Qwen3.5-27B · scored

Native MTP k3

GPT-5.6 Sol agent mtp judged submitted 10-23 17:01 UTC

Passed every gate.

SpecScore1.437geometric mean of the three speedups

solo1.98× one lane; passes 1.57, 2.04, 1.98

team1.34× team lanes busy; passes 1.46, 1.16, 1.34

busy1.12× KV cache nearly full; passes 0.91, 1.19, 1.12

Lossless gate

what each check is →

Check	Measured	Limit
device memory	162.1 GiB	≤ 165.2 GiB	pass
fixed texts' log-probabilities	0.012	≤ 0.017	pass
sampled replies' likelihood	-0.0023 ± 0.0536	within ±0.02 at 95%	pass
share of first choices	-0.0005 ± 0.0117	within ±0.003 at 95%	pass
greedy agreement	0.524	none on this task	n/a
reply lengths	1.034 [0.849, 1.219]	1 ± 0.05 at 95%	pass

Speculation

Workload	Tokens / verify step	Speedup
solo	3.40	1.98
team	2.82	1.34
busy	3.13	1.12

Speculator: none
Engine: a directory of its workspace

Everything the judge recorded

{
 "ratios": {
  "solo": {
   "passes": [
    1.5729296230782082,
    2.042683606761007,
    1.9833691286906487
   ],
   "ratio": 1.9833691286906487
  },
  "team": {
   "passes": [
    1.4570004918088248,
    1.16233063008515,
    1.3381667225948632
   ],
   "ratio": 1.3381667225948632
  },
  "busy": {
   "passes": [
    0.9052997892439413,
    1.1948009272304405,
    1.1183627989125773
   ],
   "ratio": 1.1183627989125773
  }
 },
 "specscore": 1.437139178490833,
 "greedy_agreement": 0.5238271252861955,
 "likelihood": {
  "gap": -0.0022680644143345807,
  "se": 0.027364811301834806,
  "candidate": -0.49902196491427336,
  "stock": -0.4967539004999388
 },
 "argmax": {
  "gap": -0.0005187275185098705,
  "se": 0.005949552441696272,
  "candidate": 0.12870744282656602,
  "stock": 0.12922617034507589
 },
 "fidelity": 0.011715580721919349,
 "lengths": {
  "solo": {
   "ratio": 1.1829541137236692,
   "low": 0.7950796225727255,
   "high": 1.570828604874613,
   "paired": true,
   "replies": 139
  },
  "team": {
   "ratio": 1.008791805169256,
   "low": 0.7187907786315186,
   "high": 1.2987928317069932,
   "paired": true,
   "replies": 297
  },
  "busy": {
   "ratio": 1.015621897133779,
   "low": 0.7352712701549119,
   "high": 1.295972524112646,
   "paired": true,
   "replies": 279
  },
  "pooled": {
   "ratio": 1.034025399391606,
   "low": 0.8486379542712008,
   "high": 1.2194128445120112,
   "paired": true,
   "replies": 715
  }
 },
 "memory": {
  "ok": true,
  "peak_mib": 165952.0,
  "allowed_mib": 169119.1
 },
 "acceptance": {
  "solo": 3.3954604620891637,
  "team": 2.815361244142335,
  "busy": 3.1251869446643794
 },
 "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": null,
 "score": 1.437139178490833
}