Qwen3.5-27B · scored

n-gram k=4

reference ngram judged submitted 10-23 08:29 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

solo1.03× one lane; passes 1.03, 0.96, 1.04

team0.61× team lanes busy; passes 0.72, 0.61, 0.47

busy0.70× KV cache nearly full; passes 0.67, 0.70, 0.79

Lossless gate

what each check is →

Check	Measured	Limit
device memory	162.0 GiB	≤ 165.2 GiB	pass
fixed texts' log-probabilities	0.0096	≤ 0.017	pass
sampled replies' likelihood	-0.0958 ± 0.0559	within ±0.02 at 95%	fail
share of first choices	+0.0193 ± 0.0117	within ±0.005 at 95%	fail
greedy agreement	0.177	none on this task	n/a
reply lengths	–	1 ± 0.05 at 95%	pass

Speculation

Workload	Tokens / verify step	Speedup
solo	1.99	1.03
team	2.19	0.61
busy	2.10	0.70

Speculator: none
Engine: uploaded with the submission

Everything the judge recorded

{
 "ratios": {
  "solo": {
   "passes": [
    1.0270096491284757,
    0.961892593762147,
    1.0421624361572153
   ],
   "ratio": 1.0270096491284757
  },
  "team": {
   "passes": [
    0.715984354610189,
    0.6061782200121195,
    0.46604860552735644
   ],
   "ratio": 0.6061782200121195
  },
  "busy": {
   "passes": [
    0.6699211275843489,
    0.7013618497082287,
    0.7893000649332897
   ],
   "ratio": 0.7013618497082287
  }
 },
 "specscore": 0.7586456937320034,
 "greedy_agreement": 0.17678779176607948,
 "likelihood": {
  "gap": -0.09583684354997424,
  "se": 0.028531052189067693,
  "candidate": -0.5956034926616349,
  "stock": -0.4997666491116607
 },
 "argmax": {
  "gap": 0.0193046873466918,
  "se": 0.005981825249639686,
  "candidate": 0.14976969117471292,
  "stock": 0.13046500382802112
 },
 "fidelity": 0.009576503208492149,
 "lengths": {
  "solo": {
   "ratio": 0.8417700552799744,
   "low": 0.4084173941365923,
   "high": 1.2751227164233565
  },
  "team": {
   "ratio": 1.185201104310158,
   "low": 0.7118961158985322,
   "high": 1.6585060927217838
  },
  "busy": {
   "ratio": 0.944999378678594,
   "low": 0.7106663218029403,
   "high": 1.1793324355542476
  }
 },
 "memory": {
  "ok": true,
  "peak_mib": 165878.0,
  "allowed_mib": 169119.1
 },
 "acceptance": {
  "solo": 1.9890583903208838,
  "team": 2.185207456404089,
  "busy": 2.0998419872854885
 },
 "limits": {
  "greedy_agreement_floor": null,
  "likelihood_bound": 0.02,
  "argmax_bound": 0.005,
  "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
}