{"data":{"kind":"file","path":"README.md","version_id":"dop0gnlgdxndfkhm8jc9m1eq","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1127,"modified_at":"2026-09-13T04:06:06.937000","content_hash":"bb234758d9ff6bc06626211b7e553401a0a22e739c749e6ef1f484f0a0740067"},"entries":[],"content":"# i3-logic\n\nSingle-turn logic and reasoning tasks (sudoku, ciphers, zebra, cryptarithm, BBH/BBEH subsets, GPQA-diamond, and more) from the `logic` subset of INTELLECT-3-RL. Each task is answered in one turn and graded by a task-specific in-process verifier resolved from the row's task name — pure Python, no sandbox or judge.\n\n## Taskset\n\n- **Source:** [PrimeIntellect/INTELLECT-3-RL](https://huggingface.co/datasets/PrimeIntellect/INTELLECT-3-RL) (`logic` subset, `train` split)\n- **Size:** 11,647 tasks (before task-name skips and difficulty filtering at load)\n\n## Changelog\n\n- 2026-09-03: Restore default solver network access by reverting the `network_allow=[]` default-deny policy introduced in #780; training rollouts need outbound network.\n- 2026-08-31: Yield task records on demand so bounded evaluations construct only the requested prefix.\n- 2026-07-10: Ported to the task-centric verifiers API: rewards and lifecycle hooks live on the `Task` (a `TaskData` row + behavior split), and task-facing config knobs moved from `--env.taskset.*` to `--env.taskset.task.*`. Requires `verifiers>=0.2.0` and Python `>=3.11`.\n","encoding":"utf-8","truncated":false,"total_bytes":1127},"status":null}