Researchers race to close zkEVM 52-bit proof gap by December

Researchers are working to close a 52.14-bit unresolved proof gap for the koalaIRS12 benchmark on Ethereum’s better.codes leaderboard. A snapshot taken at 15:44:47 UTC on Aug. 21 shows a 63.99-bit soundness lower certificate and a 116.13-bit attack upper certificate. The leaderboard reflected nine promoted submissions from seven solvers.

The better.codes contest, run by the Ethereum Foundation, operates two parallel tracks. The soundness track raises a machine-checked lower certificate by proving that an executable reduction-error bound meets the encoded target at a certified radius. The attack track lowers an upper certificate by certifying an unsafe suffix under the benchmark’s winning-set-density condition.

Each submission exports a pinned theorem and a verification harness. A comparator checks the exported statement against the challenge target and the Lean proof assistant kernel verifies the proof before promotion. The pinned environment fixes the theorem statement, parameter point and verification harness so successive submissions remain comparable.

KoalaIRS12 is a fixed parameter profile tied to an interleaved Reed–Solomon reduction used in proof-system research. The contest records a conservative 63.99-bit certified safe point for that parameter point and a 116.13-bit certified unsafe suffix. The numerical difference between the promoted certificates, 52.14 bits, is the live measure of what the contest has not resolved for that benchmark.

The Ethereum Foundation notes the challenge score is a spot-check quantity and should not be interpreted as a direct minus-log2 of whole-system soundness or as proof of production security. An accepted result proves the submitted theorem inside the pinned environment; broader production claims require component-level completeness, clear assumptions in formal definitions, implementation fidelity, composition arguments, and an accounting that ties component bounds to an auditable system package.

Ethereum’s zkEVM roadmap originally set a 128-bit provable-security target for December 2025, along with a maximum proof size of 300 KiB and a formal soundness argument for the recursion architecture. A subsequent update moved the M3 deadline to early December 2026 and aligned architecture-security deliverables with a Dec. 1 milestone. The Foundation’s public progress page, last synced Aug. 20, lists zkVM readiness and ISA compliance results and does not mark completion for the full early-December package.

Academic research published before the current leaderboard snapshot highlights related open questions for succinct proof systems. Researchers identified list decoding, Reed–Solomon proximity gaps, correlated agreement and mutual correlated agreement as topics that need further study. The koalaIRS12 certificates apply only to the encoded parameter point; other parameter choices, constructions and system components remain separate research items.

Closing the 52.14-bit interval on better.codes will require further promoted submissions on the soundness or attack tracks. Changes to the certificates will adjust the formally checkable boundary while the pinned theorem keeps successive results comparable. Additional system-level evidence will be required before the Foundation’s 128-bit provable-security target for a production zkEVM can be claimed.