{"data":{"kind":"file","path":"README.md","version_id":"uv4tyd6m0bov4iado1cic7an","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1210,"modified_at":"2026-10-04T16:50:11.433000","content_hash":"4c288b2af4d1468d5e0ec73fea318aa60936ca4aebc7e47affc92c09967fcc19"},"entries":[],"content":"# supernode-rediscovery\n\nRediscover causal Qwen3-4B circuits for named semantic goals.\n\nThis is a small Hub distribution for the `supernode-rediscovery` taskset. It depends on an\nversion-pinned public `circuit-supernodes==0.1.1` wheel, which owns\nall implementation modules. The distribution adds no duplicate plugin code.\n\nInstall with `prime env install wazupsteve/supernode-rediscovery`. In Verifiers v1, select\n`taskset.id = \"wazupsteve/supernode-rediscovery@0.1.2\"` and use the `circuit-null` harness for\ntasksets. All Bio-Circuit modules are installed with the shared core package.\n\nPublic data: [amitprakash2005/bio-circuit](https://huggingface.co/datasets/amitprakash2005/bio-circuit).\nFull setup: [circuit-supernodes](https://app.primeintellect.ai/dashboard/environments/wazupsteve/circuit-supernodes).\n\n**Live tools and scoring require an external CUDA GPU backend.** This release\nhas 16 evaluation tasks per taskset and no training split or validated SFT\ntrajectories. Downloading and inspecting the data works without a GPU. Follow\nthe core package README to run your own backend; no author's private credentials\nare required. This release does not supply a hosted GPU endpoint or container image.\n","encoding":"utf-8","truncated":false,"total_bytes":1210},"status":null}