{
  "schema_version": "sensebench-run-v2",
  "run_id": "vllm-gemma-4-26b-a4b-fp8-a100-p001-lexen-v1-20260614",
  "created_at": "2026-06-14T13:02:48.331932+00:00",
  "git_commit": null,
  "runner": {
    "github_handle": "vassiliphilippov",
    "name": "Vassili Philippov",
    "contact": null
  },
  "dataset": {
    "dataset_id": "lexen",
    "dataset_version": "lexen-v1",
    "dataset_revision": null,
    "content_hash": "sha256:5fd4382b93f19087a1e31f6dd7d1db17c1eb17ff80fcbe1d3fdd55c0c3ecefe8",
    "item_count": 4861
  },
  "prompt": {
    "id": "p001",
    "sensebench_version": "0.1.0"
  },
  "model": {
    "kind": "self_hosted_llm",
    "display_name": "google/gemma-4-26B-A4B-it",
    "requested_model": "hosted_vllm/google/gemma-4-26B-A4B-it",
    "resolved_model": "google/gemma-4-26B-A4B-it",
    "resolved_model_counts": {
      "google/gemma-4-26B-A4B-it": 4861
    },
    "llm_vendor": "Google",
    "source_kind": "open_source",
    "license": "gemma",
    "model_url": "https://huggingface.co/google/gemma-4-26B-A4B-it",
    "reasoning_effort": "no reasoning",
    "hf_revision": "8edbb9269ec9c3faad538ee1208a07eb46051f34",
    "quantization": "fp8",
    "inference_engine": "vllm",
    "inference_engine_version": "0.22.1",
    "container_image": "vllm/vllm-openai:v0.22.1",
    "serve_command": "exec vllm serve RedHatAI/gemma-4-26B-A4B-it-FP8-Dynamic --port 8000 --served-model-name google/gemma-4-26B-A4B-it --max-model-len 8192 --gpu-memory-utilization 0.90 --max-num-seqs 256 --no-enable-prefix-caching --language-model-only --revision 8edbb9269ec9c3faad538ee1208a07eb46051f34",
    "endpoint_base_url": "http://localhost:8000/v1"
  },
  "sampling": {
    "temperature": 0.0,
    "top_p": null,
    "max_tokens": 2048,
    "seed": null,
    "extra": {}
  },
  "policy": {
    "votes_per_item": 1,
    "semantic_reasks_per_invalid_vote": 1,
    "tie_break": "earliest_vote",
    "monosemous_policy": "short_circuit"
  },
  "machine": {
    "platform": "Linux-5.15.0-160-generic-x86_64-with-glibc2.35",
    "cpu_model": "AMD EPYC 7513 32-Core Processor",
    "cpu_cores": 128,
    "ram_gib": 1385.7,
    "gpu": {
      "name": "NVIDIA A100-SXM4-80GB",
      "count": 1,
      "vram_mib_per_gpu": 81920,
      "driver_version": "570.195.03",
      "cuda_version": "13.0"
    },
    "provider": "vast.ai",
    "instance_id": "40919271",
    "hourly_rate_usd": 1.0933147044766143
  },
  "execution": {
    "concurrency": 96,
    "warmup_call_count": 8,
    "timing": {
      "benchmark_started_at": "2026-06-14T12:59:18.917199+00:00",
      "benchmark_ended_at": "2026-06-14T13:02:48.316805+00:00",
      "benchmark_seconds": 209.39959409506992,
      "setup_seconds": 2.1766286389902234
    }
  },
  "totals": {
    "item_count": 4861,
    "correct_count": 4347,
    "accuracy": 0.894260440238634,
    "call_count": 4861,
    "usage": {
      "input_tokens": 3178178,
      "cached_input_tokens": null,
      "output_tokens": 69873,
      "reasoning_output_tokens": null
    },
    "cost": {
      "currency": "USD",
      "total_usd": 0.0635943487043262,
      "input_uncached_usd": null,
      "input_cached_usd": null,
      "output_usd": null,
      "input_uncached_unit_price_usd": null,
      "input_cached_unit_price_usd": null,
      "output_unit_price_usd": null,
      "source": "machine_time_estimate"
    },
    "elapsed_seconds": 209.39959409506992
  }
}