{
  "schema_version": "sensebench-run-v2",
  "run_id": "vllm-gemma-4-e4b-fp8-h100-p002-lexen-v1-20260616-thinking",
  "created_at": "2026-06-16T12:07:56.371195+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": "p002",
    "sensebench_version": "0.1.0"
  },
  "model": {
    "kind": "self_hosted_llm",
    "display_name": "google/gemma-4-E4B-it",
    "requested_model": "hosted_vllm/google/gemma-4-E4B-it",
    "resolved_model": "google/gemma-4-E4B-it",
    "resolved_model_counts": {
      "google/gemma-4-E4B-it": 4861
    },
    "llm_vendor": "Google",
    "source_kind": "open_source",
    "license": "gemma",
    "model_url": "https://huggingface.co/google/gemma-4-E4B-it",
    "reasoning_effort": "reasoning",
    "hf_revision": "fee6332c1abaafb77f6f9624236c63aa2f1d0187",
    "quantization": "fp8",
    "inference_engine": "vllm",
    "inference_engine_version": "0.22.1",
    "container_image": "vllm/vllm-openai:v0.22.1",
    "serve_command": "exec vllm serve google/gemma-4-E4B-it --port 8000 --served-model-name google/gemma-4-E4B-it --gpu-memory-utilization 0.90 --max-num-seqs 256 --no-enable-prefix-caching --language-model-only --max-model-len 16384 --reasoning-parser gemma4 --default-chat-template-kwargs \\{\\\"enable_thinking\\\":\\ true\\} --quantization fp8 --dtype bfloat16 --revision fee6332c1abaafb77f6f9624236c63aa2f1d0187",
    "endpoint_base_url": "http://localhost:8000/v1"
  },
  "sampling": {
    "temperature": 0.0,
    "top_p": null,
    "max_tokens": 4096,
    "seed": null,
    "extra": {}
  },
  "policy": {
    "votes_per_item": 1,
    "semantic_reasks_per_invalid_vote": 1,
    "tie_break": "earliest_vote",
    "monosemous_policy": "short_circuit",
    "shuffle_senses_per_vote": false
  },
  "machine": {
    "platform": "Linux-6.8.0-57-generic-x86_64-with-glibc2.35",
    "cpu_model": "AMD Ryzen 9 7950X 16-Core Processor",
    "cpu_cores": 32,
    "ram_gib": 187.8,
    "gpu": {
      "name": "NVIDIA H100 80GB HBM3",
      "count": 1,
      "vram_mib_per_gpu": 81559,
      "driver_version": "570.124.06",
      "cuda_version": "13.0"
    },
    "provider": "vast.ai",
    "instance_id": "41180527",
    "hourly_rate_usd": 1.6027777777777776
  },
  "execution": {
    "concurrency": 96,
    "warmup_call_count": 8,
    "timing": {
      "benchmark_started_at": "2026-06-16T12:04:22.257832+00:00",
      "benchmark_ended_at": "2026-06-16T12:07:56.369268+00:00",
      "benchmark_seconds": 214.11143363825977,
      "setup_seconds": 1.4045317005366087
    }
  },
  "totals": {
    "item_count": 4861,
    "correct_count": 4049,
    "accuracy": 0.8329561818555853,
    "call_count": 4861,
    "usage": {
      "input_tokens": 1197352,
      "cached_input_tokens": null,
      "output_tokens": 2293038,
      "reasoning_output_tokens": null
    },
    "cost": {
      "currency": "USD",
      "total_usd": 0.09532584661209557,
      "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": 214.11143363825977
  }
}