{"batch_policy":{"choice_candidates":16,"code_test_workers":4,"generation_examples":8,"padded_token_budget":8192},"batch_timings":[{"candidates":16,"kind":"choice","padded_tokens":304,"seconds":0.1297435329761356},{"candidates":16,"kind":"choice","padded_tokens":416,"seconds":0.015189048019237816},{"candidates":16,"kind":"choice","padded_tokens":544,"seconds":0.016184214036911726},{"candidates":16,"kind":"choice","padded_tokens":800,"seconds":0.023662249906919897}],"dataset_revision":"8d51e7e4887a4caaa95b3fbebbf53c0490b58bbb","device":"cuda:0","evaluator_source":true,"evidence":"CodeSOTA execution on White","execution_profile":"batched-v2","finished":1792198193,"gpu":1,"id":"efa7055f4d6480dd542a3abf","measured_cost_usd":null,"metric":"accuracy","model":"Qwen/Qwen2.5-0.5B","model_revision":"060db6499f32faf8b98477b0a26969ef7d8b9987","n":32,"notes":"Pilot subset, CodeSOTA batched-v2 execution. Separate scope from serial v1; unchanged prompt/tokenization/metric contracts.","peak_gpu_bytes":1036140544,"profile":"pilot-32","protocol":{"adapter":"sentiment","area":"Language","benchmark":"SST-2","contract":"Movie review \u2192 positive / negative","dataset":"stanfordnlp/sst2","dataset_config":null,"dataset_revision":"8d51e7e4887a4caaa95b3fbebbf53c0490b58bbb","dataset_url":"https://huggingface.co/datasets/stanfordnlp/sst2","direction":"max","id":"sentiment","license":["unknown"],"limit":32,"metric":"accuracy","models":[{"id":"Qwen/Qwen2.5-0.5B","revision":"060db6499f32faf8b98477b0a26969ef7d8b9987"},{"id":"Qwen/Qwen2.5-1.5B","revision":"8faed761d45a263340a0528343f099c05c9a4323"}],"name":"Sentiment classification","protocol_version":"codesota-pilot-v2-batched","runner_file":"evaluate_fast.py","source":"https://nlp.stanford.edu/sentiment/","split":"validation"},"resource_metrics":{"basis":"Evaluator subprocess elapsed \u00d7 one reserved GPU; includes data/model loading and CPU scoring; not kernel-active time","gpu_count":1,"gpu_slot_seconds":8.02642539597582,"measurement_version":"gpu-slot-v1"},"runner_sha256":"bc899fb160f6881408863cd8bcdf96102dd6d8b10559a5f781a8cd816103da77","scope":"1985c666f87bc1b3203e78c666ffc282afe39e7e4d54d3a5359a3ccc34bfb465","score":0.9375,"seconds":1.1125338861020282,"seed":42,"selection":"first 32 valid examples in pinned split order","status":"completed","task":"sentiment","worker":"white-gpu1"}