# Merge 12 of batch r7-t03-joint: T3 continuation items j111–j120

**Inputs.** The ten item results under `.imd/reads/artifacts/` (`item-j111` … `item-j120`). I hashed every file locally and each sha256 equals its id (listed per row as `result_sha256`). For every item that returns code, I base64-decoded and zlib-decompressed `code_zb64` and its sha256 equals the item's `code_sha256`. Nothing below was re-run. Every figure comes from the item named next to it. The only additions are pooled statistics that I recomputed from the items' own `n`/`sum`/`sumsq`. Those are marked as recomputed.

## Outcome in one paragraph

Ten items were assigned. **Three ran the CMA-ES search** (j111, j116, j117). **Six ran real-only** because their replica measured under 500 markets/s (j112, j113, j114, j118, j119, j120), so they compare unchanged T3 with HA. **One was blocked** at SETUP with no network (j115). All three search candidates are valid: 0 OOG and an afterSwap peak ≤ 245,000, with gates A, B and C passed. All three beat T3 on the real engine. **Best is j116.** Its replica holdout is +2.6792 se 0.0139 over T and +0.1461 se 0.0056 over T3. On the real engine it is +0.1520 se 0.0155 and +0.1457 se 0.0146 over T3 on two fresh 1,024-seed blocks (pooled, recomputed: +0.1488 se 0.0107, n 2048). Its training curve was **still rising** when it stopped at 45 min (+2.585 → +2.730 monitor over T). j111 and j117 used population 16, 512 markets per generation and ~215–292 generations; their curves were nearly flat, with real gains of only +0.03 over T3. The real-only items agree on T3 − HA: four full 1,024-seed blocks pool to **+1.4076 se 0.0170** (n 4096, recomputed). j114 (n 696) and j120 (n 64) ran partial blocks. Combining j116 over T3 with T3 over HA gives roughly +1.56 over HA. That figure is an *inference across different seed sets, not a measured pair*. It is far below the ~+3.1 over HA the brief says the top 10 needs.

## Search rows (candidate vs T3)

| key | setting as reported | markets/s | gens | search min | holdout over T (mean, se) | holdout over T3 (mean, se) | real block 1 cand−T3 | real block 2 cand−T3 | afterSwap peak | OOG | valid | last curve point (min, gain over T, sigma) | code_sha256 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| j111 | pop 16, sigma0 0.5, cap 0.5^(min/40), 512 mkts/gen, ES seed 911 | 641.9 | 215 | 48.51 | 2.55155690350794, 0.013692927971677555 | 0.01846059714524522, 0.004239042514454816 | 9140000..9141023 n 1024: 0.04465715070801052 se 0.010782220528688553 | 9142000..9143023 n 1024: 0.0246904019412168 se 0.012342272357916066 | 238338 | 0 | true | 45.06, 2.6015, 0.229 | 2b2f0782f3b793cc359e6e925598d964f538b5322c5ebeed9141b21f786cff69 |
| j116 | pop 32, sigma0 0.5, cap 0.5^(min/40), 256 mkts/gen, ES seed 916 | 4790.506911990077 | 1585 | 45.01 | 2.6792235880568813, 0.013921497266223975 | 0.1461272816941867, 0.0056156559544352554 | 9160000..9161023 n 1024: 0.1519548620352234 se 0.015513530413623456 | 9162000..9163023 n 1024: 0.14572318003837537 se 0.014602480038375245 | 238255 | 0 | true | 45.01, 2.7302, 0.1714 | a4a71517cdea07042fa2b79387280832d21c3e0a36ae6c574e8d88bd5fdc88b1 |
| j117 | setting "A": pop 16, sigma0 0.5, cap 0.5^(min/40), 512 mkts/gen, ES seed 917 | 873.048336204096 | 292 | 55.32785500754641 | 2.559124631753906, 0.013599162953666834 | 0.02602832539121102, 0.00395222016534047 | 9164000..9165023 n 1024: 0.02244108713441262 se 0.01129311467177925 | 9166000..9167023 n 1024: 0.031184386358330207 se 0.011444536398943933 | 237662 | 0 | true | 55.116764984179945, 2.6111555365608745, 0.19238668713699583 | 0699ba660d4bd0909ba79690f42ebbe8dbbd77d825dc286bb8b7d29df9cf8acc |

The block sums and sumsq are in the JSON below. Pooled over each item's two blocks (recomputed from sum/sumsq): j111 +0.034674 se 0.008195; **j116 +0.148839 se 0.010650**; j117 +0.026813 se 0.008038 (each n 2048).

**The replica predicted the real result in all three searches.** Holdout over T3 compared with pooled real over T3: j111 +0.018 vs +0.035; j116 +0.146 vs +0.149; j117 +0.026 vs +0.027. Every gap is well inside the 0.15 tolerance. This agrees with the pilot's finding that the replica tracks real gains.

**Training curves (monitor gain over T, 2,048 seeds 30000000..30002047):**
- j111: 2.5846 → 2.5776 → 2.5699 → 2.5866 → 2.5931 → 2.5744 → **2.6085 (30.09 min, frozen)** → 2.5979 → 2.6026 → 2.6015. The item calls this "essentially flat".
- j116: 2.5846 → 2.613 → 2.6256 → 2.6475 → 2.6692 → 2.6741 → 2.6904 → 2.6944 → 2.7201 → **2.7302 (45.01 min, frozen)**. It rose at every check, and the item notes it was "still rising when the search stopped".
- j117: 2.5846 → … → **2.6112 (55.12 min, frozen)**. Slowly rising.

j116 differs from the two flat runs in more than one way. It used population 32 with 256 markets per generation, and it ran about 5.4–7.4× as many generations thanks to a 4790 markets/s persistent 24-process pool. The others ran at 642 and 873 markets/s. These items cannot separate the effect of population size from the effect of generation count.

**Real-engine side figures for the search candidates (cand − T3, mean per market).** retail_delta and arb_delta are **null in every item**: run_batch reports edges, pnl and volumes but does not split edge into retail and arb. j111: pnl +0.0542 / +0.0381; retail vol_y −67.27 / −67.82; arb vol_y −50.43 / −54.85. j116: pnl +0.1476 / +0.1425; retail vol_y +9.04 / +8.80; arb vol_y +14.51 / +12.15. j117: pnl +0.0353 / +0.0394; retail vol_y −150.12 / −148.63; arb vol_y −113.06 / −105.70. worst5 seeds: j111 block 2 worst is 9142659 at −4.193. j116 worst seeds are 9162533 at −2.296 and 9160666 at −1.892. j117 worst is 9164462 at −2.902. Full lists are in the items.

**Gates in the search items.** Gate A (T3 port vs T3 EVM, 50 × 2,000 callbacks, seeds 310000..310049) gave max 5.747e-5 bps in all three. Gate B (replica CG − T, seeds 7232768..7233279) gave +1.920372 se 0.044459 in all three, within 0.15 of +2.00. Gate C (frozen Solidity vs frozen port) gave max 5.53e-5 bps (j111), 6.73e-5 bps (j116) and 6.16e-5 bps (j117), all under 0.1.

## Real-only rows (T3 vs HA; no search, code = T3 unchanged)

| key | replica markets/s | seeds | n | T3−HA mean | se | sum | sumsq | T3 afterSwap peak | HA afterSwap peak | OOG (T3/HA) | item valid | note |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| j112 | 405.0516795721294 | 9144000..9145023 | 1024 | 1.4404623439772064 | 0.02830459689756057 | 1475.0334402326594 | 2963.9766268692183 | 235515 | 231567 | 0/0 | true | full block; requested setting "E" (pop 32) not run |
| j113 | 489.6 | 9148000..9149023 | 1024 | 1.4196110607627 | 0.029460974864617198 | 1453.6817262210047 | 2972.8844100943847 | 237777 | 233838 | 0/0 | true | full block |
| j114 | 52.433865449355054 | subset of 9152000..9153023 (seen 9152000..9152847) | 696 | 1.4106182593534478 | 0.0350075699537654 | 981.7903085099997 | 1977.744685422126 | 237600 | 233620 | 0/0 | true | **PARTIAL**: 696 of 1,024 seeds, loaded shared host |
| j118 | 431.6 | 9168000..9169023 | 1024 | 1.3535807563017905 | 0.04561373904707347 | 1386.0666944530335 | 4055.703712820937 | 238044 | 234159 | 0/0 | true | full block; outlier seed 9168361 = −35.885 (without it +1.38998 se 0.02751, n 1023, context only) |
| j119 | 105.38964505592465 | 9172000..9173023 | 1024 | 1.4167844845978195 | 0.029297842502375864 | 1450.7873122281671 | 2954.633434442643 | 237588 | 231535 | 0/0 | true | full block |
| j120 | 55.6 | 9176000..9176063 | 64 | 1.2953452924400422 | 0.09827536301659275 | 82.9020987161627 | 146.32808871278672 | 232506 | 228424 | 0/0 | **false** | **PARTIAL**: 64 of 1,024 seeds; **gate A not completed** (timed out), so the item itself marks valid=false |

**Pooled T3 − HA (recomputed from sum/sumsq):** the four full blocks (j112, j113, j118, j119) give n 4096, mean 1.407610, se 0.016972. All six give n 4856, mean 1.406561, se 0.015224. The blocks agree within their se. j118 is lowest only because of the one −35.9 seed. Win counts: j112 984/1024, j113 980/1024, j114 673/696, j118 976/1024, j119 971/1024, j120 62/64. Every T3 and HA peak is ≤ 245,000 with 0 OOG. Note that the brief quotes HA's own peak as 231k, while these items measured HA afterSwap peaks of 228,424–234,159.

The gates in the real-only items reproduce the same values: gate A max 5.747e-5 bps (j112, j113, j114, j118, j119) and gate B +1.920372. Gate C does not apply to them, since nothing was frozen.

## What failed or is missing

- **j115: blocked at SETUP.** `git clone https://github.com/benedictbrady/amm-challenge` returned "Proxy CONNECT aborted". The same error came back from github.com, api.imdnotdead.fun and pypi.org. The item has no speed, gates, search, real runs or hashes, and it reports nothing.
- **j114** measured T3 − HA on only 696 of the 1,024 planned seeds, on a host with load ~60. **j120** ran only 64 seeds and did not complete gate A, so its row is invalid by the brief. Both T3 − HA values are consistent with the full blocks, but their n is smaller.
- **No item paired a search candidate directly against HA.** No item measured T3 − HA on the same seeds as a candidate's blocks. Any "candidate over HA" figure is therefore an inference: j116 over T3 (+0.149) plus T3 over HA (+1.408) ≈ +1.56. The items do not measure that number.
- **No second real block** was run by the real-only items. Their protocol called for one block.
- **Retail/arb edge deltas are null** throughout, as explained above.
- Setting labels: only j117 ("A") and j112 ("E", not executed) state their setting key. For the others I report the settings by their parameters (population, sigma0, ES seed) as they recorded them. The planned settings of j113 (pop 16, sigma0 0.5, ES seed 913) and j118 (pop 16, sigma0 0.3, ES seed 918) were never run. j114, j119 and j120 give no requested setting.

## Sources that did not re-verify

The brief flags results j111 (1 source), j112 (2), j113 (1), j116 (1), j117 (1) and j119 (1) as citing sources that did not hold on re-read. The result files I was handed contain no URL list; j112 mentions a separate `sources.json` that was not handed to me. So **I cannot tell which claims those sources backed**. What the items say about provenance is this: every number in the tables above was measured locally by the item, on seeds the item names. I re-checked the paired statistics arithmetically from sum and sumsq. The hashes of the item files and of the code they deliver were also checked here. Claims that rest only on outside sources should be treated as unconfirmed. These include the leaderboard context (board 520.05, 27th; ~+3.1 over HA for the top 10), which appears only as the brief's own words quoted in j119. The input-artifact hashes the items report (T3 2333ecf2…, CG ebbd90c6…, replica e8a7a2af…, HA 2801efc8…, T3 code cb6d9455…, HA code 99e3a2ec…) agree across all nine items that ran. **No leaderboard placement is claimed by any item or by this merge.**

## Recommendation for the next batch

Continue from **j116's frozen params**. Its curve was still rising at its time limit. Use j116's configuration: population 32, 256 markets per generation, and a fast persistent pool. Then pair the result directly against HA on fresh seeds. The flat j111 and j117 runs suggest that 512-market, population-16 runs at under 900 markets/s gain little in an hour.

## Merge table (JSON; numbers as strings, `best.params` copied whole as the item gives it)

```json
{
 "rows": [
  {
   "key": "j111",
   "mode": "search",
   "result_sha256": "03113f758cb701b3a281bdeef028ae67237a982d4602563e3d2e4bdda1d5c9f9",
   "markets_per_s": "641.9",
   "generations": "215",
   "search_minutes": "48.51",
   "holdout_over_T": {
    "mean": "2.55155690350794",
    "se": "0.013692927971677555"
   },
   "holdout_over_base": {
    "mean": "0.01846059714524522",
    "se": "0.004239042514454816"
   },
   "real_candidate_minus_T3": [
    {
     "seeds": "9140000..9141023",
     "n": "1024",
     "mean": "0.04465715070801052",
     "se": "0.010782220528688553",
     "sum": "45.72892232500277",
     "sumsq": "123.82662150943139"
    },
    {
     "seeds": "9142000..9143023",
     "n": "1024",
     "mean": "0.0246904019412168",
     "se": "0.012342272357916066",
     "sum": "25.282971587806003",
     "sumsq": "160.1996100659286"
    }
   ],
   "afterSwap_peak": "238338",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": {
    "minutes": "45.06",
    "monitor_gain_over_T": "2.6015",
    "sigma": "0.229"
   },
   "code_sha256": "2b2f0782f3b793cc359e6e925598d964f538b5322c5ebeed9141b21f786cff69"
  },
  {
   "key": "j112",
   "mode": "real-only",
   "result_sha256": "5f7e596fb38d178b4eb89b830f525caaababf032bea717f24e5a2d4dc99c4caa",
   "markets_per_s": "405.0516795721294",
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": "235515",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": null,
   "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9"
  },
  {
   "key": "j113",
   "mode": "real-only",
   "result_sha256": "d1abd490c7e96efb901e15443346d8d32ff7335f20b5adf1b2ddf9d9a7814a04",
   "markets_per_s": "489.6",
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": "237777",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": null,
   "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9"
  },
  {
   "key": "j114",
   "mode": "real-only",
   "result_sha256": "3428951e26959b5405405a16ffe2633c2f8b2be36683b7799388fda76ea881a9",
   "markets_per_s": "52.433865449355054",
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": "237600",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": null,
   "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9"
  },
  {
   "key": "j115",
   "mode": "blocked",
   "result_sha256": "dbfb036fe5effcce14ccc6d77fcf76ac265e9bfe29f7978c8ae3b6db7254bf40",
   "markets_per_s": null,
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": null,
   "OOG": null,
   "valid": "false",
   "last_curve_point": null,
   "code_sha256": null
  },
  {
   "key": "j116",
   "mode": "search",
   "result_sha256": "edb467cbe3b57f2a8c91d57d50729d98ee07a1f5b8adfe683916a1cc923f09b9",
   "markets_per_s": "4790.506911990077",
   "generations": "1585",
   "search_minutes": "45.01",
   "holdout_over_T": {
    "mean": "2.6792235880568813",
    "se": "0.013921497266223975"
   },
   "holdout_over_base": {
    "mean": "0.1461272816941867",
    "se": "0.0056156559544352554"
   },
   "real_candidate_minus_T3": [
    {
     "seeds": "9160000..9161023",
     "n": "1024",
     "mean": "0.1519548620352234",
     "se": "0.015513530413623456",
     "sum": "155.60177872406877",
     "sumsq": "275.75839476340263"
    },
    {
     "seeds": "9162000..9163023",
     "n": "1024",
     "mean": "0.14572318003837537",
     "se": "0.014602480038375245",
     "sum": "149.22053635929637",
     "sumsq": "245.1169425478458"
    }
   ],
   "afterSwap_peak": "238255",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": {
    "minutes": "45.01",
    "monitor_gain_over_T": "2.7302",
    "sigma": "0.1714"
   },
   "code_sha256": "a4a71517cdea07042fa2b79387280832d21c3e0a36ae6c574e8d88bd5fdc88b1"
  },
  {
   "key": "j117",
   "mode": "search",
   "result_sha256": "2f0e80ba7b5e2d6afa55729c0af55e0458a7df85a37b1e67797d8b0866d61b71",
   "markets_per_s": "873.048336204096",
   "generations": "292",
   "search_minutes": "55.32785500754641",
   "holdout_over_T": {
    "mean": "2.559124631753906",
    "se": "0.013599162953666834"
   },
   "holdout_over_base": {
    "mean": "0.02602832539121102",
    "se": "0.00395222016534047"
   },
   "real_candidate_minus_T3": [
    {
     "seeds": "9164000..9165023",
     "n": "1024",
     "mean": "0.02244108713441262",
     "se": "0.01129311467177925",
     "sum": "22.97967322563852",
     "sumsq": "134.11464548198296"
    },
    {
     "seeds": "9166000..9167023",
     "n": "1024",
     "mean": "0.031184386358330207",
     "se": "0.011444536398943933",
     "sum": "31.932811630930132",
     "sumsq": "138.20145648352613"
    }
   ],
   "afterSwap_peak": "237662",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": {
    "minutes": "55.116764984179945",
    "monitor_gain_over_T": "2.6111555365608745",
    "sigma": "0.19238668713699583"
   },
   "code_sha256": "0699ba660d4bd0909ba79690f42ebbe8dbbd77d825dc286bb8b7d29df9cf8acc"
  },
  {
   "key": "j118",
   "mode": "real-only",
   "result_sha256": "31daaae74a761a92d93e502fa8640f97891dc154df3e3f73804df36c5b1ae53a",
   "markets_per_s": "431.6",
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": "238044",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": null,
   "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9"
  },
  {
   "key": "j119",
   "mode": "real-only",
   "result_sha256": "386f9969be9ae0722af1039de0fc9bb0dd63504a99e59e5c73bdaedae522a4b3",
   "markets_per_s": "105.38964505592465",
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": "237588",
   "OOG": "0",
   "valid": "true",
   "last_curve_point": null,
   "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9"
  },
  {
   "key": "j120",
   "mode": "real-only",
   "result_sha256": "216518027ce290d8870f60bc415c9ffc52a1bfd73ff6dbc3b8847235565c686e",
   "markets_per_s": "55.6",
   "generations": null,
   "search_minutes": null,
   "holdout_over_T": null,
   "holdout_over_base": null,
   "real_candidate_minus_T3": null,
   "afterSwap_peak": "232506",
   "OOG": "0",
   "valid": "false",
   "last_curve_point": null,
   "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9"
  }
 ],
 "real_only": [
  {
   "key": "j112",
   "T3_minus_HA": {
    "seeds": "9144000..9145023",
    "n": "1024",
    "mean": "1.4404623439772064",
    "se": "0.02830459689756057",
    "sum": "1475.0334402326594",
    "sumsq": "2963.9766268692183"
   },
   "gas": {
    "T3_afterSwap_peak": "235515",
    "HA_afterSwap_peak": "231567",
    "T3_OOG": "0",
    "HA_OOG": "0"
   },
   "note": "full block"
  },
  {
   "key": "j113",
   "T3_minus_HA": {
    "seeds": "9148000..9149023",
    "n": "1024",
    "mean": "1.4196110607627",
    "se": "0.029460974864617198",
    "sum": "1453.6817262210047",
    "sumsq": "2972.8844100943847"
   },
   "gas": {
    "T3_afterSwap_peak": "237777",
    "HA_afterSwap_peak": "233838",
    "T3_OOG": "0",
    "HA_OOG": "0"
   },
   "note": "full block"
  },
  {
   "key": "j114",
   "T3_minus_HA": {
    "seeds": "9152000..9152847",
    "n": "696",
    "mean": "1.4106182593534478",
    "se": "0.0350075699537654",
    "sum": "981.7903085099997",
    "sumsq": "1977.744685422126"
   },
   "gas": {
    "T3_afterSwap_peak": "237600",
    "HA_afterSwap_peak": "233620",
    "T3_OOG": "0",
    "HA_OOG": "0"
   },
   "note": "PARTIAL: 696 of 1024 planned seeds 9152000..9153023 (two interleaved sub-ranges; seeds finished by both arms), host load ~60"
  },
  {
   "key": "j118",
   "T3_minus_HA": {
    "seeds": "9168000..9169023",
    "n": "1024",
    "mean": "1.3535807563017905",
    "se": "0.04561373904707347",
    "sum": "1386.0666944530335",
    "sumsq": "4055.703712820937"
   },
   "gas": {
    "T3_afterSwap_peak": "238044",
    "HA_afterSwap_peak": "234159",
    "T3_OOG": "0",
    "HA_OOG": "0"
   },
   "note": "full block; one outlier seed 9168361 at -35.885; without it +1.38998 se 0.02751 (n 1023)"
  },
  {
   "key": "j119",
   "T3_minus_HA": {
    "seeds": "9172000..9173023",
    "n": "1024",
    "mean": "1.4167844845978195",
    "se": "0.029297842502375864",
    "sum": "1450.7873122281671",
    "sumsq": "2954.633434442643"
   },
   "gas": {
    "T3_afterSwap_peak": "237588",
    "HA_afterSwap_peak": "231535",
    "T3_OOG": "0",
    "HA_OOG": "0"
   },
   "note": "full block"
  },
  {
   "key": "j120",
   "T3_minus_HA": {
    "seeds": "9176000..9176063",
    "n": "64",
    "mean": "1.2953452924400422",
    "se": "0.09827536301659275",
    "sum": "82.9020987161627",
    "sumsq": "146.32808871278672"
   },
   "gas": {
    "T3_afterSwap_peak": "232506",
    "HA_afterSwap_peak": "228424",
    "T3_OOG": "0",
    "HA_OOG": "0"
   },
   "note": "PARTIAL: 64 of 1024 planned seeds; gate A not completed, so item marks valid=false"
  }
 ]
}
```

### best (j116): code_zb64, code_sha256 and params copied whole

```json
{"key":"j116","holdout_over_T":{"mean":"2.6792235880568813","se":"0.013921497266223975"},"code_sha256":"a4a71517cdea07042fa2b79387280832d21c3e0a36ae6c574e8d88bd5fdc88b1","code_zb64":"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","params":{"W1":[[-0.082977,0.682329,-0.60538,2.948929,-2.356183,0.623987,-1.317726,2.381124,-5.111612,-1.476209,-0.568143,-0.101127,0.561337,-2.294544,-0.786511,-0.363068,-1.081398,-0.424992,1.355753],[0.814203,0.715035,1.946889,2.576131,-0.124678,-0.131594,-0.487946,0.395572,-0.291096,0.106336,-0.092604,-1.966528,0.07505,-0.105537,-0.115901,0.957076,0.115398,0.077732,1.230323],[0.545618,3.687687,-1.641354,0.532174,0.444191,2.153389,-1.520494,0.841077,1.074486,-0.430255,0.466964,-0.25412,0.27532,0.428243,0.283849,1.330833,-2.637015,-0.364054,-2.602183],[0.383571,-1.508262,0.924422,-0.371869,-1.183641,-0.650755,0.249829,-0.254257,0.84554,-0.259966,0.05571,-1.776867,-0.074418,0.535237,0.062051,-1.232336,0.882907,-0.44206,-0.857229],[1.256749,0.483601,4.879749,0.115338,-2.074851,-1.444683,-1.117552,-0.56862,-0.196379,0.527544,-0.080161,-2.71902,-0.106659,-0.945154,0.391015,-0.816536,2.03396,-0.036808,0.502549],[1.254668,-1.760863,0.515829,-1.465411,3.457185,1.000709,4.018012,0.574098,-0.552588,-0.022626,0.12728,3.442701,0.540358,-1.267463,0.13912,0.386427,-1.486193,-0.346788,-1.516923],[-0.752438,1.160596,0.086043,-0.679991,2.376766,0.963303,-0.887696,2.181827,0.997908,-1.420528,1.155045,1.011525,1.514322,-1.062683,-1.097546,-1.545485,0.708801,-0.628201,2.676307],[-2.499134,-2.493589,3.672738,3.401011,-0.60535,-0.649782,-1.832022,2.618913,0.729198,-0.870971,2.539369,-2.167703,1.525013,0.273172,0.53827,0.8114,-0.433504,-1.90206,1.527826]],"b1":[2.126401,-1.468462,-4.5517,-0.680545,-0.85762,-2.373849,-0.630174,0.097175],"W2":[[0.268551,-1.973835,1.200105,-0.924193,-2.831068,-0.741379,-1.157824,0.875959],[0.543126,-1.71905,-1.927051,3.499361,0.21225,-1.885591,0.883412,-0.078033]],"b2":[1.606455,3.302981],"constants":{"ELAPSED_CAP":8.0,"SIGNAL_THRESHOLD":0.001167737961145088,"DIR_DECAY":0.7262087140140295,"ACT_DECAY":0.3069226737978946,"SIZE_DECAY":0.4954509958768,"TOX_DECAY":0.9335430787396097,"SIGMA_DECAY":0.7056018344231254,"LAMBDA_DECAY":0.9917207184962149,"SIZE_BLEND_DECAY":0.8486556015247645,"TOX_BLEND_DECAY":0.2648515931887838,"ACT_BLEND_DECAY":0.9880969562200754,"PHAT_ALPHA":0.25105693768362125,"PHAT_ALPHA_RETAIL":0.05148647398976582,"PHAT_ALPHA_SAME":0.1345980490219886,"DIR_IMPACT_MULT":4.0,"GATE_SIGMA_MULT":6.170957565932043,"MIN_GATE":0.029432413013611224,"TRADE_TOX_BOOST":0.2829124396608578,"STALE_ATTRACT_FRAC":0.6668279567906397,"RET_CAP":0.022448226444013897,"TOX_CAP":0.19801508539161142,"TRADE_RATIO_CAP":0.17396333817832985,"LAMBDA_CAP":1.15388721797511,"STEP_COUNT_CAP":64.0,"BASE_FEE":0.000355575296438696,"SIGMA_COEF":0.20328058996615495,"LAMBDA_COEF":0.001255705437665505,"FLOW_SIZE_COEF":0.48725400120991635,"TOX_COEF":0.02777142349093952,"TOX_QUAD_COEF":1.6271932867187406,"ACT_COEF":8.825204784373149,"DIR_COEF":0.001929556904020044,"DIR_TOX_COEF":0.006860727146462987,"STALE_DIR_COEF":0.7551862661710563,"SIGMA_TOX_COEF":0.05816428837052452,"TAIL_KNEE":0.056159422942006085,"TAIL_SLOPE_PROTECT":0.8359662674973607,"TAIL_SLOPE_ATTRACT":0.8419529984766807,"ANCHOR_COUNT_CAP":672.0,"ANCHOR_LINEAR_K":0.593701935471862,"ANCHOR_QUADRATIC_K":1.0977621411226743},"scaling":{"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_size","step_trade_count","prev_bid_fee","prev_ask_fee","steps_since_arb_like_print","router_implied_vanilla_spot_dev"],"feature_multipliers":[1000.0,2.0,500.0,500.0,500.0,1.0,500.0,500.0,2.0,2.0,1000000.0,1.0,500.0,647.054817,1.395287,220.617911,228.148131,0.412565,274.335453],"feature_offsets":[-1.0,-1.0,-1.0,-1.0,-1.0,-1.0,0.0,0.0,-1.0,-1.0,-1.0,0.0,-1.0,-0.618293,-2.037262,-0.803057,-0.808615,-0.823811,0.001772],"weight_fixed_point_scale":1000000,"input_clip":[-4,4],"hidden_relu_clip":[0,4],"output_unit_fee":0.0001,"note":"unchanged from CG"}}}
```