# Merge 02 — batch r4-ib-wide (items w06–w10)

Long replica training of IB's 19-8-2 fee head, starting from its frozen weights, followed by one real-engine check per item.
All five inputs were read from `.imd/reads/artifacts/`. Each file's sha256 matches the hash it was given under (listed per row below).

## Summary

| key | searched | valid | mkts/s | min | gens | pop | holdout over_IB (se) | holdout over_HA | real vs_IB (se) | afterSwap peak | OOG | A/B/C |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| w06 | True | True | 275.96592585681975 | 60.70955021521077 | 77 | 16 | 0.03303069450470034 (0.004101952407878324) | 0.6851827411781026 | 0.0525245023517521 (0.015324409084365631) | 237993 | 0 | True/True/True |
| w07 | False | True | 116.57808131918684 | 0 | 0 | — | — (—) | — | 0.0 (0.0) | 237016 | 0 | True/True/True |
| w08 | True | True | 507.4334092919039 | 75.06038842398362 | 148 | 32 | 0.041942587481934385 (0.005029992164378871) | 0.6940946341553367 | 0.06288579925944027 (0.01383265485095135) | 237039 | 0 | True/True/True |
| w09 | False | True | 95.247954873156 | 0 | 0 | — | — (—) | — | 0.0 (0.0) | 234946 | 0 | True/True/True |
| w10 | False | False | — | — | — | — | — (—) | — | — (—) | — | — | None/None/None |

- **Best: w08**. It is the valid searched item with the highest holdout gain over IB. code_sha256 `5dfd0a55d44c3d48ffd101ed650b0e0ded5ae201349c2c723ae817162ce4773a`.
- Only w06 and w08 trained. Both improve IB by only a few hundredths, which is far below the roughly +3.1 over HA that the goal needs. See `notes`.
- w07 and w09 hit the speed fallback and ran no search. w10 was blocked at setup. Both are explained in `notes`.
- Sources: w06–w09 each had 3 cited sources that failed re-check. This mainly affects their claims about input hashes and the Gate B reference, not their measurements. See `notes`.
- Source of every figure: the item result file named by `result_sha256` in its row. No figure was computed by this merge. Numbers in `rows` are strings and `null` means the item did not measure the value. `weights` and `best.params` are copied exactly from the results.

## Merged object

```json
{
 "rows": [
  {
   "key": "w06",
   "result_sha256": "6e83add27cd040ca1c4df5d2656f2de1397719931c4a6101d082ebeeb4eaca0a",
   "markets_per_s": "275.96592585681975",
   "search_minutes": "60.70955021521077",
   "generations": "77",
   "population": "16",
   "markets_per_generation": "512",
   "holdout_over_IB": "0.03303069450470034",
   "holdout_over_IB_se": "0.004101952407878324",
   "holdout_over_HA": "0.6851827411781026",
   "holdout_over_HA_se": "0.00797221391056306",
   "real_vs_IB_n": "512",
   "real_vs_IB_mean_diff": "0.0525245023517521",
   "real_vs_IB_se": "0.015324409084365631",
   "afterSwap_peak": "237993",
   "OOG": "0",
   "gates": {
    "A": true,
    "B": true,
    "C": true
   },
   "valid": true,
   "searched": true
  },
  {
   "key": "w07",
   "result_sha256": "3591cc3230a24fd285726f1e6eefe213ee0212ed9221a6d52d4be4711905fa56",
   "markets_per_s": "116.57808131918684",
   "search_minutes": "0",
   "generations": "0",
   "population": null,
   "markets_per_generation": null,
   "holdout_over_IB": null,
   "holdout_over_IB_se": null,
   "holdout_over_HA": null,
   "holdout_over_HA_se": null,
   "real_vs_IB_n": "1056",
   "real_vs_IB_mean_diff": "0.0",
   "real_vs_IB_se": "0.0",
   "afterSwap_peak": "237016",
   "OOG": "0",
   "gates": {
    "A": true,
    "B": true,
    "C": true
   },
   "valid": true,
   "searched": false
  },
  {
   "key": "w08",
   "result_sha256": "7aa96ad733ad72e17b458c49d590811c271555bd32d61943277c654825930ce5",
   "markets_per_s": "507.4334092919039",
   "search_minutes": "75.06038842398362",
   "generations": "148",
   "population": "32",
   "markets_per_generation": "512",
   "holdout_over_IB": "0.041942587481934385",
   "holdout_over_IB_se": "0.005029992164378871",
   "holdout_over_HA": "0.6940946341553367",
   "holdout_over_HA_se": "0.008589827695524686",
   "real_vs_IB_n": "512",
   "real_vs_IB_mean_diff": "0.06288579925944027",
   "real_vs_IB_se": "0.01383265485095135",
   "afterSwap_peak": "237039",
   "OOG": "0",
   "gates": {
    "A": true,
    "B": true,
    "C": true
   },
   "valid": true,
   "searched": true
  },
  {
   "key": "w09",
   "result_sha256": "da43ed1fae14054d73767cad20f595f80c81a131b0a2f5ae4f6fb8a5a085dc13",
   "markets_per_s": "95.247954873156",
   "search_minutes": "0",
   "generations": "0",
   "population": null,
   "markets_per_generation": null,
   "holdout_over_IB": null,
   "holdout_over_IB_se": null,
   "holdout_over_HA": null,
   "holdout_over_HA_se": null,
   "real_vs_IB_n": "1536",
   "real_vs_IB_mean_diff": "0.0",
   "real_vs_IB_se": "0.0",
   "afterSwap_peak": "234946",
   "OOG": "0",
   "gates": {
    "A": true,
    "B": true,
    "C": true
   },
   "valid": true,
   "searched": false
  },
  {
   "key": "w10",
   "result_sha256": "4712e16bbbaaa6a82ed87fe7a3edc2619ce5a3f92f472648166f016b0a4abf35",
   "markets_per_s": null,
   "search_minutes": null,
   "generations": null,
   "population": null,
   "markets_per_generation": null,
   "holdout_over_IB": null,
   "holdout_over_IB_se": null,
   "holdout_over_HA": null,
   "holdout_over_HA_se": null,
   "real_vs_IB_n": null,
   "real_vs_IB_mean_diff": null,
   "real_vs_IB_se": null,
   "afterSwap_peak": null,
   "OOG": null,
   "gates": {
    "A": null,
    "B": null,
    "C": null
   },
   "valid": false,
   "searched": false
  }
 ],
 "weights": {
  "w06": [
   0.549334,
   1.687597,
   -0.446712,
   3.633784,
   -2.840337,
   -0.763271,
   -1.073162,
   1.561127,
   -5.041441,
   0.354249,
   2.106027,
   -0.477865,
   0.381477,
   1.409349,
   0.292361,
   -1.520592,
   0.343909,
   0.321628,
   0.485126,
   0.529009,
   1.159744,
   2.605723,
   2.91636,
   0.719161,
   -1.011648,
   -0.05567,
   -0.194834,
   -0.125496,
   -0.069357,
   -0.218565,
   -0.8542,
   0.297246,
   -1.27636,
   -1.086565,
   1.312439,
   0.407593,
   -0.824731,
   0.252784,
   1.372173,
   2.895227,
   0.292387,
   0.366965,
   1.157772,
   1.36978,
   -1.672947,
   1.236553,
   0.564553,
   -1.043537,
   0.184543,
   -3.047025,
   -0.583106,
   -0.448905,
   0.639696,
   0.780119,
   -0.311008,
   -1.010258,
   -1.197642,
   -1.50179,
   -1.441085,
   1.338951,
   -0.364959,
   -0.798654,
   0.666673,
   -1.095909,
   1.624617,
   -1.686415,
   -0.162621,
   0.509203,
   -2.287506,
   -1.633197,
   -0.087189,
   0.330625,
   0.840818,
   1.753095,
   0.422579,
   1.681342,
   0.214116,
   0.723548,
   2.527124,
   -0.18472,
   -2.059287,
   0.180596,
   1.357115,
   1.066312,
   -0.161755,
   0.18521,
   0.010158,
   -2.037194,
   -0.180595,
   -0.531305,
   0.047506,
   -1.512642,
   2.198802,
   -0.484875,
   -0.138352,
   2.505087,
   -1.153859,
   1.816105,
   -1.008775,
   2.021776,
   1.488557,
   0.084204,
   -0.01828,
   -0.549215,
   0.093151,
   -0.225738,
   3.383157,
   1.267958,
   -0.573983,
   -1.357704,
   0.577669,
   -0.064152,
   1.125371,
   0.155884,
   -1.035609,
   1.369079,
   0.260992,
   -2.862956,
   1.541565,
   0.174667,
   -0.146475,
   -0.12893,
   0.635111,
   0.081126,
   0.853639,
   0.184534,
   -0.794364,
   0.815151,
   -0.128265,
   0.095566,
   -0.173536,
   0.545552,
   -0.397818,
   -2.109688,
   -1.624275,
   2.944831,
   3.073511,
   -0.229344,
   -2.497889,
   -1.031583,
   2.868815,
   1.026783,
   -1.728121,
   2.596928,
   -2.173214,
   0.334294,
   0.42301,
   -0.109699,
   2.410922,
   1.184518,
   0.273911,
   -0.655177,
   2.297993,
   -1.734579,
   -3.345959,
   -1.782727,
   -0.699529,
   -1.655852,
   -2.673111,
   -0.851723,
   0.512142,
   -1.655367,
   0.054065,
   -0.464514,
   -2.949447,
   -0.529975,
   -2.304723,
   0.851415,
   0.403918,
   -1.47663,
   -1.245486,
   0.586482,
   1.976294,
   -1.166057,
   0.71628,
   0.302605,
   1.885727,
   2.05856
  ],
  "w08": [
   0.76818,
   1.492611,
   -0.82184,
   3.22623,
   -2.742334,
   -0.496728,
   -1.107862,
   1.237659,
   -5.023633,
   -0.1657,
   1.833447,
   -0.547024,
   0.322372,
   1.37529,
   0.364681,
   -1.043102,
   0.530069,
   0.240353,
   0.219488,
   0.592175,
   0.925336,
   2.44196,
   2.874909,
   0.609473,
   -0.950093,
   -0.220445,
   -0.235336,
   -0.075041,
   -0.08705,
   -0.150224,
   -0.971107,
   -0.017733,
   -1.285051,
   -0.906677,
   1.438549,
   0.49411,
   -0.685943,
   0.106729,
   1.316907,
   3.00805,
   0.072407,
   0.470065,
   1.011851,
   1.797496,
   -1.567939,
   1.322709,
   0.910816,
   -0.936133,
   0.195664,
   -3.211737,
   -0.679618,
   -0.258938,
   0.235077,
   0.70412,
   -0.388873,
   -0.799709,
   -1.062115,
   -1.25772,
   -0.772287,
   1.386393,
   -0.40182,
   -0.369839,
   1.271398,
   -1.016775,
   1.713782,
   -1.555966,
   -0.161421,
   0.195151,
   -2.307722,
   -1.601025,
   -0.00937,
   -0.024311,
   0.903026,
   1.986951,
   0.27561,
   1.783493,
   0.284298,
   0.82343,
   2.501842,
   -0.329539,
   -2.029555,
   0.086215,
   1.36587,
   1.167533,
   -0.143093,
   0.227483,
   0.011422,
   -1.958674,
   -0.057782,
   -0.543676,
   0.221279,
   -1.319585,
   2.379594,
   -0.396339,
   -0.133795,
   2.535464,
   -1.158153,
   1.796431,
   -0.953909,
   1.716174,
   1.317991,
   0.136639,
   0.018529,
   -0.768033,
   0.198061,
   -0.192303,
   3.320716,
   0.90425,
   -0.456196,
   -1.145337,
   0.606393,
   0.082938,
   1.287636,
   0.193141,
   -1.049991,
   1.212686,
   0.251807,
   -2.892951,
   1.504471,
   -0.203362,
   -0.369515,
   -0.116136,
   0.653416,
   -0.105238,
   0.825544,
   0.356032,
   -0.946531,
   0.638486,
   -0.017644,
   0.130364,
   0.270786,
   0.993226,
   -0.3799,
   -2.118129,
   -1.672671,
   2.730096,
   3.093645,
   -0.458675,
   -2.613598,
   -1.402909,
   2.903498,
   1.333096,
   -1.585719,
   2.620219,
   -2.232235,
   0.261143,
   0.722685,
   -0.502692,
   2.135685,
   0.708834,
   0.162446,
   -0.846015,
   1.652872,
   -1.518703,
   -3.244752,
   -1.477866,
   -0.947461,
   -1.65868,
   -2.631638,
   -0.725693,
   0.475163,
   -1.739058,
   0.274451,
   -0.240457,
   -2.887586,
   -0.248875,
   -2.370849,
   0.722041,
   0.366591,
   -1.38459,
   -1.312942,
   0.440014,
   1.942611,
   -1.273108,
   0.749713,
   0.492264,
   1.768973,
   1.989157
  ]
 },
 "best": {
  "key": "w08",
  "params": {
   "W1": [
    [
     0.76818,
     1.492611,
     -0.82184,
     3.22623,
     -2.742334,
     -0.496728,
     -1.107862,
     1.237659,
     -5.023633,
     -0.1657,
     1.833447,
     -0.547024,
     0.322372,
     1.37529,
     0.364681,
     -1.043102,
     0.530069,
     0.240353,
     0.219488
    ],
    [
     0.592175,
     0.925336,
     2.44196,
     2.874909,
     0.609473,
     -0.950093,
     -0.220445,
     -0.235336,
     -0.075041,
     -0.08705,
     -0.150224,
     -0.971107,
     -0.017733,
     -1.285051,
     -0.906677,
     1.438549,
     0.49411,
     -0.685943,
     0.106729
    ],
    [
     1.316907,
     3.00805,
     0.072407,
     0.470065,
     1.011851,
     1.797496,
     -1.567939,
     1.322709,
     0.910816,
     -0.936133,
     0.195664,
     -3.211737,
     -0.679618,
     -0.258938,
     0.235077,
     0.70412,
     -0.388873,
     -0.799709,
     -1.062115
    ],
    [
     -1.25772,
     -0.772287,
     1.386393,
     -0.40182,
     -0.369839,
     1.271398,
     -1.016775,
     1.713782,
     -1.555966,
     -0.161421,
     0.195151,
     -2.307722,
     -1.601025,
     -0.00937,
     -0.024311,
     0.903026,
     1.986951,
     0.27561,
     1.783493
    ],
    [
     0.284298,
     0.82343,
     2.501842,
     -0.329539,
     -2.029555,
     0.086215,
     1.36587,
     1.167533,
     -0.143093,
     0.227483,
     0.011422,
     -1.958674,
     -0.057782,
     -0.543676,
     0.221279,
     -1.319585,
     2.379594,
     -0.396339,
     -0.133795
    ],
    [
     2.535464,
     -1.158153,
     1.796431,
     -0.953909,
     1.716174,
     1.317991,
     0.136639,
     0.018529,
     -0.768033,
     0.198061,
     -0.192303,
     3.320716,
     0.90425,
     -0.456196,
     -1.145337,
     0.606393,
     0.082938,
     1.287636,
     0.193141
    ],
    [
     -1.049991,
     1.212686,
     0.251807,
     -2.892951,
     1.504471,
     -0.203362,
     -0.369515,
     -0.116136,
     0.653416,
     -0.105238,
     0.825544,
     0.356032,
     -0.946531,
     0.638486,
     -0.017644,
     0.130364,
     0.270786,
     0.993226,
     -0.3799
    ],
    [
     -2.118129,
     -1.672671,
     2.730096,
     3.093645,
     -0.458675,
     -2.613598,
     -1.402909,
     2.903498,
     1.333096,
     -1.585719,
     2.620219,
     -2.232235,
     0.261143,
     0.722685,
     -0.502692,
     2.135685,
     0.708834,
     0.162446,
     -0.846015
    ]
   ],
   "b1": [
    1.652872,
    -1.518703,
    -3.244752,
    -1.477866,
    -0.947461,
    -1.65868,
    -2.631638,
    -0.725693
   ],
   "W2": [
    [
     0.475163,
     -1.739058,
     0.274451,
     -0.240457,
     -2.887586,
     -0.248875,
     -2.370849,
     0.722041
    ],
    [
     0.366591,
     -1.38459,
     -1.312942,
     0.440014,
     1.942611,
     -1.273108,
     0.749713,
     0.492264
    ]
   ],
   "b2": [
    1.768973,
    1.989157
   ],
   "hidden_units": 8,
   "input_count": 19,
   "layout": "flat vector = W1 (8x19 row-major), b1(8), W2 (2x8 row-major: bid row then ask row), b2(2)",
   "weight_fixed_point_scale": 1000000,
   "feature_fixed_point_scale": 1000000,
   "output_unit_fee": 0.0001,
   "feature_order": [
    "sigmaHat",
    "lambdaHat",
    "sizeHat",
    "toxEma",
    "actEma",
    "direction",
    "(spot-staleHat)/staleHat",
    "signed_trade_ratio",
    "arb_probability",
    "first_in_step",
    "uncertainty",
    "min(silent,8)/8",
    "anchor_mean",
    "raw_amountY",
    "step_trade_count",
    "prev_bid_fee",
    "prev_ask_fee",
    "steps_since_arb_like_print",
    "router_implied_vanilla_spot"
   ],
   "feature_multipliers": [
    1000,
    2,
    500,
    500,
    500,
    1,
    500,
    500,
    2,
    2,
    1000000,
    1,
    500,
    0.05,
    0.5,
    333.333333,
    333.333333,
    0.125,
    500
   ],
   "feature_offsets": [
    -1,
    -1,
    -1,
    -1,
    -1,
    -1,
    0,
    0,
    -1,
    -1,
    -1,
    0,
    -1,
    -1,
    -0.5,
    -1,
    -1,
    -1,
    0
   ],
   "feature_clip": [
    -4,
    4
   ],
   "hidden_relu_clip": [
    0,
    4
   ],
   "toxicity_anchor": {
    "count_cap": 674,
    "linear_k_wad": 712696014807650816,
    "quadratic_k_wad": 1295240753826310656
   }
  },
  "code_zb64": "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",
  "code_sha256": "5dfd0a55d44c3d48ffd101ed650b0e0ded5ae201349c2c723ae817162ce4773a"
 },
 "notes": [
  "w10: ran no search and is not valid. Setup stopped at `uv venv` with exit 126 ('scratch/bin/uv: cannot execute binary file'); the item reports that its host was Darwin arm64 and the setup downloads an x86_64 Linux uv. It measured nothing: no speed, gates, holdout, real-engine run or gas. It reports that the simulator wheel hash check passed before the failure. It produced no candidate (code_zb64, code_sha256 and params are null).",
  "w07: ran no search (searched=false). It measured replica speed at 116.578 markets/s (an earlier run that overlapped the parity process measured 109.871) and took the brief's speed fallback. It delivers the unchanged IB (code_sha256 301941829a44\u2026d686, same params as IB) with no holdout. Its real vs_IB of 0 with se 0 compares the IB arm against itself, as the item says, so it does not show any gain. What it did measure on real markets: IB vs HA on 1056 seeds (7584032\u20137585087), mean_diff 0.6642513815075862, se 0.021755339484222994, OOG 0, IB afterSwap peak 237016. Seeds 7584000\u20137584031 were dropped after a serialization bug and never rerun. Its Gate C uses the Gate A replay because the frozen output is byte-identical to IB.",
  "w09: ran no search (searched=false). It measured replica speed at 95.248 markets/s and took the speed fallback. It delivers the unchanged IB (same code_sha256 and params as w07) with no holdout. Its vs_IB = 0 is IB compared with itself. What it measured on real markets: IB vs HA on 1536 seeds (7592000\u20137593535), mean_diff 0.6498760072868056, se 0.0192648148126186, OOG 0, IB afterSwap peak 234946, HA afterSwap peak 231555. Its Gate C repeats Gate A (same code and weights). Its real retail/arb deltas are Y-volume differences, not edge.",
  "w06 and w08 are the only searched items, and both are valid: Gates A/B/C passed, OOG 0 and afterSwap peak at most 245,000. Both measured holdout on the same 8192 replica seeds (40000000\u201340008191) and monitor on the same 2048 seeds (30000000\u201330002047). Both report the same initial-IB monitor gain over T, 1.8095714223753105. Their holdout gains over IB can therefore be compared directly: w08 0.041942587481934385 (se 0.005029992164378871) vs w06 0.03303069450470034 (se 0.004101952407878324). The gap is within two combined standard errors, so 'best' is the higher point estimate, not a clear winner. Their real-engine checks use different fresh seed blocks: w06 7580000\u20137580511 and w08 7588000\u20137588511.",
  "Size of the gains: both searched candidates beat IB by only about +0.03 to +0.06 on the replica holdout and on real markets (w06 real vs_IB 0.0525245023517521 se 0.015324409084365631; w08 real vs_IB 0.06288579925944027 se 0.01383265485095135). Against HA, the holdout gains are w06 0.6851827411781026 and w08 0.6940946341553367, and the real gains are w06 0.7067310415787678 (se 0.038053443609535005) and w08 0.659927865646242 (se 0.03460751527018567). The goal needs about +3.1 over HA, and no item in this group gets close. Both monitor curves were nearly flat: w06 went from 1.8096 to 1.8228 in 60.7 min, and w08 from 1.8096 to 1.8551 in 75.1 min, with sigma capped to about 0.009\u20130.012 by the end.",
  "Replica vs real agreement (no new numbers): w06 holdout over_IB 0.0330 vs real 0.0525, and w08 0.0419 vs 0.0629. Both pairs agree well within the goal's ~0.1 replica-accuracy figure.",
  "Retail/arb deltas are in different units across items, so they are not comparable. w06 reports real EDGE deltas reconstructed from traces: vs IB, retail -0.05826141007674013 (se 0.015563453669901545) and arb 0.11078591242849026 (se 0.00670908456815188). w07 reports reconstructed EDGE deltas vs HA. w08 and w09 report real Y-VOLUME deltas, because the engine exposes no edge split. w08 also reports replica holdout edge deltas vs IB: retail -0.052500830088624474 and arb 0.09444341757055962.",
  "Items disagree on Gate A. All four measuring items replayed the same IB port against the same 50 sequences (seeds 310000\u2013310049, 100000 swap callbacks). w06 and w07 report max |fee diff| 6.64202309125228e-05 bps; w08 and w09 report 5.224879067594612e-05 bps. Both pass the 0.01 bps threshold by a wide margin, so the outcome is the same. The difference probably comes from different callback generators (w09 says it adapted the RS2 differential-test generator), but this is not resolved here.",
  "All four measuring items report the same Gate B: replica IB\u2212T = 1.708612109706731 (se 0.04425219850674254) on seeds 7132768\u20137133279, against the reference 1.749 with tolerance 0.15. The replica is deterministic, so identical values are expected. The 1.749 reference was supplied by the task, and no item re-measured it on the real engine (w06 says so explicitly).",
  "Shared diagnostic seed: w06 and w08 each ran one IB-only real-engine diagnostic on seed 7590000, outside their confirmation blocks. Both say it was not used for selection or confirmation.",
  "Source re-check: w06, w07, w08 and w09 each cited 3 sources that did not hold when re-read (a page that could not be read, or one that does not say what it was quoted for). Their sources.json files were not given to this merge, so it cannot tell which claims those were. In these results, the claims that rest on outside sources are the input-file sha256 checks against the api.imdnotdead.fun artifact pages (T, HA, replica, IB, IB port, RS2 port), the pinned repository/wheel, and the Gate B reference 1.749. Weigh those accordingly. The measured numbers (speed, search, holdout, real-engine, gas) are each item's own report and do not depend on those sources. This merge independently checked two things: each input file's sha256 matches the hash it was given under, and each candidate's code_zb64 decodes to the code_sha256 it states (w06 ce003381\u2026ecfec, w07/w09 301941829a44\u2026d686, w08 5dfd0a55\u2026773a).",
  "w07 and w09 report valid=true, but this only means the unchanged IB passed the gates and gas checks. Neither is a trained candidate, so neither appears under `weights` or is considered for `best`."
 ]
}
```