizer’s official overview body was read. Where the sources below include rules, those rules were also checked. No account access, actual entry, data download or agreement to rules was performed.
These requirements concern the AI/ML task and do not establish blanket permission to use generative AI assistance.
The official body reviewed in this check does not list prizes. This is not extended to a claim of 0 KRW cash prizes or that all organizer materials omit prizes.
Sources
Check the original notice for current rules. English includes editorial renderings of stored notices and organizer names, not verified official English names. Original records are preserved. Excerpts may be incomplete; consult the linked sources for full conditions.