# r6-j5-joint merge 05 (items k41-k50)

Merged from the ten item results in `.imd/reads/artifacts/` (item-k41 to item-k50), each identified by the `result_sha256` in its row.

**Summary.** No item improved on J5. Two items ran a search: k44 froze J5 itself, and k47 froze a policy that scored 2.2990 over T on the holdout (J5 scores 2.3075; paired difference -0.0085 se 0.0053). `best` therefore holds J5 unchanged (row k44). Across the 9 items where J5 is the candidate, pooled J5 minus IB on the real engine is +0.5432 se 0.0094 over 8032 paired seeds; all candidates had 0 OOG and afterSwap peaks up to 238,937. k48 and k50 are partial, and k50 ran no holdout or gates A/C; see `notes`.

```json
{
"rows": [
{"key":"k41","mode":"real-only","result_sha256":"05448d5b9523f1f1cc6e1ed889f4d4e32375bbcd17d1f9d12313cfdfa86d62ea","markets_per_s":214.36068713348618,"generations":0,"search_minutes":0,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5504384437155975,"real_vs_IB_se":0.024654201850779125,"real_vs_IB_n":1024,"afterSwap_peak":238878,"OOG":0,"valid":true,"last_curve_point":null},
{"key":"k42","mode":"real-only","result_sha256":"04bf0d1c2b18db6a66603f73b9473ff0417cedc6658aca65d8909d0744917f67","markets_per_s":163.80495155887488,"generations":0,"search_minutes":0,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5563861433453503,"real_vs_IB_se":0.024876610762135293,"real_vs_IB_n":1024,"afterSwap_peak":238221,"OOG":0,"valid":true,"last_curve_point":[0,2.3568067935702803,null]},
{"key":"k43","mode":"real-only","result_sha256":"5b9e85317ef6ac804c39310ac0ed6a35b91914b3ecc5ce2e94122d72b290166a","markets_per_s":205.2,"generations":0,"search_minutes":0,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5084611843461581,"real_vs_IB_se":0.024224630738609018,"real_vs_IB_n":1024,"afterSwap_peak":237573,"OOG":0,"valid":true,"last_curve_point":null},
{"key":"k44","mode":"search","result_sha256":"afd7a3992ede0cfbcd7d8b879c7d62a57f4c8e60cbb0563fd985682035ffe898","markets_per_s":846.7,"generations":47,"search_minutes":30.07,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5390241772246698,"real_vs_IB_se":0.027370393483788356,"real_vs_IB_n":1024,"afterSwap_peak":238221,"OOG":0,"valid":true,"last_curve_point":[30.07,2.3376136145879647,0.4991799373246464]},
{"key":"k45","mode":"real-only","result_sha256":"ac6b94ed64dc3ea1c25bb9d363eee1590779be7f642feb11d7f8ae7c7a2815c1","markets_per_s":103.2383250572368,"generations":0,"search_minutes":0,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5633219278752516,"real_vs_IB_se":0.02489315957908245,"real_vs_IB_n":1024,"afterSwap_peak":238322,"OOG":0,"valid":true,"last_curve_point":null},
{"key":"k46","mode":"real-only","result_sha256":"8e9786d7733a854039f47e58f59b92d3966a028e473f805edd8995ecb21aa773","markets_per_s":295.9714559414134,"generations":0,"search_minutes":0,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5531974471335414,"real_vs_IB_se":0.028993992005455906,"real_vs_IB_n":1024,"afterSwap_peak":238232,"OOG":0,"valid":true,"last_curve_point":[0,2.3568067935702803,null]},
{"key":"k47","mode":"replica-search+real","result_sha256":"63ff195505c6b6db500b9ed3174c85f4e6678853ec0e289780cbed0f8887490b","markets_per_s":487.30615257709843,"generations":175,"search_minutes":45.25,"holdout_over_T":2.2990269541009445,"holdout_over_T_se":0.013732606685024297,"holdout_over_base":-0.008517114560603077,"holdout_over_base_se":0.005295434574417922,"real_vs_IB_mean":0.557461731467046,"real_vs_IB_se":0.034851184042424405,"real_vs_IB_n":1024,"afterSwap_peak":238060,"OOG":0,"valid":true,"last_curve_point":[45.19,2.3628791813169854,0.3520259455342695]},
{"key":"k48","mode":"real-only","result_sha256":"0658162d8a91018aba8d1f1ba0c8c3223ea1c7a5540cf2e47a0819c06f8a1960","markets_per_s":157.2028245920821,"generations":0,"search_minutes":0,"holdout_over_T":null,"holdout_over_T_se":null,"holdout_over_base":null,"holdout_over_base_se":null,"real_vs_IB_mean":0.5197795547996793,"real_vs_IB_se":0.04482978267763355,"real_vs_IB_n":352,"afterSwap_peak":235374,"OOG":0,"valid":true,"last_curve_point":null},
{"key":"k49","mode":"real-only","result_sha256":"a743efadabc0e0a19bb29e42112277f3d8df35c6bb972c940f6c72c1c6d2892f","markets_per_s":180.05978586157477,"generations":0,"search_minutes":0,"holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"holdout_over_base":0,"holdout_over_base_se":0,"real_vs_IB_mean":0.5516612299940767,"real_vs_IB_se":0.029119152301005435,"real_vs_IB_n":1024,"afterSwap_peak":237600,"OOG":0,"valid":true,"last_curve_point":null},
{"key":"k50","mode":"real-only","result_sha256":"a90ce7b7e044e07578beab32f04744deca4b819dd1bdddc627ade324ef4961b0","markets_per_s":161.6,"generations":0,"search_minutes":0,"holdout_over_T":null,"holdout_over_T_se":null,"holdout_over_base":null,"holdout_over_base_se":null,"real_vs_IB_mean":0.5190681811687325,"real_vs_IB_se":0.03931866126557076,"real_vs_IB_n":512,"afterSwap_peak":238937,"OOG":0,"valid":{"J5":true,"IB":true},"last_curve_point":null}
],
"best": {"key":"k44","mode":"search","holdout_over_T":2.307544068661547,"holdout_over_T_se":0.013528690579041706,"code_zb64":"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","code_sha256":"dd53db1fbe42628e040c6c7b7a82689867ab601263c5e740506299a3a186b38c","params":{"architecture":"19-8-2 as IA (ReLU clipped [0,4], inputs clipped [-4,4], output bps added to T-path fees)","W1":[[1.15129,-0.996585,-1.360612,2.864062,-2.166269,1.89742,-0.696507,2.175748,-3.58687,-2.449428,-0.696308,-0.558013,0.414019,-1.949829,-1.111019,0.52261,0.039733,-0.631826,0.495709],[0.86271,0.254613,1.646103,3.037976,1.718562,0.780098,-0.197718,-0.184908,-0.437817,0.190946,-0.202232,-1.333669,0.413344,-0.244053,-0.058662,1.013436,-0.128968,0.139462,0.195157],[1.729115,2.575792,-0.147932,1.629057,1.061114,4.271396,-1.566701,1.396977,1.6111,-1.150991,0.180765,-1.441042,-0.441326,0.394615,0.291057,1.177919,-1.248081,0.126277,-1.173105],[-0.664494,-0.965412,1.563603,-0.425824,-1.598845,1.349881,0.036501,0.697289,-0.445311,0.369449,0.32802,-0.940209,-2.098654,0.833946,-0.08927,0.548337,3.384914,0.014411,-0.487982],[1.022564,0.779154,4.569973,0.12783,-2.062091,-2.137189,-0.453636,-0.91211,-0.260035,0.136344,0.045042,-1.584326,0.291057,-1.076427,-0.006439,-0.275677,2.576729,-0.043329,-0.250119],[1.274221,-2.125229,0.477101,-0.440908,2.15979,0.969918,2.157772,1.49217,-0.795849,0.290526,0.042831,3.728527,0.655144,-0.751888,0.053949,1.711916,-0.967529,-0.845371,0.51631],[-2.154359,0.323503,0.502496,-1.692203,2.351653,1.316477,-0.470729,1.546865,1.607024,-0.69916,0.730603,1.579664,0.852441,0.6062,-0.560468,-0.257002,1.122865,-0.208801,1.201277],[-3.431737,-1.947345,4.449738,3.631164,-1.291438,-1.559901,-1.265114,0.949285,2.503463,-0.876593,2.612494,-1.483027,0.781478,-0.841417,0.463619,0.746681,0.865164,-1.373002,1.526777]],"b1":[0.497572,-0.831682,-2.767245,-0.125743,-1.222987,-1.550949,-0.777339,-0.161961],"W2":[[0.037931,-2.360691,0.520992,-0.817009,-3.147262,-0.447985,0.471992,0.567987],[0.587518,-2.30336,-1.969661,2.252433,0.778349,-1.707478,0.144538,0.56426]],"b2":[0.8623,1.118605],"constants":{"ELAPSED_CAP":7,"SIGNAL_THRESHOLD":0.001168085308968538,"DIR_DECAY":0.6708732125720838,"ACT_DECAY":0.49682908632969613,"SIZE_DECAY":0.4728256252769685,"TOX_DECAY":0.935365149604538,"SIGMA_DECAY":0.6696245950727084,"LAMBDA_DECAY":0.987870739046352,"SIZE_BLEND_DECAY":0.8483380461513357,"TOX_BLEND_DECAY":0.24478681527482282,"ACT_BLEND_DECAY":0.9846490771136527,"PHAT_ALPHA":0.247625310137791,"PHAT_ALPHA_RETAIL":0.0507874078770196,"PHAT_ALPHA_SAME":0.14851350675375644,"DIR_IMPACT_MULT":5,"GATE_SIGMA_MULT":5.424109880048015,"MIN_GATE":0.02752861145289368,"TRADE_TOX_BOOST":0.29535921748716293,"STALE_ATTRACT_FRAC":0.7339436627363265,"RET_CAP":0.02077647001542073,"TOX_CAP":0.18285021586819034,"TRADE_RATIO_CAP":0.2114237530741308,"LAMBDA_CAP":1.3277634863952608,"STEP_COUNT_CAP":56,"BASE_FEE":0.00051082991832555,"SIGMA_COEF":0.19158029221698078,"LAMBDA_COEF":0.001362871137046423,"FLOW_SIZE_COEF":0.47371206396001814,"TOX_COEF":0.02450003373115383,"TOX_QUAD_COEF":1.322845864299775,"ACT_COEF":8.240528322972024,"DIR_COEF":0.003178306275011696,"DIR_TOX_COEF":0.0104643196179574,"STALE_DIR_COEF":0.7451429374420652,"SIGMA_TOX_COEF":0.045521188875803024,"TAIL_KNEE":0.04944397677675635,"TAIL_SLOPE_PROTECT":0.8769115691956163,"TAIL_SLOPE_ATTRACT":0.8632618839360807,"ANCHOR_COUNT_CAP":564,"ANCHOR_LINEAR_K":0.7532823694587746,"ANCHOR_QUADRATIC_K":1.301162029835076},"scaling":{"feature_multipliers":[1000,2,500,500,500,1,500,500,2,2,1000000,1,500,647.054817,1.395287,220.617911,228.148131,0.412565,274.335453],"feature_offsets":[-1,-1,-1,-1,-1,-1,0,0,-1,-1,-1,0,-1,-0.618293,-2.037262,-0.803057,-0.808615,-0.823811,0.001772],"weight_fixed_point_scale":1000000}}},
"real_only": [
{"key":"k41","seeds":[8460000,8461023],"n":1024,"mean":0.5504384437155975,"se":0.024654201850779125,"sum":563.6489663647718,"sumsq":946.9872451621744},
{"key":"k42","seeds":[8464000,8465023],"n":1024,"mean":0.5563861433453503,"se":0.024876610762135293,"sum":569.7394107856387,"sumsq":965.268230212348},
{"key":"k43","seeds":[8468000,8469023],"n":1024,"mean":0.5084611843461581,"se":0.024224630738609018,"sum":520.6642527704659,"sumsq":879.4753672195725},
{"key":"k45","seeds":[8476000,8477023],"n":1024,"mean":0.5633219278752516,"se":0.02489315957908245,"sum":576.8416541442576,"sumsq":974.083465536345},
{"key":"k46","seeds":[8480000,8481023],"n":1024,"mean":0.5531974471335414,"se":0.028993992005455906,"sum":566.4741858647463,"sumsq":1193.9983094712263},
{"key":"k48","seeds":[8488000,8488351],"n":352,"mean":0.5197795547996793,"se":0.04482978267763355,"sum":182.9624032894871,"sumsq":343.4037341595609},
{"key":"k49","seeds":[8492000,8493023],"n":1024,"mean":0.5516612299940767,"se":0.029119152301005435,"sum":564.9010995139346,"sumsq":1199.879597173246},
{"key":"k50","seeds":[8496000,8496511],"n":512,"mean":0.5190681811687325,"se":0.03931866126557076,"sum":265.76290875839106,"sumsq":542.420923863576}
],
"notes": [
"BEST: selected as the brief says (valid search rows, highest holdout over_T). Search rows are k44 (over_T 2.307544, se 0.013529) and k47 (over_T 2.299027, se 0.013733). k44 wins, but k44's frozen point is J5 itself (search.frozen_point: no later monitored CMA mean exceeded J5; code_sha256 dd53db1f, same as every real-only row), so `best` is unchanged J5, not a search improvement. No search in this group produced a candidate better than J5 on the holdout. [k44.search.frozen_point; k44.holdout]",
"k47 is the only new policy (code_sha256 0abbcf07..., 178 of 178 head values and 39 of 41 constants differ from J5): holdout over_T 2.299027 vs J5 2.307544, paired over_base -0.008517 se 0.005295 against J5, i.e. indistinguishable and slightly lower (k47.holdout). Its code and params are not copied here; they are in artifacts/goal/batches/r6-j5-joint/items/k47/result.json (result_sha256 in its row).",
"WHETHER CURVES WERE STILL RISING (monitor gain over T, 2048 seeds, se about 0.029): k44 never returned to its J5 start: 2.3568 at 0 min, 2.3238 at 20.4 min, 2.3376 at 30.07 min. k47 sat 0.01-0.03 below J5 until its last row: 2.3568 at 0 min, 2.3481 at 40.19, 2.3629 at 45.19, so +0.0061 over J5 at the final point, inside the se, and its holdout says no gain. So neither curve shows the sustained rise reported for the pilot; k47's final segment rose but the holdout does not support a real gain. These were 47 and 175 generations, against 1496 for J5's own search (J5 artifact search.generations). Real-only rows have no training curve; k42 and k46 show one diagnostic J5 point [0, 2.3568, null] (their curve fields).",
"REAL ENGINE, J5 vs IB, pooled over the 9 items with frozen J5 as the candidate (k41-k46, k48-k50, seed sets disjoint): n 8032, sum 4362.955638969755, sumsq 8127.79853648845, mean_diff 0.543197, se 0.009448 (computed by this merge from the items' sum and sumsq). Per-item means run 0.5085 (k43) to 0.5633 (k45), each se 0.024-0.045. k47 (a different candidate) is +0.5575 se 0.0349, n 1024 (k47.real.vs_IB), within noise of J5.",
"REAL vs HA CONTEXT: this group's items did not measure over-HA; J5's artifact reports real vs_HA +1.219 se 0.038 on 512 seeds (api.imd.fun/artifacts/fe2320b5..., real.vs_HA), against the about +3.1 over HA the goal needs for the top 10. No item in this group claims or measures a top-10 placement, and none improved J5.",
"GAS: all items reported 0 OOG on the candidate over their real seeds (where reported, also IB), and afterSwap peaks 235,374-238,937, under the 245,000 limit (rows). afterInitialize peak 243,234 on the J5 rows that report it (e.g. k43/k44/k47/k50 gas), recorded for reference; the 245,000 limit is defined on afterSwap only.",
"RETAIL/ARB DELTAS vs IB on the real engine, Y volumes, J5 minus IB unless noted (se from the items where given): k41 retail +68.20 se 7.44, arb -166.96 se 4.88; k42 +66.09 se 7.46, -173.62 se 4.89; k43 +66.26 se 7.55, -168.45 se 5.02; k44 not computed (run_batch exposes only totals and the reconstruction artifact was unavailable to k44); k45 +63.13 se 7.46, -174.83 se 4.88; k46 +72.54 se 7.77, -175.86 se 4.89; k48 +48.38, -182.33 (n 352); k49 +72.31 se 7.76, -173.59 se 5.05; k50 +47.96 se 10.68, -183.01 se 7.02 (n 512). k47 candidate is +228.04 se 8.93, +22.24 se 5.50, a different pattern (its candidate differs from J5). Edge-split deltas were computed by k42, k46 and k49 on 960 of 1024 seeds: retail edge -0.512/-0.507/-0.528, arb edge +1.063/+1.062/+1.079 (k42.real.retail_delta/arb_delta; k46/k49 real.vs_IB). Edge split was not computed by k41, k43, k44, k45, k48, k50; k47 reports a replica split only.",
"WORST SEEDS: k47's worst single seed against IB is -17.18 at 8484262 (k47.real.vs_IB.worst5), far beyond the worst J5 single-seed value, -5.97 at 8492876 (k49). Holdout worst5 in k41/k46 reports -9.87 against T for J5. Only k47 has this outlier; it is one seed and has not been investigated further.",
"MISSING OR PARTIAL WORK: k48 completed only 352 of 1024 planned real seeds (8488000..8488351; heavily loaded host, commands killed at the 8-minute limit) and has no holdout (k48.holdout: not run). k50 completed only 512 paired IB seeds out of 1024 (8496000..8496511; J5 alone ran all 1024), and has no holdout and no gate A or C run (k50.gates A/C run:false; holdout null). k50's `valid` is an object {J5:true,IB:true}, kept verbatim in its row, and k50 did not run gates A or C, so its valid does not cover them. k44 stopped at 30.07 search minutes, below the requested 45 (k44.notes). k44 reported no retail/arb deltas.",
"PROTOCOL DEVIATION: k47's first speed measurement was 487.3 markets/s, under the 500 threshold, but it ran the search on the strength of five repeats at 516-526/s (k47.speed, k47.notes); the row's markets_per_s is the first measurement. k44 and k47 are the only items that reached a search. Eight items took the real-only branch on speed: k41 214, k42 164, k43 205, k45 103 (only 2 processes), k46 296, k48 157, k49 180, k50 162. The low speeds track host load (k43: load about 30, the later holdout ran near 550/s; k48: load 25-65). Rows set generations and search_minutes to 0 for no-search items.",
"VALIDITY SCOPE: every item except k50 states `valid` true on its own terms; gates A, B and C pass where run (max_fee_diff_bps up to 0.00089, k44; k50 did not run A or C). The A/C callback generators differ between items (k41: lognormal(-4,1) amounts; k45: fraction-of-reserve lognormal(-7,1.2); k46: lognormal(-6,1.2); k48: simulator sequences; k49: lognormal(-5,1)). k44's valid is `true` although it fell short of 45 search minutes; k47's is `true` despite the speed-rule deviation above.",
"KNOWN FRAGILITY, not a real-engine failure: J5 reverts with Panic 0x12 (division by zero) on unbounded synthetic callback paths where cumulative trades push spot to about 7e7 (k41.gates.additional_stress at sequence 310037, callback 1424; k43.notes; k45.notes; k46.notes; k49.gates.additional_stress; k42.notes). All these items then ran bounded generators for gates A/C. The real engine recorded 0 reverts in every item; the items do not show that real markets cannot reach such a state, so this evidence covers the real-engine seeds run here, not those extreme states.",
"HOLDOUT DUPLICATION: every holdout-running item reports J5 over T exactly 2.307544068661547 on seeds 40000000..40008191 because they evaluate the same deterministic replica on the same seeds; these rows are not independent evidence. k42 and k46 also report retail +4.5243 / arb -2.2168 over T on the replica holdout (k42.holdout; k46.holdout), consistent with each other.",
"CHECKS BY THIS MERGE: all ten result files match the SHA-256 given in the brief; each code_zb64 decodes and matches its code_sha256; sum and sumsq reproduce each item's mean_diff and se; k41, k42, k46 per-seed lists reproduce sum and sumsq; J5 rows carry byte-identical code_zb64 and identical weights and constants. Upstream artifacts were re-fetched from api.imd.fun/artifacts/<id> and their sha256 equals the id for J5 fe2320b5, IA df0ddfdf, IB afd5183e, replica e8a7a2af, HA 2801efc8; J5's artifact matches k44 on code_zb64, the four weight blocks and the constants (artifact constants are [start, value] pairs; items hold the value). Upstream J5 artifact reports holdout over_T 2.307544 and real vs IA +0.8762, the values items used. The artifact's scaling fields say 'unchanged from IA params', and the items' numeric scaling arrays were not compared against IA here.",
"FLAGGED SOURCES: the orchestrator reports cited sources that did not re-verify for every item (k41: 2, k42: 3, k43: 2, k44: 2, k45: 2, k46: 2, k47: 3, k48: 1, k49: 2, k50: 1). The result files contain no URLs or sources list, and the items' source files were not given to this merge, so which claims those sources backed cannot be determined here. The columns above are local run measurements and do not rest on cited pages; however, they are unaudited self-reports (several items say so explicitly: k42, k45, k46, k47 notes), recorded here as such. The only outside references visible in the results are artifact hashes and reference values taken from those artifacts; the re-fetches in the previous note confirm the ones concerning J5, IA, IB, the replica and HA. Any other statement an item makes about an outside page should be given less weight.",
"TRANSCRIPTION SLIPS IN NOTES (values elsewhere are consistent): k50's note writes 'IB artifact afd51838', while the id is afd5183e... (k41/k42/k45/k46 hashes fields); k43's note writes IB code '301941829' where others give 30194182. Neither affects the numbers."
]
}
```