Qwen3.5-27B · scored

n-gram k=4

reference ngram judged submitted 10-23 16:36 UTC

Lossless gate failed: sampled replies are more or less likely under the target than the stock engine's. SpecScore 0.

SpecScore0.000geometric mean of the three speedups

solo1.25× one lane; passes 0.97, 1.38, 1.25

team0.61× team lanes busy; passes 0.61, 0.58, 0.67

busy0.66× KV cache nearly full; passes 0.66, 0.59, 0.73

Lossless gate

what each check is →

Check	Measured	Limit
device memory	162.2 GiB	≤ 165.2 GiB	pass
fixed texts' log-probabilities	0.011	≤ 0.017	pass
sampled replies' likelihood	-0.1384 ± 0.0546	within ±0.02 at 95%	fail
share of first choices	+0.0242 ± 0.0121	within ±0.003 at 95%	fail
greedy agreement	0.171	none on this task	n/a
reply lengths	–	1 ± 0.05 at 95%	pass

Speculation

Workload	Tokens / verify step	Speedup
solo	2.27	1.25
team	2.23	0.61
busy	2.19	0.66

Speculator: none
Engine: uploaded with the submission

Everything the judge recorded

{
 "ratios": {
  "solo": {
   "passes": [
    0.9682280530255188,
    1.3755768909202668,
    1.2507875608798378
   ],
   "ratio": 1.2507875608798378
  },
  "team": {
   "passes": [
    0.6117581528411916,
    0.5805712112708303,
    0.6704269997018997
   ],
   "ratio": 0.6117581528411916
  },
  "busy": {
   "passes": [
    0.6605919925549087,
    0.5927833136222279,
    0.7313040214261721
   ],
   "ratio": 0.6605919925549087
  }
 },
 "specscore": 0.7965851539265519,
 "greedy_agreement": 0.17062816403507672,
 "likelihood": {
  "gap": -0.1383933118854142,
  "se": 0.02783365896533421,
  "candidate": -0.6188419346285234,
  "stock": -0.48044862274310923
 },
 "argmax": {
  "gap": 0.02420134511637803,
  "se": 0.006169234238388912,
  "candidate": 0.15366005845584613,
  "stock": 0.1294587133394681
 },
 "fidelity": 0.0107374798445607,
 "lengths": {
  "solo": {
   "ratio": 1.0631152412849663,
   "low": 0.5333479670573782,
   "high": 1.5928825155125543
  },
  "team": {
   "ratio": 0.7947758024696878,
   "low": 0.557584708140958,
   "high": 1.0319668967984177
  },
  "busy": {
   "ratio": 0.8141641272537549,
   "low": 0.572087961772141,
   "high": 1.0562402927353687
  }
 },
 "memory": {
  "ok": true,
  "peak_mib": 166114.0,
  "allowed_mib": 169119.1
 },
 "acceptance": {
  "solo": 2.268160029498525,
  "team": 2.2337280085624456,
  "busy": 2.19380121125757
 },
 "limits": {
  "greedy_agreement_floor": null,
  "likelihood_bound": 0.02,
  "argmax_bound": 0.003,
  "fidelity_ceiling": 0.017,
  "length_margin": 0.05,
  "memory_slack_mib": 4096
 },
 "reason": "sampled replies are more or less likely under the target than the stock engine's",
 "score": 0.0
}