BMS

Start typing to search the publication.

Local EVM Verify Pilot · Figure 01

Local EVM Verify Pilot v1 — result plate

A fully local Qwen3.8-27B (Q4_K_M) agent writing executable proofs of contract vulnerability across three conditions and three curated tasks. Feasibility pilot, one run per cell; not an official EVMbench score.

Back to all figures

A fully local Qwen3.8-27B (Q4_K_M) agent writing executable proofs of contract vulnerability across three conditions and three curated tasks. Feasibility pilot, one run per cell; not an official EVMbench score.
Collection in progress: 0 of 54 runs terminal. Verified-proof cells are pending and are never shown as zero.Deterministic evaluator in disposable, network-isolated Docker containers; not official EVMbench grading.
Academic PNGAcademic SVG 1200×675 social card Read the full pilot