Record: X-WING — Shared N-gram Tables + Cubric (val_bpb=0.5644)#800

Closed

newjordan wants to merge 257 commits intoopenai:mainfrom

newjordan:submission/xwing

This pull request is big! We're only showing the most recent 250 commits

Commits on Mar 21, 2026

Add FA3→FA2→SDPA fallback chain for pod restart resilience

Octavian
and
claude
committed
Revert FA3 fallback chain — was unauthorized code change to baseline + untested v2

Octavian
and
claude
committed
Fix FA3 NaN: cast qkv to bf16 before FA3 call, disable dynamo DDP opt

Octavian
and
claude
committed
Add 2-seed validation scripts for exp A/B/C

Octavian
and
claude
committed

Commits on Mar 22, 2026

Log exp A/B results: both behind baseline, zlib fallback bug found

Octavian
and
claude
committed
Fix XSA NaN: position 0 has no valid targets when self-mask + causal combine

Octavian
and
claude
committed
Disable XSA in ttt_only run — manual attention too slow vs FA3

Octavian
and
claude
committed
Add run_v2_ttt_noXSA.sh — TTT v2 + temp scaling, all FA3, max speed

Octavian
and
claude
committed
Restore XSA_LAST_N=3 in run_v2_ttt_only.sh (keep existing test intact)

Octavian
and
claude
committed
Log v2 TTT-only + XSA=3 result: 1.1982 BPB (worse than 1.1301 baseline)

Octavian
and
claude
committed
Strip verbose logging from v2 train loop — match baseline format

Octavian
and
claude
committed
Log v2 noXSA result: 1.1538/1.1315 BPB — TTT v2 hurt, no edge over baseline

Octavian
and
claude
committed
Log exp_a/b/c results: all worse than 1.1301 baseline, exp_c never ran

Octavian
and
claude
committed
Add exp D: TTT 8 epochs + stride 32 (eval-only improvement)

Octavian
and
claude
committed
Add SAM (Sharpness-Aware Minimization) option for TTT

Octavian
and
claude
committed
Add baseline reproduction script — verify 1.1303 on current FA3 build

Octavian
and
claude
committed
Add SAM to baseline TTT — test sharpness-aware adaptation on proven code

Octavian
and
claude
committed
Log exp D result: 1.1295 BPB — new best (-0.0008 vs baseline)

Octavian
and
claude
committed
Log exp D seed 42: 1.1307 BPB — confirms improvement (mean 1.1301)

Octavian
and
claude
committed
Add exp_d SAM variant — TTT 8ep + stride 32 + sharpness-aware TTT

Octavian
and
claude
committed
Log exp D seed 7: 1.1313 BPB but 16.18 MB — over size limit

Octavian
and
claude
committed
Add Partial RoPE + LN Scale (from PR #315) to sota254 + run_sam

Octavian
and
claude
committed
Add exp_d/run_sam_clean.sh — pure SAM A/B test, no other changes

Octavian
and
claude
committed
Log exp D seeds 7+137: both over size limit

Octavian
and
claude
committed
Add Sponge Bath experiment: TTT 8ep + stride 32 eval-only improvement
Octavian
committed
Add PR#315 clone with TTT 8ep + run script
Octavian
committed
Log D+SAM+PR315tricks: 1.1274 BPB new best; add SAM to pr315 run script
Octavian
committed
Log PR315+TTT results: 1.1240 BPB (invalidated — TTT now banned)
Octavian
committed
Add PR374 enchilada experiment: 12L/2KV/2.75xMLP + train@1024 + EMA

Octavian
and
claude
committed
Add RunPod 8xH100 setup guide — every gotcha we've hit 3 times

Octavian
and
claude
committed
Add fractal cadence experiment: F/N alternation pattern

Octavian
and
claude
committed
Add PR374-safe: EMA + earlier QAT + longer warmdown only, shape unchanged

Octavian
and
claude
committed
Add PR374-depth: 12L/4KV/2.625xMLP (same params, +1 layer) + EMA + QAT + warmdown

Octavian
and
claude
committed
Add minified train_gpt (30KB vs 74KB) with EMA+SWA edge

Octavian
and
claude
committed
Fix log0 keyword arg bug in minified script

Octavian
and
claude
committed
Fix Block kwarg: layer_idx -> li in minified script

Octavian
and
claude
committed
Fix a.te -> a.tie_embeddings in minified script

Octavian
and
claude
committed
Add pr374_submit: trimmed winner for submission

Octavian
and
claude
committed
Remove duplicate stride=64 eval from submit script

Octavian
and
claude
committed
Revert "Remove duplicate stride=64 eval from submit script"
Octavian
committed
Add fractal cadence H100 script: 4 unique × 3 loops + ortho positions

Octavian
and
claude
committed
Add fractal cadence long run (1.6575 BPB @ 3929 steps) + H100 script

Octavian
and
claude
committed
Add pr374_slim: comment-stripped pr374_safe (67KB vs 74KB, -6.4KB)

Octavian
and
claude
committed
Add fractal cadence auto-research sweep runner

Octavian
and
claude
committed
Fix FA3 head_dim limit: 12 heads / 6 kv_heads for dim=768 (head_dim=64)

Octavian
and
claude
committed
Add pr374_ttt: PR374 + EMA + TTT(20ep,SAM) — rock the house

Octavian
and
claude
committed
Fix early QAT trigger: skip warmdown/QAT for first 100 steps

Octavian
and
claude
committed
Optimize fractal: tuned H100 defaults + focused sweep grid

Octavian
and
claude
committed
Add TTT eval: test-time training on already-graded tokens

Octavian
and
claude
committed
Log depth + TTT results: 1.1223 BPB (12L+TTT20ep+SAM)

Octavian
and
claude
committed
Save pr374_throne: our #1 non-TTT record holder (1.1243 BPB)

Octavian
and
claude
committed
Add autonomous overnight fractal optimizer (Karpathy auto-research)

Octavian
and
claude
committed
Rewrite autoresearch: Qwen-guided fractal optimizer via Ollama

Octavian
and
claude
committed
Add 3 leapfrog variants based on PR#414 SOTA (1.1233 BPB)

Octavian
and
claude
committed
Add pod setup script for leapfrog variants (v1/v2/v3)

Octavian
and
claude
committed
Fix setup_pod.sh: apply FA3 build fixes from RUNPOD_SETUP.md

Octavian
and
claude
committed
Update fractal H100 with overnight Qwen findings

Octavian
and
claude
committed
Fix v1 TTT burst: apply EMA first, then QAT-aware sharpening

Octavian
and
claude
committed
Add inner-TTT to fractal eval: recursive weight improvement per loop

Octavian
and
claude
committed
v1 combo: burst+QAT before EMA + 15 GPTQ percentiles

Octavian
and
claude
committed
Isolate TTT to fractal layers only: blocks + loop_pos + skip_weights

Octavian
and
claude
committed
Add TTT drift gate: snap block weights back toward trained originals

Octavian
and
claude
committed
Add v4: burst + distill + train_seq_len=1024 for more steps

Octavian
and
claude
committed
Add SOTA autoresearch: Qwen-guided edge finding for 1.1233 record

Octavian
and
claude
committed
Fix TTT graph bug + upgrade to 3 blocks x 896d (23M params, ~14MB)

Octavian
and
claude
committed
Save leapfrog experiment results: 6 variants tested, v1 wins at 1.12319

Octavian
and
claude
committed
Fractal v2: dim=960 (25.7M, ~15.8MB), Muon LR back to 0.025

Octavian
and
claude
committed
Disable TTT in v2 (backward graph bug), focus on bigger model

Octavian
and
claude
committed
Add v5: QAT percentile fix + TrigramHash + EMA-SWA blend

Octavian
and
claude
committed
Fix TTT graph bug: fresh vc + detached inputs per loop

Octavian
and
claude
committed
Two v2 scripts: with TTT and without, for A/B comparison

Octavian
and
claude
committed
Fractal v3: MLP 3.0->3.3 fills 16MB budget (27.4M params, ~15.2MB)

Octavian
and
claude
committed
Fractal v4: 1.5x batch (1.18M tokens/step) to use spare VRAM

Octavian
and
claude
committed
Log The Frugendorff: fractal cadence baseline at 1.2113 BPB

Octavian
and
claude
committed
Add v6: fractal 6L×2 loops (12 effective) + 480d/16H/4xMLP

Octavian
and
claude
committed
Frugendorff v5: TTT warmup-then-freeze to capture 1.19 peak

Octavian
and
claude
committed
Log all Frugendorff results: v3 TTT peak 1.1901, v4 batch 1.2186

Octavian
and
claude
committed
Fix v6: 512d/16H/8KV (head_dim=32, FA3 requires multiple of 8)

Octavian
and
claude
committed
Log v5/v6 results + add fractal sweep script

Octavian
and
claude
committed
Frugendorff 8-hour single GPU longrun

Octavian
and
claude
committed

Commits on Mar 23, 2026

Complete Frugendorff documentation: all runs, findings, architecture analysis

Octavian
and
claude
committed
Add v7: legal score-first TTT eval (PR #461 recipe)

Octavian
and
claude
committed
Add v8: mutual learning — flat GPT + fractal GPT co-training

Octavian
and
claude
committed
v7: add TTT early stopping — train only first 60 chunks

Octavian
and
claude
committed
Update v6 to sweep winner: 4×3/512d/8H/4KV/4xMLP

Octavian
and
claude
committed
Log v7 TTT results: early stop at 60 chunks = 1.12312 (best)

Octavian
and
claude
committed
The Frugendorff Squared: 6L x 2 loops, dim=640, MLP 4x, 28.5M params

Octavian
and
claude
committed
Log higher LR (0.030) + MTP results — both worse

Octavian
and
claude
committed
The Frugendorff Squared: 1.1478 BPB — 0.025 from world lead

Octavian
and
claude
committed
Stack distillation onto Frugendorff Squared

Octavian
and
claude
committed
Prep Frugendorff Squared PR submission (draft — needs image)

Octavian
and
claude
committed
Add train script to Frugendorff submission folder

Octavian
and
claude
committed
TTT stabilization: EMA scoring + fixed cosine LR + embed freeze

Octavian
and
claude
committed
Add GPTQ quantization: Hessian-aware error compensation for int6

Octavian
and
claude
committed
Stack low-hanging fruit: fix GPTQ, earlier QAT, matched clipping

Octavian
and
claude
committed
Add post-quant training burst: repair quant damage before eval

Octavian
and
claude
committed
Remove post-quant burst — illegal (uses training data during eval)

Octavian
and
claude
committed
Add selective int8 for sensitive layers (GPTQ-int8 upgrade path)

Octavian
and
claude
committed
Frugendorff V3: v7 full stack + fractal loops (GPTQ + TTT + PQB)

Octavian
and
claude
committed
Add AdamW TTT option — PR #462 shows 5x better TTT gain vs SGD

Octavian
and
claude
committed
Frugendorff V3: SwiGLU + AdamW TTT + 8K bigram (PR#462 techniques)

Octavian
and
claude
committed
Fix DDP unused params: add find_unused_parameters=True

Octavian
and
claude
committed
SwiGLU + GPTQ: PR #462 architecture with our quant improvements

Octavian
and
claude
committed
Add OptRot: Hadamard rotation before GPTQ quantization

Octavian
and
claude
committed
Frugendorff V3: Star-ReLU SwiGLU + VRL + MLP 4.0 (fits 16MB)

Octavian
and
claude
committed
Log session results: 1.0763 SwiGLU (size problem), 1.1215 v7 (submitted)

Octavian
and
claude
committed
Frugendorff V4: GEPA-matched config for 16MB compression

Octavian
and
claude
committed
Frugendorff V4 A/B: 5x2 and 4x2 variants for H100 calibration

Octavian
and
claude
committed
Add short TTT experiment: no-EMA, 50 chunks, SGD

Octavian
and
claude
committed
Add XSA-all(11) + short TTT run script

Octavian
and
claude
committed
Record: v7 GPTQ + Short TTT — val_bpb 1.1207 (seed 1337)

Octavian
and
claude
committed
Frugendorff compression shim on SwiGLU SOTA (1.0763 model)

Octavian
and
claude
committed
Add v7 smooth run: proper warmdown + XSA-all

Octavian
and
claude
committed
Reduce bigram buckets 2048→1792 for XSA-11 size fit

Octavian
and
claude
committed
Frugendorff stacked: LeakyReLU(0.5)² + VRL + decoder_lr_mult=2.0

Octavian
and
claude
committed
Add research batch script + GPTQ env var tuning

Octavian
and
claude
committed
Sanitize: remove all illegal TTT from Frugendorff scripts

Octavian
and
claude
committed
Add Frugendorff SwiGLU sweep: 9 configs across 3 batches

Octavian
and
claude
committed
EMERGENCY: Disable all illegal TTT across entire project

Octavian
and
claude
committed
CRITICAL: Remove illegal TTT from train_gpt_swiglu.py

Octavian
and
claude
committed
Add SwiGLU Frugendorff compression sweep (6 configs)

Octavian
and
claude
committed
SwiGLU F1: VRL + LeakyReLU(0.5)² + legal TTT + int8 attn.proj + seq2048

Octavian
and
claude
committed
Fix DDP: find_unused_parameters=True for VRL block 0

Octavian
and
claude
committed
Add GPTQ re-quantization sweep (no training, ~2 min)

Octavian
and
claude
committed
Save Gold Standard (GS): v7 GPTQ 1.1206 BPB — 3 copies

Octavian
and
claude
committed
Restore original train_gpt_swiglu.py, new F1 variant as separate file

Octavian
and
claude
committed
Log SwiGLU F1 run: 1.1208 BPB (20.6MB, needs compression)

Octavian
and
claude
committed
Add TTT calibration sweep: 11-config grid to close 1.11 BPB gap

Octavian
and
claude
committed
Add GPTQ re-quant sweep for GS v7 — no retraining needed

Octavian
and
claude
committed
Submission: XSA-11 + GPTQ block64/percdamp002

Octavian
and
claude
committed

Commits on Mar 24, 2026

Add int5 sizing sweep — test how many params fit at int5

Octavian
and
claude
committed
Int5 test: 33.6M params (MHA 8/8, MLP 3.5x) with int5 GPTQ

Octavian
and
claude
committed
Add 576+ single-file runner and aggressive edge test launchers
Octavian
committed
Add TTT sweep infra + Vast.ai pipeline + int5mix run record

Octavian
and
claude
committed
Add RunPod setup script for TTT calibration sweep

Octavian
and
claude
committed
Add Micro Crawler architecture — asymmetric fractal with flat+crawler split

Octavian
and
claude
committed
Update micro crawler to balanced 4f+2cx2 config at dim=640

Octavian
and
claude
committed
Add crawler cadence + per-loop GPTQ to H100 micro crawler

Octavian
and
claude
committed
Cadence N/N/N/N/C + activation-aware per-loop GPTQ blending

Octavian
and
claude
committed
Pre-flight: backup scripts + pod setup for micro crawler 8xH100

Octavian
and
claude
committed
Recursive cadence: ramp N count as LR warms down

Octavian
and
claude
committed
Frugendorff v2: trigram hash + lr_mul compile-spike fix

Octavian
and
claude
committed
Fix lr_mul compile spike + shrink dim 640→624 for artifact budget

Octavian
and
claude
committed
Freeze run1 scripts — 1.1377 sliding window (broken LR, dim=640)

Octavian
and
claude
committed
Fix dim 624→630 (must be divisible by num_heads=10)

Octavian
and
claude
committed
Fix dim 630→620 (head_dim must be even for RoPE)

Octavian
and
claude
committed
Fix dim 620→560 (FA3 needs head_dim multiple of 8: 560/10=56)

Octavian
and
claude
committed
Add quantfix training script: learned scales, earlier QAT, selective int8

Octavian
and
claude
committed
Run2: GPTQ Hessian quant + dim=640 + trigram_vocab=2048

Octavian
and
claude
committed
Launcher: dim=640, trigram_vocab=2048 (fit budget, test lr_mul fix only)

Octavian
and
claude
committed
Keep trigram_vocab=8192 to match run1 — isolate lr_mul fix only

Octavian
and
claude
committed
Run3: Self-referential crawler with deliberation gate

Octavian
and
claude
committed
Fix DDP unused param error: touch gate on normalize steps

Octavian
and
claude
committed
Run3: push dim to 720 for max capacity

Octavian
and
claude
committed
Run4: self-referential crawler at dim=720

Octavian
and
claude
committed
Run5: persistent deliberation — gate fires every step

Octavian
and
claude
committed
Run6: run1 base (1.1377) + persistent deliberation + GPTQ

Octavian
and
claude
committed
Fix run6 launcher: dim 624→640

Octavian
and
claude
committed
Run7: run1 base + GPTQ only, no gate — the safe play

Octavian
and
claude
committed
Run8: PD + GPTQ + fixed cadence 2 — keep EMA fresh

Octavian
and
claude
committed
Run8: bidirectional PD — consensus_ref as learned Parameter

Octavian
and
claude
committed
Save micro crawler sweep results + experiment log

Octavian
and
claude
committed
Science log: all micro crawler H100 results + findings

Octavian
and
claude
committed
Cadence ablation science: H1 + H2 fronts, 8 arms, RunPod-ready

Octavian
and
claude
committed
Fix: total_mem → total_memory for PyTorch 2.9+

Octavian
and
claude
committed
Add diagnostic training script with polar blending + cadence support

Octavian
and
claude
committed
H3 hypothesis: per-block cadence gradient (funnel/diamond/inverse)

Octavian
and
claude
committed
H1: cadence 0 (no C-steps) at full 1.0 scale — the critical test

Octavian
and
claude
committed
H2: inverse architecture 2f+4cx2 — skinny head, deep recursion

Octavian
and
claude
committed
Run 8 production script with zero C-steps — clean A/B vs Run 8

Octavian
and
claude
committed
Science log: H1+H2 cadence sweep results + cad0 full-scale data

Octavian
and
claude
committed
DEFINITIVE: cad0 beats Run 8 on production script (1.1325 vs 1.1355)

Octavian
and
claude
committed
H4: crawler bank on GS v7 U-Net — best trick on fastest car

Octavian
and
claude
committed
H4 results: crawler bank learns better per step but loses on sliding

Octavian
and
claude
committed
f1: add PR587 baseline plus low-rank correction and stable distill
Octavian
committed
F1 v1: accuracy profile — corr rank 256 + distillation

Octavian
and
claude
committed
F1 v2: accuracy profile + crawler bank at U-Net bottleneck

Octavian
and
claude
committed
H4 F/G: fast small model (8L/384d) crawler bank test

Octavian
and
claude
committed
Organize all Frugendorff research: ablations, scripts, results

Octavian
and
claude
committed
f1: add legal leaderboard profile and activation knob
Octavian
committed
f1: add isolated sota xsa4+bg1536 speed-safe concept
Octavian
committed
Skiptrace: crawler bank fires every 10 steps, injects decaying delta

Octavian
and
claude
committed
Cubric: temporal weight sharing via skiptrace — research lab

Octavian
and
claude
committed
records: log f1 legal-lb xsa4/bg1536 1.1195 candidate run
Octavian
committed
SOTA: F1 legal LB profile — 1.1195 (seed 1337), 1.1200 (seed 42)

Octavian
and
claude
committed
Update Frugendorff PR draft with full research findings + warning labels

Octavian
and
claude
committed
Reframe PR: recursion is challenged, not dead — per-step benefit is real

Octavian
and
claude
committed
NEW RECORD: seed 2045 hits 1.1190 post-TTT — best single seed

Octavian
and
claude
committed
H4: bottleneck crawler scripts for H100 pod

Octavian
and
claude
committed
f1: create sota garage with three baseline cars
Octavian
committed
Fix H4 Arm C crash: remove PD gate, use simple sequential looping

Octavian
and
claude
committed

Commits on Mar 25, 2026

Add F1 architecture triplet run script
Octavian
committed
H5-H8: next research hypotheses — skiptrace, trigram, noisy QAT, weight sharing

Octavian
and
claude
committed
Add isolated legal 5-gram eval A/B for car02
Octavian
committed
Fix hashed n-gram eval probability safety
Octavian
committed
Add compile stability switches for isolated n-gram eval
Octavian
committed
car02: add budgeted ngram eval cutoff and partial reporting
Octavian
committed
Distribute n-gram eval across all 8 GPUs — ~8x speedup

Octavian
and
claude
committed
research: standardize hypothesis tagging across garage and experiments
Octavian
committed
Cubric: visualization concept — model internals as art

Octavian
and
claude
committed
Alpha sweep: eval-only script, no retraining needed

Octavian
and
claude
committed
Alpha sweep eval script + SDPA/GQA fallback for non-FA3 GPUs

Octavian
and
claude
committed
Fix: move GPTQ calibration inside training phase

Octavian
and
claude
committed
Multi-order backoff (2-7) + entropy-adaptive alpha + GPTQ fix

Octavian
and
claude
committed
Backoff 7-gram run script + results prep

Octavian
and
claude
committed
Cubric n-gram accumulator: online alpha adaptation from n-gram reliability

Octavian
and
claude
committed
Cubric n-gram: full A/B sweep with all eval-time variants + moonshot prep

Octavian
and
claude
committed
Moonshot: neural alpha head — learned n-gram interpolation weight

Octavian
and
claude
committed
Cubric cadence accumulator: N/N/N/C pattern for n-gram table optimization

Octavian
and
claude
committed
Podracing II: Electric Bugaloo — 0.9620 BPB (best seed), mean 0.9823

Octavian
and
claude
committed
Cubric cadence C arm: single run script for cadence=10 balanced test

Octavian
and
claude
committed
Cubric cadence B arm: single run script for cadence=4 aggressive test

Octavian
and
claude
committed
Podracer garage: SOTA train_gpt.py + run.sh × 3 safety copies

Octavian
and
claude
committed
Fix: revert SOTA, cubric on separate copy only

Octavian
and
claude
committed
Add n-gram parameter sweep + cubric lite eval variant

Octavian
and
claude
committed
Add podracer_red safe experiment lane
Octavian
committed
Fix corrupted SOTA: restore real PR train_gpt.py (147bbccc)

Octavian
and
claude
committed
Add Vast.ai 8xH100 n-gram sweep script

Octavian
and
claude
committed
Podracer lanes: green (cubric lite), purple (clean experimental), frozen backups

Octavian
and
claude
committed
Add podracer green2 lane (cubric lite copy for iteration)

Octavian
and
claude
committed
Racing profiles: green + purple with aggressive n-gram params

Octavian
and
claude
committed
podracer_red: add cubric-lite racing profile
Octavian
committed
podracer_red: add dedicated racing profile run scripts
Octavian
committed
Podracing III submission prep: cubric lite 0.9362 BPB (seed 2045)

Octavian
and
claude
committed
Green2: maxed cubric — floor=0.05, cap=4.0, adapt=1.05/0.95, alpha_max=0.80

Octavian
and
claude
committed
Podracing III: Cubric Lite — 0.9362 BPB (3-seed mean)

Octavian
and
claude
committed
Fix: add cubric_cadence to Hyperparameters class in green + green2

Octavian
and
claude
committed
Fix green2: add COMPILE_FULLGRAPH=0 to match working runs

Octavian
and
claude
committed

Commits on Mar 26, 2026

xwing_red: add pod-ready PR779 test launch script
Octavian
committed
X-WING: shared n-gram tables + cubric — podracer engine with PR#779 insight

Octavian
and
claude
committed
Move X-WING to concepts/xwing (beside podracer, not inside)

Octavian
and
claude
committed
Add X-wing-Red_1 alpha-head variant from podracer_green2
Octavian
committed
X-WING eval speedup: bincount + batch_seqs 128

Octavian
and
claude
committed
Speed up X-wing Red 1 eval: bincount + batch_seqs 128

Octavian
and
claude
committed
X-WING submission: 0.5641 BPB 2-seed mean (new world record)

Octavian
and
claude
committed
Merge remote-tracking branch 'upstream/main' into submission/xwing
Octavian
committed
X-WING: 0.5644 BPB 3-seed mean — new world record

Octavian
and
claude
committed
X-WING v2: per-order entropy gating + cubric (targeting #798)

Octavian
and
claude
committed
Add eval-only script for fast 2x2 ablation grid

Octavian
and
claude
committed
Add 2x2 ablation grid script for cubric × per-order centers

Octavian
and
claude
committed
X-WING Brown: shared tables + per-order entropy gating, no cubric

Octavian
and
claude
committed
X-WING Fast: safe speed boosts, no quality loss

Octavian
and
claude
committed
X-WING Brown II: brown per-order gating + safe speed boosts

Octavian
and
claude
committed
X-WING Yellow: 2D cubric (order × entropy_bin) pattern recognizer

Octavian
and
claude
committed
X-WING Yellow II: 3D cubric monster + orders 2-9

Octavian
and
claude
committed
X-WING Yellow II: 3D cubric + complementary training — THE MONSTER

Octavian
and
claude
committed
X-WING Yellow III: warm-start cubric at proven converged values

Octavian
and
claude
committed
X-WING Yellow IV: UNCHARTED — ceiling 2.5 + 16M buckets + orders 2-10

Octavian
and
claude
committed
X-WING Yellow II submission: 0.4896 BPB — 3D cubric + complementary training

Octavian
and
claude
committed
X-WING 3D Cubric: 0.4820 BPB (3-seed mean, std 0.0002)

Octavian
and
claude
committed