{
  "caveats": [
    "Bounded tuning pilot; incomplete window growth cannot pass the full workload grade. Lab retrieval and rolling-history evidence only. No summaries. Shared engine cache intervals include overlapping users and are not per-user hits. Gateway path, not authenticated document ingestion or customer acceptance.",
    "Weights file SHA-256 was not recorded for public reproduction."
  ],
  "engine": {
    "name": "NVIDIA NGC vLLM",
    "recipe": "plain",
    "version": "0.29.0+5013de39.dev"
  },
  "format": "olir/0.1",
  "history": {
    "corrections": [],
    "supersedes": []
  },
  "id": "042d9bd08ed94402bc2085e5",
  "measured_at": "2026-10-07T16:31:48.247200+00:00",
  "metrics": [
    {
      "basis": null,
      "n": null,
      "name": "roll_ttft_max_seconds",
      "scope": "per_person",
      "source_field": "roll_ttft_max_seconds",
      "statistic": "max",
      "unit": "s",
      "value": 4.037177833961323
    },
    {
      "basis": null,
      "n": null,
      "name": "requests_ok",
      "scope": "machine_total",
      "source_field": "requests_ok",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "turn_number",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].turn_number",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 20229
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 3.962868105038069
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 71.94677230783611
    },
    {
      "basis": null,
      "n": null,
      "name": "content_tokens_added",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].content_tokens_added",
      "statistic": "single",
      "unit": "tokens",
      "value": 20000
    },
    {
      "basis": null,
      "n": null,
      "name": "cumulative_content_tokens",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].cumulative_content_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 20000
    },
    {
      "basis": null,
      "n": null,
      "name": "dropped_messages",
      "scope": "per_person",
      "source_field": "rolling_context_turns[0].dropped_messages",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "engine_interval_cache_hit_rate",
      "scope": "machine_total",
      "source_field": "rolling_context_turns[0].engine_interval_cache_hit_rate",
      "statistic": "single",
      "unit": "ratio",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "user_number",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].user_number",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "turn_number",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].turn_number",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "prompt_tokens",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].prompt_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 20288
    },
    {
      "basis": null,
      "n": null,
      "name": "ttft_s",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].ttft_s",
      "statistic": "single",
      "unit": "s",
      "value": 4.037177833961323
    },
    {
      "basis": null,
      "n": null,
      "name": "decode_tps",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].decode_tps",
      "statistic": "single",
      "unit": "tok/s",
      "value": 72.27615355001669
    },
    {
      "basis": null,
      "n": null,
      "name": "content_tokens_added",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].content_tokens_added",
      "statistic": "single",
      "unit": "tokens",
      "value": 20000
    },
    {
      "basis": null,
      "n": null,
      "name": "cumulative_content_tokens",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].cumulative_content_tokens",
      "statistic": "single",
      "unit": "tokens",
      "value": 40000
    },
    {
      "basis": null,
      "n": null,
      "name": "dropped_messages",
      "scope": "per_person",
      "source_field": "rolling_context_turns[1].dropped_messages",
      "statistic": "single",
      "unit": "count",
      "value": 2
    },
    {
      "basis": null,
      "n": null,
      "name": "engine_interval_cache_hit_rate",
      "scope": "machine_total",
      "source_field": "rolling_context_turns[1].engine_interval_cache_hit_rate",
      "statistic": "single",
      "unit": "ratio",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "roll_ttft_p50_seconds",
      "scope": "per_person",
      "source_field": "roll_ttft_p50_seconds",
      "statistic": "median",
      "unit": "s",
      "value": 4.037177833961323
    },
    {
      "basis": null,
      "n": null,
      "name": "rolled_off_abstained",
      "scope": "machine_total",
      "source_field": "rolled_off_abstained",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "rolled_off_invalid",
      "scope": "machine_total",
      "source_field": "rolled_off_invalid",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "rolled_off_wrong",
      "scope": "machine_total",
      "source_field": "rolled_off_wrong",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "normal_ttft_max_seconds",
      "scope": "per_person",
      "source_field": "normal_ttft_max_seconds",
      "statistic": "max",
      "unit": "s",
      "value": 3.962868105038069
    },
    {
      "basis": null,
      "n": null,
      "name": "rolled_off_correct",
      "scope": "machine_total",
      "source_field": "rolled_off_correct",
      "statistic": "single",
      "unit": "count",
      "value": 0
    },
    {
      "basis": null,
      "n": null,
      "name": "normal_ttft_p50_seconds",
      "scope": "per_person",
      "source_field": "normal_ttft_p50_seconds",
      "statistic": "median",
      "unit": "s",
      "value": 3.962868105038069
    },
    {
      "basis": null,
      "n": null,
      "name": "rolled_off_total",
      "scope": "machine_total",
      "source_field": "rolled_off_total",
      "statistic": "single",
      "unit": "count",
      "value": 1
    },
    {
      "basis": null,
      "n": null,
      "name": "requests_total",
      "scope": "machine_total",
      "source_field": "requests_total",
      "statistic": "single",
      "unit": "count",
      "value": 2
    }
  ],
  "model": {
    "artifact_sha256": [],
    "publisher_repo": "Qwen/Qwen3.6-35B-A3B",
    "publisher_revision": "995ad96eacd98c81ed38be0c5b274b04031597b0",
    "variant": "NVFP4 variant",
    "weights_dtype": "NVFP4",
    "weights_format": "safetensors"
  },
  "outcome": {
    "gates": [
      {
        "met": false,
        "name": "requests",
        "observed": "1/2"
      }
    ],
    "reasons": [
      "1/2 requests passed."
    ],
    "status": "fail"
  },
  "publication_group": "pre-standard",
  "published": null,
  "publisher": {
    "name": "Skilak Keep"
  },
  "quality": [],
  "record_date": "2026-10-07",
  "reproduce": {
    "driver": null,
    "fixture_sha256": null,
    "harness": {
      "name": "keep-rolling-context-v1",
      "version": "1"
    },
    "harness_sha256": null,
    "os": null,
    "workload_name": null,
    "workload_version": null
  },
  "requests": [],
  "scenario": {
    "id": "rolling-context",
    "version": "1"
  },
  "settings": {
    "concurrency": 1,
    "configured_context_tokens": 262144,
    "context_tokens": 262144,
    "max_output_tokens": 256
  },
  "setup_complete": false,
  "setup_id": "47a03fd9bbfe612c42f992e302c0e2fe",
  "standard": "pre-standard",
  "system": {
    "driver": null,
    "hardware": "NVIDIA GB10, 128 GB unified memory",
    "os": null
  },
  "workload": {
    "cache": null,
    "content_tokens": null,
    "document": null,
    "input_tokens": null,
    "output_tokens": 256,
    "repetitions": null,
    "users": 1,
    "warmup": null
  }
}