Reasoning & Test-Time Compute · 2025

Absolute Zero: Reinforced Self-play Reasoning with Zero Data

Andrew Zhao, Gao Huang

A single model proposes and solves its own verifiable tasks with no human-curated data, reaching SOTA math and code reasoning and removing the human-question bottleneck in RLVR.

Editorial record

Plain-language summary

A single model proposes and solves its own verifiable tasks with no human-curated data, reaching SOTA math and code reasoning and removing the human-question bottleneck in RLVR.

Knowledge graph

Relationships

Descendants

Source record

Provenance

Record ID
P-644
Record created
2026-07-13
Last reviewed
2026-07-14
Record version
2

Citation caveat: Citation metadata is approximate and marked unverified in the source dataset.