Drop7 Research

Type to search every page, record and vocabulary term.

GitHubPlay

P-SOL-3: T0 timing, all-M=6 fidelity ladder, uniform certified-cheap-engine M=6 corpus, and the G1/G2 offline gates

Full protocol: runs/RUN-20260823T191900Z-b9f8f80d/kimi-k3-psol3-design.md (P-SOL-3, Kimi K3), frozen as written; this record fixes its decision structure.

preregisteredtier: CHECKalgorithmicno-gameplaypublic-policyEX-20260824-psol3-m6-ladder-corpus-gate-2d0167ad
On this page
Linked approachThe fast engine learns reveal sampling
Theories
Dates
Created
Updated
Record idEX-20260824-psol3-m6-ladder-corpus-gate-2d0167ad

No explanation has been written for this record yet.

Technical recordThe registered protocolEX-20260824-psol3-m6-ladder-corpus-gate-2d0167ad
Hypothesis
Full protocol: runs/RUN-20260823T191900Z-b9f8f80d/kimi-k3-psol3-design.md (P-SOL-3, Kimi K3), frozen as written; this record fixes its decision structure. Stages: (T0, ~1 CPU-h, seed-free) measure continuation-duty rates for fast d1-M6, d2-M6, d3-N7M6 on 20 seed-free roots x 7 x K8 x H40 under an exclusive resource lease; re-price every stage; S0 halt with a no-run record if fewer than 6,000 corpus roots fit 18 CPU-h at the measured d1-M6 rate; if d3-M6 exceeds 1.5x 0.177 s/move the ladder drops R 64->48 with an amended power note before any ladder root is drawn. (S1 ladder) certify fast d1-M6 and d2-M6 against fast d3-N7M6 (no native arm - parity is exact per E-FAST-M6; no M<6 arm - M=1 is dead per the v2 guardrail) at the corpus operating point K8 H40 with CRN shared across siblings and engines; R=64 in cohorts 16 then 48, reusing the retained v2 1,200-root seed-free pool iff its manifest shows no M=1-conditioned selection (else regenerate ~0.2 CPU-h); certification = mean within-root Kendall tau LB95 >= 0.75 AND top-1 LB95 >= 0.80 (root bootstrap, 10,000 resamples, seed 0xb0071eaf); early stop drops an arm with tau LB95 < 0.5 after cohort 1. (Corpus) uniform labels from the cheapest certified engine on 2,048 whole origin games (lease sub-range 0xa5217000-0xa52177ff) x 7 harvested roots = 14,336 roots x 7 siblings x K8 x H40 (~17.8 CPU-h at the 2 ms/move d1-M6 estimate, re-priced at T0; ~1 GB disk, full continuations not retained); fallback if only d2 certifies: uniform d2-M6, K4 H32, ~1,300 roots, explicitly a pilot. (Training) as P-SOL v2 with two changes: SE-weighted pairs w_ij=|mu_i-mu_j|/sqrt(se_i^2+se_j^2) clipped to [0,2], and near-tie oversampling x2 for roots with label top-two gap <= 500 points capped at 50% of total loss weight; labels are KM restricted mean lifetime at H=40 with censor rates recorded. (Gate) G1/G2 unchanged from v2 on the 512-game / 4,096-root gate set (lease sub-range 0xa5219000-0xa52191ff) with exact D4 sibling values - the comparator is independent of the corpus engine.
Arms
ArmNameEntry pointManifest
Candidatepsol3-m6-cheap-engine-corpus-nnueapproaches/lifetime-objective/fast-reveal-sampling/fast-factored-search.hpp–
Comparatorincumbent played-action LeafNet and exact D1/D2/D4 orderingsapproaches/lifetime-objective/learned-leaf/train_leaf.py–
Classification
algorithmic
Information boundary
public-policy
Benchmark tier
CHECK
Lifecycle
preregistered
Theories tested
Primary metric
G1: student top-1 vs exact-D4 argmax on the 4,096-root gate set, against 0.55 absolute and incumbent+0.03 (paired McNemar LB > 0)
Secondary metrics
  • T0 measured ms/move per engine with SD; re-priced budget table
  • S1 per-engine mean within-root Kendall tau and top-1 agreement with bootstrap CIs, per cohort
  • G2 target quality (label-argmax vs D4), fit quality vs 0.7x split-half ceiling, coverage slice
  • censor rate at H=40 (preregistered check: median within-root rate > 35% raises H to 48 and re-prices x0.83; below 6,000 roots halts)
  • near-tie rate (preregistered check: outside [20%,60%] rescales oversampling to fixed effective weight, no re-labeling)
  • init-seed spread over 5 seeds; per-stage CPU-h vs the budget table
Statistical unit
root
Uncertainty method
root/origin cluster bootstrap (10,000 resamples, seed 0xb0071eaf); paired McNemar for the incumbent comparison
Data role
no-gameplay
Seed leases
  • SL-20260823T215000Z-a5216000
Whole-origin split
yes
Reuse disclosure
T0 and the ladder consume zero leased seeds (seed-free roots; the retained v2 rung-1 pool is reused only if manifest-clean). The training lease SL-20260823T215000Z-a5216000, still reserved and never opened, is opened only for the corpus (0xa5217000-0xa52177ff) and gate set (0xa5219000-0xa52191ff) after S1 certifies an engine; gate-set origins become development-read on use and never train.
Pass criteria
  1. T0/S0: at least 6,000 corpus roots affordable within 18 CPU-h at measured rates.
  2. S1: at least one of d1-M6/d2-M6 certified (tau LB95 >= 0.75 AND top-1 LB95 >= 0.80 vs fast d3-N7M6).
  3. G1: student top-1 >= 0.55 AND >= incumbent + 0.03 (McNemar LB > 0); marginal band [0.52,0.55) proceeds to a d4s5-only screen as in v1/v2.
  4. G2 reported in every outcome; label-argmax >= 0.60 required to call the target right.
On pass
Register the gameplay successor (P-SOL-v1 section 8 unchanged: blend tuning on reusable-development cohorts, 256-game paired screens at d4s5/d4s7 with rarity and futility stops, plus the D3 N7M6 deployment arm) before any gameplay seed is opened.
On fail
S0: no-run record. S1 double failure: valid negative refuting depth-second-order at K8 H40; the d3-M6 small-corpus and scale-out routes are named, not run. G1 fail: routed by G2 exactly as v2 (target wrong vs fit wrong vs coverage wrong).
Gate fixed before controlled data
yes
Resources
Wall seconds21600
CPU threads26
Max host bytes34359738368
Max GPU bytes17179869184
GPU devicesgfx1151 (training only)
Stop conditions
  1. S0 as above.
  2. S1 early stop: drop an arm at tau LB95 < 0.5 after cohort 1; both dropped halts the experiment.
  3. Censor and near-tie preregistered checks as in metrics.
  4. Any stage exceeding 150% of its re-priced budget halts for coordinator review before the 2.3 CPU-h reserve is spent; the reserve never buys extra roots and never repeats a failed gate.
  5. Budget to the offline gate: 35.0 CPU-h cap per the design's table (T0 1.0, S1 2.4 + <=7.0, corpus roots 1.0, labels 17.8, training 2.0, gate 1.5, reserve 2.3).
Expected artifacts
  • runs/RUN-20260823T191900Z-b9f8f80d/kimi-k3-psol3-design.md
  • runs/<psol3-run>/t0/timing.json
  • runs/<psol3-run>/ladder/{cohort1,cohort2}.json and certification.json
  • runs/<psol3-run>/corpus/ summary
  • runs/<psol3-run>/gate/g1-g2.json
Amendments
none recorded
Agent contextHow to extend this record

To add a reader-facing explanation, write web/content/research/EX-20260824-psol3-m6-ladder-corpus-gate-2d0167ad.mdx; it renders above this record on the next request. The registered protocol itself is in the technical record above.

Record file: research/experiments/EX-20260824-psol3-m6-ladder-corpus-gate-2d0167ad.json, validated against research/schemas/experiment-v1.schema.json. Protocol hash: b132397899ef701b7335e3095c70855d538013036772217793d96f4a43d7e11b.

Registered by Claude Code / claude-fable-5.