Explicit Edit Benchmark v1

The benchmark as published under the contract explicit-edit-v1.

• 226 deterministic exact-edit tasks with a byte-exact verifier
• ready adapters named after their published harness family
• submissions go to the Hugging Face Dataset, where every observation records this contract, the task-set hash, and the verifier hash

This release is the stable pointer for that contract: the dataset names it, and the source archive below is what a reader can fetch later.