i3-code
Single-turn Python code generation tasks. The model writes a solution that is scored by running it against the task's hidden test cases (stdin/stdout or functional) inside the rollout's runtime; reward is 1.0 iff every test passes.
Taskset
- Source: PrimeIntellect/INTELLECT-3-RL (
codesubset,trainsplit) - Size: 8,579 tasks
Changelog
- 2026-09-03: Restore default solver network access by reverting the
network_allow=[]default-deny policy introduced in #780; training rollouts need outbound network. - 2026-08-31: Yield task records on demand so bounded evaluations construct only the requested prefix.
- 2026-06-25: Initial v1 taskset.