# Merge 01 — batch r5-cg-continue (items k01–k05)

This merges the five item results under `.imd/reads/artifacts/item-k01` … `item-k05`. Each item ran long training of CG's 19-8-2 head on the replica, starting from CG's frozen weights, and then checked the result on the real engine. Every number below is copied from the item result named in its row. I computed nothing new. Before merging I checked two things: each file's sha256 matches the value the brief gives, and each item's `code_zb64` decompresses to a source whose sha256 equals its stated `code_sha256`. Those values also appear in the JSON below. None of the five results has a `blocked` field, so `blocked` is `null` in every row.

## Summary

| key | markets/s | search min / gens | holdout over_CG (se) | real vs_CG mean (se, n) | real vs_IB mean (se, n) | afterSwap peak | OOG | gates A/B/C | candidate |
|---|---|---|---|---|---|---|---|---|---|
| k01 | 2246.0 (8 procs; 4909.7 on 30) | 42.22 / 1276 | +0.1101 (0.0046) | +0.1147 (0.0179, 512) | +0.3160 (0.0204, 1024) | 237626 | 0 | pass/pass/pass | new weights, code 0ab25ecd… |
| k02 | 239.6 | 72.66 / 70 | +0.0077 (0.0020) | +0.0176 (0.0083, 512) | +0.1799 (0.0215, 1024) | 235478 | 0 | pass/pass/pass | new weights, code 02b4f502… |
| k03 | 412.3 (4 procs) | 49.49 / 150 | 0.0 (CG unchanged) | 0.0 (identity, 512) | +0.1683 (0.0199, 1024) | 237819 | 0 | pass/pass/pass | CG unchanged, e1ea8fcb… |
| k04 | 313.4 | 76.21 / 97 | 0.0 (CG unchanged) | 0.0 (identity, 512) | +0.1708 (0.0201, 1024) | 237848 | 0 | pass/pass/pass | CG unchanged, e1ea8fcb… |
| k05 | 86.7 | 0 / 0 (no search) | 0.0 (CG unchanged) | 0.0 (identity, 976) | +0.1819 (0.0224, 976) | 235599 | 0 | pass/pass/pass | CG unchanged, e1ea8fcb… |

**Best: k01.** It is valid, passes all three gates, and has the highest holdout over_CG: +0.110 se 0.005 on 8,192 replica markets, seeds 40000000..40008191. On the real engine it gained +0.115 se 0.018 over CG (n 512, seeds 8000000..8000511). That agrees with the replica holdout. The k01 result also reports +0.960 se 0.032 over HA (n 512). That is still far below the roughly +3.1 over HA the goal says the top 10 needs. None of the items claims a leaderboard placement.

## Notes and disagreements

- **Three of the five items made no new policy.** In k03 (150 generations) and k04 (97 generations), no point on the monitor beat CG, so each froze CG unchanged. k05 measured the replica at 86.7 markets/s, below the 200 markets/s cutoff its brief set, so it ran no search at all. For these three items, `vs_CG` and holdout `over_CG` are exactly 0 because the candidate is identical to CG, not because anything was measured. Their real-engine rows are fresh CG-vs-IB and CG-vs-HA baselines on new seeds.
- **k02 found only a small gain.** It ran 70 generations with population 32 and sigma0 0.06. Its holdout gain over CG is +0.0077 se 0.0020 and its real gain is +0.0176 se 0.0083. Both are positive but small. k02 measured its own replica speed at 239.6 markets/s.
- **Replica speeds disagree a lot between items:** 86.7 to 412 markets/s for k02–k05 against 2,246 (8 procs) for k01. The items ran on different machines and core counts (k03 says it had 4 cores; k01 says its machine was shared and used 30 processes). This explains why k01 could run 1,276 generations while the others ran 0 to 150. k01's step size (sigma0 0.13) was also larger than k02's and k03's (0.06) and k04's (0.10). k03 and k04 note that their small steps may have kept them from finding a gain; k03 marks that as an untested suggestion.
- **Gate B is identical in all five items:** replica CG−T +1.9204 se 0.0445 against a real reference of 2.0, n 512, seeds 7232768..7233279.
- **Gate A agrees across items:** max fee difference 8.81e-05 bps. Gate C for the new candidates: k01 7.78e-05 bps, k02 8.15e-05 bps. Both are under 0.1 bps.
- **CG's holdout over_T differs slightly.** k01, k04 and k05 report 1.9519172921936434. k03 reports 1.9519157504253348, with se 0.011938024 against 0.011938005. The difference is about 1.5e-6, which is negligible but not bit-identical.
- **over_IB is not a separate measurement.** Every item derives holdout over_IB as over_T − 1.7690, the reference value its brief supplied. No item simulated an IB arm on the holdout.
- **Real-engine samples differ.** Each item used its own fresh seed block: k01 8000000.., k02 8004000.., k03 8008000.., k04 8012000.., k05 8016000... k05 cut its samples to 976 per arm (512 for HA) because of time. In k01, `means.Candidate` covers n 1088: 64 extra seeds (8001024..8001087) from chunk overshoot are included in that mean, but k01 says they are excluded from every paired comparison.
- **Gas:** every candidate arm reports 0 OOG and an afterSwap peak between 235,478 and 237,848, all within the 245,000 limit.

## Sources that did not hold on re-read

The plane flagged some cited sources that did not hold when re-read: k01 (1), k02 (1), k03 (2), k04 (3), k05 (3). Those items' `sources.json` files were not handed to this merge. I therefore cannot tell which URL or passage failed, and I could not re-check them here. Note that k04's result says its own check passed (`checks.live_source_quotes: {passed: true, n: 8}`), which conflicts with the plane's later re-read.

As far as the result files show, the cited sources are the upstream artifacts (HA, T, replica, port, KIT, CG, IB on api.imdnotdead.fun) and the challenge repository. The items use those only for provenance hashes. Every item states that its speed, gate, holdout, real-engine and gas figures are its own local measurements, not quotes. So the flags cast doubt on the provenance claims (that the artifact files at those URLs say what was quoted), not on the measured numbers. The sha256 values each item reports for those files agree across all five items.

## Merged JSON

Numbers are copied as given. `markets_per_s` is each item's primary speed figure: k01's 8-process figure, k03's 4-process figure. `holdout.se` collects each item's se for over_T and over_CG; every item's over_IB se equals its over_T se. In every row, `afterSwap_peak` and `OOG` belong to the candidate arm; for k05 that is the CG arm, which is the candidate. `settings` is copied verbatim from each item's `search.settings`. `best.code_zb64` is base64 of zlib-compressed Solidity.

```json
{"rows":[
{"key":"k01","result_sha256":"2023d343e67559907a8da0c8b178d0e52ade6dae29064ff3b373da650755b5fa","blocked":null,"markets_per_s":2246.0,"search_minutes":42.22,"generations":1276,"settings":{"method":"pycma CMA-ES over 178 head weights (W1 19x8, b1, W2 2x8, b2); multipliers/offsets fixed","start":"CG params (frozen policy), not restart","CMA_stds":"CG search_settings.CMA_stds per group (W1 0.16007, b1 0.18565, W2 0.12726, b2 0.11859)","popsize":32,"sigma0":0.13,"sigma_cap":"sigma <= 0.13*0.5^(minutes/40), enforced after each tell","es_seed":601,"markets_per_generation":256,"markets_rule":"speed >= 2000 markets/s so 2x128; same 256 markets for every proposal; contiguous blocks from 1500000000","fitness":"mean replica edge","proposals_rounded":1e-06,"monitor":"30000000..30002047 every 10 generations, ES mean and generation best, paired gain over CG","selected":"best@1140","selected_monitor_gain_over_CG":0.10515063363556332},"holdout":{"over_T":2.0619789717380543,"over_CG":0.11006167954441087,"over_IB":0.29297897173805443,"se":{"over_T":0.012180134631977608,"over_CG":0.004554952728974956}},"real":{"vs_IB":{"n":1024,"mean_diff":0.3159530124817927,"se":0.020365042092018816,"sum":323.5358847813557,"sumsq":536.6785526911397},"vs_CG":{"n":512,"mean_diff":0.11469920783846532,"se":0.017890078101032988,"sum":58.725994413294245,"sumsq":90.47242718636709}},"afterSwap_peak":237626,"OOG":0,"gates":{"A":true,"B":true,"C":true},"all_gates_passed":true,"valid":true,"code_sha256":"0ab25ecd54f9a816119546408b08906ede14f74faa99d9c10cdfec7d6d6c1be6"},
{"key":"k02","result_sha256":"a906727c81bdc99bc28f8e049c9df319578a04a8f23075a6fb42aab4f5f6318c","blocked":null,"markets_per_s":239.63327721379457,"search_minutes":72.6625479358668,"generations":70,"settings":{"population":32,"seed":602,"sigma0":0.06,"CMA_stds":{"W1":0.16007274507182634,"b1":0.18564858000307327,"W2":0.12726224434596164,"b2":0.1185932470281719,"rule":"0.1*rms of HA W1 (13 old cols)/b1/W2/b2; new W1 columns start at 0 with W1 std"},"train_n":512,"train_start":1510000000,"monitor":[30000000,30002047],"processes":8,"sigma_ceiling_rule":"0.06 * 0.5 ** (training_minutes / 40); clamped before ask, after tell, and after final monitor overhead","start":"CG frozen params; never restart","common_markets_per_proposal":true,"monitor_every_generations":10,"monitor_points":["ES mean","generation best"]},"holdout":{"over_T":1.9595954862417346,"over_CG":0.007678194048091155,"over_IB":0.19059548624173472,"se":{"over_T":0.011961602860798191,"over_CG":0.0019865584367904725}},"real":{"vs_IB":{"n":1024,"mean_diff":0.17985469109448576,"se":0.02145433032747639,"sum":184.17120368075342,"sumsq":515.2999735035669},"vs_CG":{"n":512,"mean_diff":0.017629521150736016,"se":0.008281855961991932,"sum":9.02631482917684,"sumsq":18.104243007243216}},"afterSwap_peak":235478,"OOG":0,"gates":{"A":true,"B":true,"C":true},"all_gates_passed":true,"valid":true,"code_sha256":"02b4f502af8e57de7deb7920e3a73267844d8d3a83797abd257b15336a239e96"},
{"key":"k03","result_sha256":"243a754585687c9428994f3cd3ab646258bbd263ccd85d482bc639526cc5878a","blocked":null,"markets_per_s":412.29330367733775,"search_minutes":49.485,"generations":150,"settings":{"method":"pycma CMAEvolutionStrategy over 178 head weights, normalized coords x=(w-w_CG)/CMA_std, start x=0 (CG params)","CMA_stds":{"W1":0.16007274507182634,"b1":0.18564858000307327,"W2":0.12726224434596164,"b2":0.1185932470281719},"population":16,"sigma0":0.06,"sigma_cap":"sigma <= 0.06*0.5**(search_minutes/40), enforced after every tell","es_seed":603,"train_markets_per_generation":512,"common_markets_per_generation":true,"train_seeds":[1520000000,1520076799],"fitness":"mean replica edge over the generation's 512 markets (T baseline is common, so ranking equals gain over T)","monitor_seeds":[30000000,30002047],"monitor_every":10,"monitor_selection":"ES mean and generation best scored paired vs CG; keep max; CG if none beats it","weights_rounded":1e-06,"processes":4,"segments":"7 segments under 8 min each, pickled state in scratch"},"holdout":{"over_T":1.9519157504253348,"over_CG":0.0,"over_IB":0.18291575042533492,"se":{"over_T":0.011938024070038173,"over_CG":0.0}},"real":{"vs_IB":{"n":1024,"mean_diff":0.1682834843117217,"se":0.019904069246870944,"sum":172.322287935203,"sumsq":444.0097372628629},"vs_CG":{"n":512,"mean_diff":0.0,"se":0.0,"sum":0.0,"sumsq":0.0}},"afterSwap_peak":237819,"OOG":0,"gates":{"A":true,"B":true,"C":true},"all_gates_passed":true,"valid":true,"code_sha256":"e1ea8fcb105e00754d17061efbe4b299a634b0fcc1b5c293d65b631654b7204b"},
{"key":"k04","result_sha256":"a96e8a22e2c9294848bf9623ae706fd03efd0052c33c5560f950c0f1bd214ff8","blocked":null,"markets_per_s":313.4413702187676,"search_minutes":76.20533454518299,"generations":97,"settings":{"population":32,"dimensions":178,"markets_per_generation":256,"sigma0":0.1,"sigma_ceiling":"0.10 * 0.5**(training_minutes/40)","ES_seed":604,"training_seed_start":1530000000,"training_seed_end":1530024831,"monitor_seeds":[30000000,30002047],"monitor_every_generations":10,"CMA_stds":{"W1":0.16007274507182634,"b1":0.18564858000307327,"W2":0.12726224434596164,"b2":0.1185932470281719,"rule":"0.1*rms of HA W1 (13 old cols)/b1/W2/b2; new W1 columns start at 0 with W1 std"},"initial_weights":"CG.params, never CG.restart","coordinates":"W1 row-major (152), b1 (8), W2 row-major (16), b2 (2)","objective":"mean replica edge, same block for every proposal in generation","selection":"best monitor point among CG, every tenth generation ES mean and generation best","quantization":1e-06},"holdout":{"over_T":1.9519172921936434,"over_CG":0.0,"over_IB":0.1829172921936435,"se":{"over_T":0.011938004553452971,"over_CG":0.0}},"real":{"vs_IB":{"n":1024,"mean_diff":0.17084393645423684,"se":0.02011014175189102,"sum":174.94419092913853,"sumsq":453.53683080580737},"vs_CG":{"n":512,"mean_diff":0.0,"se":0.0,"sum":0.0,"sumsq":0.0}},"afterSwap_peak":237848,"OOG":0,"gates":{"A":true,"B":true,"C":true},"all_gates_passed":true,"valid":true,"code_sha256":"e1ea8fcb105e00754d17061efbe4b299a634b0fcc1b5c293d65b631654b7204b"},
{"key":"k05","result_sha256":"0847ff15871f9091367e268ff496402a2852f1236633b04a8d00524902269d5f","blocked":null,"markets_per_s":86.65745095289543,"search_minutes":0,"generations":0,"settings":{"mode":"required low-speed fallback","cutoff_markets_per_second":200,"reason":"Measured replica speed below 200 markets/s; no CMA-ES search permitted by brief.","planned_sigma0":0.13,"ES_seed":605,"popsize":32,"CMA_stds":{"W1":0.16007274507182634,"b1":0.18564858000307327,"W2":0.12726224434596164,"b2":0.1185932470281719,"rule":"0.1*rms of HA W1 (13 old cols)/b1/W2/b2; new W1 columns start at 0 with W1 std"},"train_markets_per_generation":128,"train_seed_start":1540000000,"monitor_seed_start":30000000,"monitor_seed_end":30002047,"training_performed":false,"initialization":"CG frozen params; restart never used"},"holdout":{"over_T":1.9519172921936434,"over_CG":0.0,"over_IB":0.1829172921936435,"se":{"over_T":0.011938004553452971,"over_CG":0.0}},"real":{"vs_IB":{"n":976,"mean_diff":0.18194513715612826,"se":0.02244372246080817,"sum":177.57845386438117,"sumsq":511.65013323186014},"vs_CG":{"n":976,"mean_diff":0.0,"se":0.0,"sum":0.0,"sumsq":0.0}},"afterSwap_peak":235599,"OOG":0,"gates":{"A":true,"B":true,"C":true},"all_gates_passed":true,"valid":true,"code_sha256":"e1ea8fcb105e00754d17061efbe4b299a634b0fcc1b5c293d65b631654b7204b"}
],
"best":{"key":"k01","result_sha256":"2023d343e67559907a8da0c8b178d0e52ade6dae29064ff3b373da650755b5fa","weights":[0.377567,0.356046,-0.667638,2.545459,-2.260978,-0.358841,-1.277071,2.078663,-4.097256,-1.81732,-0.903212,-0.696789,0.846616,-1.686496,0.656082,-0.146955,-0.842327,-0.254459,1.319435,0.398249,1.295391,1.39291,2.798965,0.286757,0.83181,-0.520457,-0.143574,-0.473977,0.543226,-0.174266,-1.914084,0.01205,-0.302114,0.038458,1.872365,-0.059224,0.089439,0.665201,1.395311,3.766688,-1.08101,0.963463,1.927527,2.304594,-2.193684,1.531812,1.493116,-1.440604,0.377747,-1.422682,-0.291479,0.391839,-0.331577,2.058578,-3.024463,-0.147274,-1.866862,0.091288,-1.546162,-0.051021,-0.435912,-1.141754,-0.53952,0.030253,0.426611,0.716964,0.192733,0.219679,-1.410481,-0.172716,0.167249,0.287884,-0.817676,1.661155,-0.316893,-0.834477,0.851131,0.690635,4.615599,-0.673141,-2.614258,-1.153956,-0.479063,-0.446467,-0.229665,0.03459,0.039046,-1.923889,0.001416,-0.842296,0.082825,-0.282464,2.018368,-0.072694,0.496722,1.373015,-1.637644,1.149117,-1.958229,3.578705,1.26264,2.24435,1.179905,-0.939547,-0.329401,-0.100196,3.589758,0.689255,-0.915996,-0.006865,0.081806,-1.486218,-0.363154,-1.176678,-2.127154,1.014625,-0.896049,-0.439361,2.414874,1.114218,-1.451955,2.150983,0.992382,-0.90161,1.464122,1.502271,1.582672,0.087486,-1.029623,-1.754763,0.342278,-1.158969,2.535865,-2.071498,-3.200743,3.486804,3.954261,-1.050095,-0.839952,-2.067388,2.111688,1.078617,-0.143585,2.485138,-2.055018,1.178728,0.150554,0.888245,1.734261,-0.69275,-1.714357,1.223643,1.338398,-1.552128,-4.778387,-0.944473,-1.612411,-0.882288,-1.135033,0.482454,0.367941,-2.109449,0.409917,-0.370609,-3.674363,-0.3956,-0.731814,1.107762,0.68397,-1.657232,-1.345954,3.721155,0.15111,-1.714835,0.651806,-0.066568,1.394891,2.34391],"code_zb64":"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","code_sha256":"0ab25ecd54f9a816119546408b08906ede14f74faa99d9c10cdfec7d6d6c1be6"}}
```