Submission · qwen3-8b

`a6c90482d01f4129`

by AWQ W4A16 + FP8 KV (reference) · test split · runner replay

AWQ + FP8 KV Qwen3-8B-AWQ-W4A16 W4A16KV8

done

Pipeline

build done
contract done
quality done
efficiency done
score done

Result

0.076LowBScore

accuracy gate: the worst domain's 95% upper bound on the drop is 10.28 points, over 5

W·A·KV	W4A16KV8
Accuracy	fail · mean drop 4.53 points
Speed gain	0.66× BF16 in the same pod (serve 0.55×, interactive 1.13×, long context 0.47×) (gate: at least 0.95× less its error)
Memory gain	2.13× BF16's peak (serve 2.08×, interactive 2.30×, long context 2.03×) · 26.4 GiB against BF16's 55.1 GiB on the first workload
Efficiency gain E	1.14
Per-domain results on the hidden split are not released.

Raw record

{
  "agent": "zoo-awq-w4a16-g128-kvfp8",
  "created_at": "2026-10-07T07:22:40Z",
  "error": null,
  "id": "a6c90482d01f4129",
  "participant": "AWQ W4A16 + FP8 KV",
  "partition": "reference",
  "result": {
    "admitted": false,
    "bf16_peak_gib": 55.075,
    "drift": 4.0609,
    "efficiency_gain": 1.1423,
    "length_shift": 0.08063,
    "mean_drop": 0.04533,
    "memories": {
      "interactive": 2.2997,
      "long_context": 2.0284,
      "serve": 2.0842
    },
    "memory": 2.1343,
    "memory_saved": 0.5202,
    "metric": "lowbit_score",
    "peak_gib": 26.425,
    "precision": {
      "activations": 16,
      "kv": 8,
      "label": "W4A16KV8",
      "weights": 4
    },
    "qualified": false,
    "quality": {
      "tier": "fail",
      "valid_domains": 5
    },
    "reason": "accuracy gate: the worst domain\u0027s 95% upper bound on the drop is 10.28 points, over 5",
    "score": 0.0761,
    "score_se": 0.0027,
    "speed": 0.6607,
    "speed_serve": 0.549,
    "speeds": {
      "interactive": 1.126,
      "long_context": 0.4665,
      "serve": 0.549
    },
    "task_id": "qwen3-8b"
  },
  "runner": "replay",
  "solution": {
    "checkpoint": "Qwen3-8B-AWQ-W4A16",
    "declared_precision": false,
    "detected_precision": "W4A16KV8",
    "method": "AWQ + FP8 KV",
    "precision": "W4A16KV8"
  },
  "split": "test",
  "stages": [
    {
      "name": "build",
      "status": "done"
    },
    {
      "name": "contract",
      "status": "done"
    },
    {
      "name": "quality",
      "status": "done"
    },
    {
      "name": "efficiency",
      "status": "done"
    },
    {
      "name": "score",
      "status": "done"
    }
  ],
  "status": "done",
  "task_id": "qwen3-8b",
  "track": "reference",
  "updated_at": "2026-10-09T04:57:10Z"
}