Qwen3.5-27B · scored

MTP k=3

reference mtp judged submitted 10-23 08:29 UTC

Passed every gate.

SpecScore1.336geometric mean of the three speedups

solo1.74× one lane; passes 1.74, 1.91, 1.55

team1.36× team lanes busy; passes 1.52, 1.36, 1.34

busy1.01× KV cache nearly full; passes 1.20, 0.99, 1.01

Lossless gate

what each check is →

Check	Measured	Limit
device memory	161.5 GiB	≤ 165.2 GiB	pass
fixed texts' log-probabilities	0.013	≤ 0.02	pass
sampled replies' likelihood	-0.0155 ± 0.0728	within ±0.05 at 95%	pass
share of first choices	+0.0005 ± 0.0155	within ±0.01 at 95%	pass
greedy agreement	0.501	none on this task	n/a
reply lengths	–	1 ± 0.05 at 95%	pass

Speculation

Workload	Tokens / verify step	Speedup
solo	3.45	1.74
team	3.21	1.36
busy	3.06	1.01

Speculator: none
Engine: uploaded with the submission

Everything the judge recorded

{
 "ratios": {
  "solo": {
   "passes": [
    1.741380187085606,
    1.9086301786642677,
    1.552344122127689
   ],
   "ratio": 1.741380187085606
  },
  "team": {
   "passes": [
    1.5247457393237156,
    1.35639398524987,
    1.3365145800784035
   ],
   "ratio": 1.35639398524987
  },
  "busy": {
   "passes": [
    1.2003751237197713,
    0.9891232395566902,
    1.0102061963447386
   ],
   "ratio": 1.0102061963447386
  }
 },
 "specscore": 1.3362770020647952,
 "greedy_agreement": 0.5013839637570825,
 "likelihood": {
  "gap": -0.015491351971005152,
  "se": 0.03713565690745868,
  "candidate": -0.5073671184145867,
  "stock": -0.49187576644358155
 },
 "argmax": {
  "gap": 0.0005287911896301178,
  "se": 0.007906602150099045,
  "candidate": 0.1315890028849277,
  "stock": 0.13106021169529758
 },
 "fidelity": 0.0125198151439468,
 "lengths": {
  "solo": {
   "ratio": 0.8905585508058926,
   "low": 0.5833602601866454,
   "high": 1.1977568414251398
  },
  "team": {
   "ratio": 1.101256035953066,
   "low": 0.8042274760837631,
   "high": 1.398284595822369
  },
  "busy": {
   "ratio": 0.924187431399389,
   "low": 0.7087002374348272,
   "high": 1.1396746253639507
  }
 },
 "memory": {
  "ok": true,
  "peak_mib": 165416.0,
  "allowed_mib": 169119.1
 },
 "acceptance": {
  "solo": 3.4549578226961644,
  "team": 3.2051407525690885,
  "busy": 3.0621626042859478
 },
 "limits": {
  "greedy_agreement_floor": null,
  "likelihood_bound": 0.05,
  "argmax_bound": 0.01,
  "fidelity_ceiling": 0.02,
  "length_margin": 0.05,
  "memory_slack_mib": 4096
 },
 "reason": null,
 "score": 1.3362770020647952
}