id stringlengths 36 36 | tier stringclasses 1
value | github_login stringclasses 179
values | created_at stringlengths 32 32 | status stringclasses 2
values | sha256 stringlengths 64 64 | source_bytes int64 341 260k | leaderboard_rank int64 | source large_stringlengths 341 260k | metadata_json large_stringlengths 645 734 | result_json large_stringlengths 5 4.8k |
|---|---|---|---|---|---|---|---|---|---|---|
fef6a407-f976-4813-8698-65ea8f2efd4c | easy | DDanlov | 2026-08-13 19:42:22.020416+00:00 | succeeded | 939f58d818c3f953053dfcbc068030b31142e50f323a87b6d7f84ab5df0f2dd5 | 26,904 | null | from __future__ import annotations
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
try:
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
except ... | {
"id": "fef6a407-f976-4813-8698-65ea8f2efd4c",
"created_at": "2026-08-13 19:42:22.020416+00:00",
"db_md5": "e3de2d9e8ccead933e2b8372c4ddba6b",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "4c7ad94d-7be6-4f7f-8f4e-637872f06475",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.308775201216343,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08333333387970925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3420049254573994,
"example_count": 100,... |
feffa4b2-dc9a-4015-8fed-0f43a33f0804 | easy | chad-atexpedient | 2026-08-05 22:50:06.236919+00:00 | succeeded | 9282a3a7d00bc6d102bc70304efba49bb93fa54aab7b6eeffc2a5409fec3d99c | 11,132 | null | """PR-C: fixed-depth R028F with full immutable input recall."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
LANES = 6
CHANNELS = ... | {
"id": "feffa4b2-dc9a-4015-8fed-0f43a33f0804",
"created_at": "2026-08-05 22:50:06.236919+00:00",
"db_md5": "f10c988d0e81dfa119323776e1cf7390",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "16e538ba-45fa-461d-96a5-6c0150efb234",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.175557365944177,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00791666670391957
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1725889513845402,
"example_count": 600,... |
ff081248-f600-40c6-a133-045783f76c68 | easy | EyimofeA | 2026-08-10 13:53:16.776996+00:00 | succeeded | 21d8cff8feeb82c53bac0de652748f32cceaaac77eef4798640a95a1e0f674e5 | 10,693 | null | """Competition-legal multi-lane local recurrent grid.
The model uses generic learned scratch lanes and tied local updates. It has no
arithmetic trace, carry target, task solver, or hard-coded numeric transition.
Training uses only evaluator-provided final labels.
"""
from __future__ import annotations
import math
imp... | {
"id": "ff081248-f600-40c6-a133-045783f76c68",
"created_at": "2026-08-10 13:53:16.776996+00:00",
"db_md5": "8fb3e1990d041360f7b0940dc8a0f399",
"submitter": "mof",
"github_login": "EyimofeA",
"run_id": "f6efda2c-8e48-4a6b-a14c-4d392a30c0ae",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.4161855361952975,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003333333353511989
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3504607677459717,
"example_count": 60... |
ff08c27a-7ca4-4f88-a716-09602ce10d13 | easy | shreyash-chonkie | 2026-08-24 20:56:09.328403+00:00 | succeeded | c4faa44cf9b04caba9e331cce26ed8e1a7a4002f3b68ac3fa79ad2fc2212275d | 15,807 | null | """Eight-step recurrent attention with local loss trends and separation."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
... | {
"id": "ff08c27a-7ca4-4f88-a716-09602ce10d13",
"created_at": "2026-08-24 20:56:09.328403+00:00",
"db_md5": "23100c6111f7ed7c8f71bb2887e59d80",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "eeaf6560-beaf-4a83-96ac-0196fa624b44",
"tier": "easy",
"dataset_id": "e6",
"status": "... | {
"score": {
"mean_loss": 4.030854225158691,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.026801803149282932
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.574614524841309,
"example_count": 60,
... |
ff0a36f5-8b83-4316-9c4d-e84c217c15e0 | easy | oupadhyay | 2026-08-08 09:34:37.740215+00:00 | succeeded | d94c570c352412c2f347b893c3890fd70cba3ab3de844fcfafb49a1e47b29d29 | 7,426 | null | """Generic universal recurrent Transformer candidate."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
D = 256
HEADS = 8
FFN = 768
SCRATCH = 32... | {
"id": "ff0a36f5-8b83-4316-9c4d-e84c217c15e0",
"created_at": "2026-08-08 09:34:37.740215+00:00",
"db_md5": "39607075a8a454c92d9443a035783a1e",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "a9952746-43bb-43bf-8219-4ff9569cc45f",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 3.1694401128402165,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0087500002173086
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2927026748657227,
"example_count": 600,... |
ff0c5ded-b9de-4cb9-b956-06772369e46a | easy | newjordan | 2026-08-22 18:38:15.727050+00:00 | succeeded | 7e3f4064db354271d59ffaab3e31042f7e33f3aa32214d67f40ed53b9793465a | 18,977 | null | """Neural Transition Cell: learned local computation with tied recurrence.
The model parses the public decimal fields only into fixed-width one-hot
registers. A small convolutional gated cell, shared across digit positions
and refinement steps, learns one state transition from evaluator labels. The
same learned tran... | {
"id": "ff0c5ded-b9de-4cb9-b956-06772369e46a",
"created_at": "2026-08-22 18:38:15.727050+00:00",
"db_md5": "e95d54770614860813b319808cc7c85e",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "bfde1b96-2dd0-4e2c-920a-a05c3c87b2d8",
"tier": "easy",
"dataset_id": "e6",
"status": "succeed... | {
"score": {
"mean_loss": 1.345544844865799,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.6353603899478912
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.9919772744178772,
"example_count": 60,
... |
ff15705f-4923-47c7-901b-a2dc48747c94 | easy | DDanlov | 2026-08-09 00:14:29.480928+00:00 | failed | bc84a28bc573fc5e52deedbbae09c79a7f780a6b0bf7c9bcc9873b5fe06fbf68 | 15,978 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with eigenvalue ma... | {
"id": "ff15705f-4923-47c7-901b-a2dc48747c94",
"created_at": "2026-08-09 00:14:29.480928+00:00",
"db_md5": "ab717b6b65e8c5fac5e0ad68a12f59a5",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "8f6e808b-aa02-498e-b942-8c091454d7e7",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
ff159c5f-1a26-48a6-bec5-5202d88feeac | easy | DESU-CLUB | 2026-08-24 10:41:48.866238+00:00 | succeeded | c08e97fdaa7801930216c9f5dc862a783045e65f0870f69365bb3995d78f5011 | 10,580 | null | """Huginn-style looped Transformer whose iteration count is read from the T token.
prelude -> [core applied T times, input re-injected each step] -> coda
Forked from experiments/_baselines/adamw_transformer.py; the Block, RMSNorm and
optimizer are unchanged so that the only varied axis is the depth mechanism.
"""
fr... | {
"id": "ff159c5f-1a26-48a6-bec5-5202d88feeac",
"created_at": "2026-08-24 10:41:48.866238+00:00",
"db_md5": "89acb12baa48f7227e6269e0df3aab96",
"submitter": "Low Keng Hoong, Warren",
"github_login": "DESU-CLUB",
"run_id": "2d319347-e6d7-4f98-80e6-ec16232cbd96",
"tier": "easy",
"dataset_id": "e6",
"sta... | {
"score": {
"mean_loss": 8.171068136006186,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08040540992609552
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.9129319190979,
"example_count": 60,
... |
ff1875eb-cf8f-4ca0-91e2-ce2bd9b37afb | easy | viridale | 2026-08-03 21:43:47.871240+00:00 | succeeded | 4cda0ca962056cf33a776d62d228d1dc0ee5863efb6b881a441e94825634814b | 144,966 | null | """CUDA-wide mismatch-triggered broad evidence scan.
hypothesis: H100 fallback latency is dominated by Python/kernel launch overhead;
a CUDA-only 4194304-atom chunk uses measured memory headroom to reduce it.
axis: arch
target: revised h1 Max T at least1 with faster unsupported-route commitment.
expected_delta: pr... | {
"id": "ff1875eb-cf8f-4ca0-91e2-ce2bd9b37afb",
"created_at": "2026-08-03 21:43:47.871240+00:00",
"db_md5": "990f3a734d7da60a1d7b7352bea851f2",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "a3a00b7c-4798-4746-9627-ade3b5398028",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 0.003020370119402878,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.003020370181758508,
"example_count": 600,
... |
ff19b106-b607-4ac3-8175-af5e229f6470 | easy | 0Chris5R | 2026-08-07 09:33:42.094889+00:00 | succeeded | e05836b28dca21757c73c53a09009c1c020939c2f0ad9f3a89c67755fc09ad27 | 12,651 | null | """Learned convolutional-GRU over an aligned field-by-digit workspace.
The public field markers and decimal significance define only the layout of
the neural workspace. Every transition and output is learned; there is no
arithmetic routine, recurrence assumption, lookup table, or generated target.
"""
from __future__... | {
"id": "ff19b106-b607-4ac3-8175-af5e229f6470",
"created_at": "2026-08-07 09:33:42.094889+00:00",
"db_md5": "dcc37bc7661a9fa872e067430ddd8c3b",
"submitter": "Chris ",
"github_login": "0Chris5R",
"run_id": "cdf03135-dc3d-4db1-a10d-52f38fb6c4d3",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 7.333926918958732,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333358168601
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.839098177683193,
"example_count": 600,... |
ff1a1146-c5d6-4bdb-9583-88403cc99cd8 | easy | jordanrubin | 2026-08-12 05:48:21.702463+00:00 | succeeded | bedf3454ce1c03710e3324993d467ae463c1638427b29d6e399f344c0613b17d | 24,919 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "ff1a1146-c5d6-4bdb-9583-88403cc99cd8",
"created_at": "2026-08-12 05:48:21.702463+00:00",
"db_md5": "50830d58ed5ea7674110ea515b39ce84",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "16213ccf-47a1-4f05-a7d9-1122034aa1a7",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.6899150686301443,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01666666637174785
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1814218396028595,
"example_count": 600... |
ff1b4bdc-5323-4856-a344-d9ab357d9950 | easy | oupadhyay | 2026-08-16 17:50:34.387914+00:00 | succeeded | 933c843abaa7a388adaa05e1792f94eb9436821515cce10428ffe4f0fd1366be | 4,272 | null | """Dynamic conditional T1-weight screen: weight4_d64."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,TokenLossBatch,assert_model_state
W,D,H=None,64,32
class C:
def __init__(s,vocab... | {
"id": "ff1b4bdc-5323-4856-a344-d9ab357d9950",
"created_at": "2026-08-16 17:50:34.387914+00:00",
"db_md5": "e707e35a76fa3f43dc42335fe8867ceb",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "8139a18a-c12c-46d9-86d6-0105e1510b63",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.3007966718003496,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007083333345750968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.16382624001781,
"example_count": 600,... |
ff1cea41-d7b8-4885-840e-41c1a6c56bdb | easy | Dandandan | 2026-08-04 18:16:16.635159+00:00 | succeeded | 9205c2ff8ed1151c01a30024979eb97c465adf90d0b2f94c5f4ef4574c2e5d8f | 17,274 | null | """Multi-start learned spectral recurrence for One Layer Deeper.
Every coefficient that determines a prediction is randomly initialized and
updated end to end. Shared candidate transitions are unrolled to the prompt's
training depth; self-calibration and an integer-boundary loss favor a reusable
functional root withou... | {
"id": "ff1cea41-d7b8-4885-840e-41c1a6c56bdb",
"created_at": "2026-08-04 18:16:16.635159+00:00",
"db_md5": "788d65c6bd5adfda3d297ed60ee070af",
"submitter": "Daniël Heres",
"github_login": "Dandandan",
"run_id": "c38fbda8-3ae9-4f6e-8e97-686fd9bb4f4d",
"tier": "easy",
"dataset_id": "e4",
"status": "suc... | {
"score": {
"mean_loss": 1.1314449047815334e-05,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.1313680157761708e-05,
"example_count": 1200,
... |
ff206746-ca07-4c7a-8068-07fec9a49f4c | easy | ddoan | 2026-08-19 05:15:28.499665+00:00 | succeeded | 946c0e986e33f1b29f2b149740fa2221100dc968bed8d3023b309b052104007f | 13,259 | null | """Contest-contract T=1 Newton difference-ladder experiment.
The model is a generic degree-three falling-factorial executor over decimal
prompt fields. Its four coefficients are learned; no target polynomial is
stored. The boundary-preserving loss finds unit-successor windows among the
T=1 rows already present in th... | {
"id": "ff206746-ca07-4c7a-8068-07fec9a49f4c",
"created_at": "2026-08-19 05:15:28.499665+00:00",
"db_md5": "2ec0bf54c548fd117c723a24ff31794d",
"submitter": "Doug Doan",
"github_login": "ddoan",
"run_id": "454f0bf2-a568-469a-b5d2-5e900bd03996",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 15.58665657043457,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0283333333209157
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 15.295624732971191,
"example_count": 100,
... |
ff279779-cd31-4483-854e-ff7abdfd267e | easy | arthurfeeney | 2026-08-07 07:32:01.788600+00:00 | succeeded | f57d5ba23656ca80f0703a51811648d794d50b84d40ba246db2ad1554b887f6b | 12,917 | null | r"""
Notes
1. goal is basically learn y = G(x_0, T, N).
G is always a recurrence `x_t+1 = f(x_t, t) mod N`.
Shouldn't be possible in general since f isn't determinable from a dataset...
I.e., multiple f can generate the same training dataset.
2. can't really use any info on structure of f...
"""
from __future... | {
"id": "ff279779-cd31-4483-854e-ff7abdfd267e",
"created_at": "2026-08-07 07:32:01.788600+00:00",
"db_md5": "ab239f853de695e0e19a63ef12458e76",
"submitter": "Arthur",
"github_login": "arthurfeeney",
"run_id": "c54d11f9-2b51-4128-84e1-7ede21f61443",
"tier": "easy",
"dataset_id": "e1",
"status": "succee... | {
"score": {
"mean_loss": 2.7418967485427856,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01833333307877183
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.3842382431030273,
"example_count": 100... |
ff28ad45-9f3e-4ad7-b60f-dba2295beb57 | easy | yashkant | 2026-08-13 00:18:09.889214+00:00 | succeeded | bead31070d046316b97e524980c156066652abcfb5708bf358c8e3fceecc7ee2 | 31,392 | null | """Round1702 one-projection-per-T Dykstra recursion on the D416 carrier.
Each parsed public-T cycle applies one shared learned smooth projection and
alternates the A/B set code and correction memory across cycles. The immutable
anchor is reinjected on every call; all active cycle outputs are averaged, and
terminal CE... | {
"id": "ff28ad45-9f3e-4ad7-b60f-dba2295beb57",
"created_at": "2026-08-13 00:18:09.889214+00:00",
"db_md5": "7343addd98e8e1ebc60baa83d62a705a",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "ea8e1732-e61e-4767-9deb-c6d7d848a749",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 4.121008336544037,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3033333495259285
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.660740375518799,
"example_count": 300,
... |
ff360787-17c7-43b5-8bf2-ad5a7d7186ff | easy | DDanlov | 2026-08-21 07:13:16.337462+00:00 | failed | 8a82f39f97234fd5867c1b9fd4433b9efc27be6e5b3d48566b4dae1aca301faa | 25,755 | null | from __future__ import annotations
import math
import time
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except Imp... | {
"id": "ff360787-17c7-43b5-8bf2-ad5a7d7186ff",
"created_at": "2026-08-21 07:13:16.337462+00:00",
"db_md5": "e6866b515d7b8f95e2194e140c819e69",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "e85d7d7a-b815-4ed9-990a-ff0cca5c218d",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
ff3cdcb1-a61f-4dcd-8bb0-0461501fe1ed | easy | DDanlov | 2026-08-08 10:59:24.880797+00:00 | succeeded | 8cc530aaa3cfd96917b619df350832d07b4d0a3e1aa4c6b3a74195c792f0eff0 | 15,561 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with eigenvalue ma... | {
"id": "ff3cdcb1-a61f-4dcd-8bb0-0461501fe1ed",
"created_at": "2026-08-08 10:59:24.880797+00:00",
"db_md5": "e8b5f06a26c5725fa3754b502958c7c2",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "cf54287a-b58f-4829-8030-88cc00dc1709",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 140.10529291243327,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.014999999664723873
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 134.20655822753906,
"example_count": 10... |
ff3e79ab-2b26-438e-be92-daef1d30f329 | easy | jordanrubin | 2026-08-06 19:59:06.254751+00:00 | succeeded | 09586f90ffa296e3bdbb25e6a0b059c7579344900d2f7d9eb12b0e902c4848f6 | 15,390 | null | """Bidirectional transformer, D=256 x 3 blocks, grokking-tuned AdamW.
Design notes:
- Inputs are padded to config.max_seq_len inside forward and logits sliced
back, so every training batch presents one static shape (compile/cudagraph
friendly, and uniform kernel shapes even in eager).
- Attention mask gets an iden... | {
"id": "ff3e79ab-2b26-438e-be92-daef1d30f329",
"created_at": "2026-08-06 19:59:06.254751+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "4ad42c39-5cf0-49fd-84b8-0cf82c1e981b",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.518913074679319,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.18249999467904368
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.490227658297266,
"example_count": 300,
... |
ff401388-6b0c-41ed-b5e3-79d055c38c5c | easy | newjordan | 2026-08-09 20:40:15.709934+00:00 | succeeded | c62384bf1996d7a46847b6c53519b16add45744b3efce861e92058af708b2309 | 94,093 | null | """Helix-Crawler — a recursive-weight-shared "crawler" ALU for modular squaring.
The task is r_{i+1} = r_i^2 mod n, chained T times: a *recurrence* over a single
operator. This build embodies the crawler / Frugendorff lineage (spark:~/sota_crawler)
mapped onto that structure. n never enters as a table row (unseen modu... | {
"id": "ff401388-6b0c-41ed-b5e3-79d055c38c5c",
"created_at": "2026-08-09 20:40:15.709934+00:00",
"db_md5": "814c97436c37f58c4124ff91f2e40326",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "5b074bd7-a51c-405c-b329-c747e6b349a7",
"tier": "easy",
"dataset_id": "e1",
"status": "succeed... | {
"score": {
"mean_loss": 0.0033483841689303517,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.003363973693922162,
"example_count": 100,
... |
ff477ca6-a0bd-4ba8-8b75-cdf1f4e48289 | easy | DDanlov | 2026-08-27 21:57:10.073695+00:00 | succeeded | 258443eddda663c3b73d8e81f6fe068aec433813cd6c23c96c29a57911739135 | 39,442 | null | import math
import time
from dataclasses import dataclass
from typing import Optional, Tuple, Dict, Any, List, Union
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from onelayerdeeper.api import Submission, OptimizerBundle, ModelSpec, OptimizerSpec
except ... | {
"id": "ff477ca6-a0bd-4ba8-8b75-cdf1f4e48289",
"created_at": "2026-08-27 21:57:10.073695+00:00",
"db_md5": "a10babc2aef379b68a74a70a00477458",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "f161c282-e079-4054-bdc5-6d462b4e6d83",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.741232991218567,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333334140479565
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.742947578430176,
"example_count": 600... |
ff4ade01-f480-4c25-88d9-4ecfded1a887 | easy | DDanlov | 2026-08-28 21:56:45.829400+00:00 | succeeded | 64342cbe504e0078f1a90b72d2c63d5eff6f1da56172e6beb86487f86c87994d | 15,065 | null | """
Trial 24: Model C: D=256, H=4, M=8, u=2 (T=16)
"""
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from dataclasses import dataclass
from typing import Optional, Tuple, Dict, Any, List, Union
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, T... | {
"id": "ff4ade01-f480-4c25-88d9-4ecfded1a887",
"created_at": "2026-08-28 21:56:45.829400+00:00",
"db_md5": "57eab46928dc166e9ee0d7114e1fba13",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "e18f4fa8-b904-43bc-adc8-7b36a228f8ca",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.403151887088872,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005000000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4370712500249323,
"example_count": 600... |
ff4b62c1-e115-4c9a-bc71-bd3ec317d025 | easy | viridale | 2026-08-04 22:22:23.374071+00:00 | succeeded | 18659897460356ffe2d076c83ae4f04db528c215926890b47b712e530d35996c | 7,862 | null | """Clean width-256 autonomous recurrent token model.
hypothesis: v174's 512-channel cell is capacity-heavy for a 17-token short-sequence
problem and completes too few updates; halving width should provide substantially
more learning through the same 16-step shared computation and improve transfer.
axis: arch
t... | {
"id": "ff4b62c1-e115-4c9a-bc71-bd3ec317d025",
"created_at": "2026-08-04 22:22:23.374071+00:00",
"db_md5": "585d8a196f922606c6f2df04040f5a85",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "59f5ef8b-083d-416e-8d96-17c94868793b",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.2602587491400365,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0075
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.27669243026191,
"example_count": 800,
"ex... |
ff5737bd-877d-4c06-8616-c28512bfad55 | easy | adwhit | 2026-08-13 13:35:20.591997+00:00 | succeeded | 2b22e7bef73f68b1801b8b74c416332afaac204182efe7caa9406879ca1393ee | 17,862 | null | """DeepThinker v4a: depth-recurrent transformer with manual CUDA-graph capture.
The evaluation harness is kernel-launch-bound on unrolled-loop models
(measured on H100/Easy: eager 206 steps/60s with the GPU >95% idle;
torch.compile default -> 120 steps net; reduce-overhead -> crash). v4a
captures the ENTIRE training u... | {
"id": "ff5737bd-877d-4c06-8616-c28512bfad55",
"created_at": "2026-08-13 13:35:20.591997+00:00",
"db_md5": "11593a427e174a98473e896293d113d8",
"submitter": "adwhit",
"github_login": "adwhit",
"run_id": "03c97ec7-03d3-4bb2-9324-d2dfe5b421e1",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
... | {
"score": {
"mean_loss": 3.117512583732605,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04833333380520344
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.5133771896362305,
"example_count": 100,... |
ff589720-0059-4648-8ea1-ccf55bd8d0d7 | easy | viridale | 2026-08-02 00:08:27.327625+00:00 | succeeded | a383352d982444f24a09aef69160eca315bf3e0a51d7d4a1498f1399f0f2c84d | 22,650 | null | """One-edit exact-program beam with cumulative hard evidence.
hypothesis: post-reset Hard rejects v076 after a full 284,883-step flat trace,
despite e3/m1 64/64 and repaired 24-bit factor/decode reach. The leading
remaining cause is absence of Hard's minimally modified recurrence from
the closed 1,536-map ... | {
"id": "ff589720-0059-4648-8ea1-ccf55bd8d0d7",
"created_at": "2026-08-02 00:08:27.327625+00:00",
"db_md5": "69fb7a0d2b8ad0e8a0cbeb0f45bc39e2",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "1223fa82-a577-4322-9099-f0008bbcf96b",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 0.003014546214626436,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.00301454615599774,
"example_count": 800,
"... |
ff64f232-e57e-4af2-bde8-a8ade105d0dd | easy | k-penchev | 2026-08-22 21:52:18.910610+00:00 | succeeded | fc8884a3cd9d42599b06dc444cf9f0c6618847ba7cebdd664e9d44b8c0560426 | 4,612 | null | """Two-block untied Transformer with digit places and input recall."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 128
... | {
"id": "ff64f232-e57e-4af2-bde8-a8ade105d0dd",
"created_at": "2026-08-22 21:52:18.910610+00:00",
"db_md5": "143aa40a3df036dabb261603512a50de",
"submitter": "Kaloyan Penchev",
"github_login": "k-penchev",
"run_id": "94fe05f8-9b58-4ce8-8274-8ff326528433",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 2.439367175102234,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.029999999329447746
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.69362211227417,
"example_count": 100,
... |
ff6c5789-b78d-4fa8-a94f-00e72d863d45 | easy | velocizapkar | 2026-08-11 21:16:36.329609+00:00 | failed | 9b3506a19b564842eb478940e972654f4603bcee23afd3ef817a66915c49f057 | 10,712 | null | """BDH associative reasoner for One Layer Deeper.
This is a generic learned recurrent architecture. It contains no parser,
task-specific arithmetic, data augmentation, persistent cross-example state,
participant-owned backward pass, or hidden training work.
"""
from __future__ import annotations
import math
import... | {
"id": "ff6c5789-b78d-4fa8-a94f-00e72d863d45",
"created_at": "2026-08-11 21:16:36.329609+00:00",
"db_md5": "4ba5503ffb8dc09ae49077a677fc69bb",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "5dad5c24-3fb9-4220-9007-fff33a437a26",
"tier": "easy",
"dataset_id": "e2",
"statu... | null
|
ff6e88c2-5653-4dd8-80c3-5de6bab2e151 | easy | KaustubhKumar05 | 2026-08-29 05:03:24.070476+00:00 | succeeded | 89285038b28e448631ad6034fb5a7007c1cf5db04998c8d5583268facf41ab02 | 12,682 | null | """op2_t1w -- stock cell_op2 + T=1 loss upweighting (the scored rung).
Forked from cell_op2. ORIGINAL DOC BELOW.
cell_op2 -- ONE shared squaring operator with internal depth, applied T times.
WHY THIS SHAPE (verified on the real data, not assumed):
answer(T=t) == x squared t times mod N 20000 / 20000 e... | {
"id": "ff6e88c2-5653-4dd8-80c3-5de6bab2e151",
"created_at": "2026-08-29 05:03:24.070476+00:00",
"db_md5": "de146efc427260a28a698b29f1f546c3",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "20a073b7-9dbc-43bb-90ed-258f114a4b1a",
"tier": "easy",
"dataset_id": "e5",
"status": "succee... | {
"score": {
"mean_loss": 2.4304395339074016,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333370586235
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3609126702017846,
"example_count": 60... |
ff740364-6533-4668-b3f5-35a7404855c5 | easy | yashkant | 2026-08-14 16:07:41.201178+00:00 | succeeded | 61c361cdd1c65d8ddb082d926e5112b412a3ee960b6352bf755951507f4e4f0b | 35,452 | null | """Round1871 BCH3 scale/LR factorial crossover.
Both arms preserve the exact Round1866 graph, loss, RNG order, AdamW, EMA
.995, and fully differentiable BCH operation. At LR 5.5e-3, the candidate
uses cubic scale 1.25 and the matched control uses scale 1.0.
"""
from __future__ import annotations
import math
import ... | {
"id": "ff740364-6533-4668-b3f5-35a7404855c5",
"created_at": "2026-08-14 16:07:41.201178+00:00",
"db_md5": "9b590b315148193ac3cf3a4c126c949d",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "d470de19-5aca-4376-9b54-4c6307d7532e",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.125759609523584,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016250000158324836
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.12568181705515,
"example_count": 800,
... |
ff797ab5-5ea6-4af2-9408-e0343f04cdf2 | easy | newjordan | 2026-08-22 18:14:36.730137+00:00 | failed | d8781c04ab9139aaa048d3c15cae106b07c5446e49e952cb662107f2be47e588 | 18,998 | null | """Neural Transition Cell: learned local computation with tied recurrence.
The model parses the public decimal fields only into fixed-width one-hot
registers. A small convolutional gated cell, shared across digit positions
and refinement steps, learns one state transition from evaluator labels. The
same learned tran... | {
"id": "ff797ab5-5ea6-4af2-9408-e0343f04cdf2",
"created_at": "2026-08-22 18:14:36.730137+00:00",
"db_md5": "7d4066b8203055a2d20eed268344c6ef",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "a0bfd646-0b29-4755-89aa-660e855e2b19",
"tier": "easy",
"dataset_id": "e6",
"status": "failed"... | null
|
ff79a8ad-f627-4e02-991d-54fdf398e152 | easy | alirezashirvani-jr | 2026-08-29 12:08:46.156407+00:00 | succeeded | 342200170e15fa3e8465d13973f52c114f9d1341dbc9f9cb2a61402ea1949133 | 32,928 | null | from __future__ import annotations
import math
import time
from contextlib import nullcontext
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
MODEL_... | {
"id": "ff79a8ad-f627-4e02-991d-54fdf398e152",
"created_at": "2026-08-29 12:08:46.156407+00:00",
"db_md5": "865a7d4d3b356344be80118eb172f236",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "2f4e9fd1-7159-49f9-a880-70933a927619",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 2.9123879387840494,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03416666659216086
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.948183298110962,
"example_count": 600,... |
ff7c2c8e-d837-4332-890b-f6bfd9bdd51f | easy | chuk-uzowihe | 2026-08-09 19:24:07.504361+00:00 | succeeded | 109cf7187730e4702d8a3efe0eaa463e60c9856138c3a4f0b50d3f1b0eb70e9f | 36,588 | null | """One Layer Deeper submission: depth-routed MoEUT over a 2D digit lattice.
v9: the recurrent block is a single MoEUT layer (refs/moeut) — SwitchHead
attention + sigma-MoE feedforward — whose routers are conditioned on the
loop depth (Fourier features of the iteration index) so one shared bank can
express phase-depend... | {
"id": "ff7c2c8e-d837-4332-890b-f6bfd9bdd51f",
"created_at": "2026-08-09 19:24:07.504361+00:00",
"db_md5": "34e342a6328c7302f3041195cfaf52c8",
"submitter": "Chuk Uzowihe",
"github_login": "chuk-uzowihe",
"run_id": "d09f5937-580e-412d-b438-ab8073f0f615",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 5.4781174659729,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07666666433215141
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.775765895843506,
"example_count": 100,
... |
ff7e9336-19a6-4aa9-9761-8d397ca9f25f | easy | sapient-sapiens | 2026-08-22 18:23:42.183727+00:00 | succeeded | 3d9bd901ad01bf30cfb61f7bce2738d56efa4e4f279b0fb9d49966cb8f69d488 | 19,798 | null | """Batch-16 eight-layer T-private Q+V TRM with HybridMuon and SAM for E5.
Research hypothesis: batch 16 improves H100 utilization and fixed-clock learning
relative to the inherited batch 8 recipe.
Primary experimental variable: training batch size, 8 versus 16, with model,
optimizer, LR, SAM, and loss held fixed. The... | {
"id": "ff7e9336-19a6-4aa9-9761-8d397ca9f25f",
"created_at": "2026-08-22 18:23:42.183727+00:00",
"db_md5": "bfbac46ffdc94d21f7e01cb419f0c35f",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "9899682c-d238-4000-aefe-9c4c040e9486",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.2777814740213564,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002500000024835269
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2820006929705494,
"example_count": 60... |
ff86620e-75b9-4623-b8bd-ca2718716731 | easy | KaustubhKumar05 | 2026-08-25 16:19:30.866093+00:00 | succeeded | cede8f877bf90dc9d38af3879dc3680a73b6693631c3476085830cfc30a955d0 | 16,027 | null | """final_v1: culmination candidate. v8 (N-FiLM) + Muon momentum 0.5 + tail-EMA.
A small MLP embeds the modulus digits (mean-pooled over the N span) and
emits per-channel scale/shift applied to the cell input and a bias on the
update gate, every iteration. Hypernetwork-lite: the executor is modulated
by N rather than h... | {
"id": "ff86620e-75b9-4623-b8bd-ca2718716731",
"created_at": "2026-08-25 16:19:30.866093+00:00",
"db_md5": "4c081115900db9bcea126ae905aeee44",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "46ed0c8c-f6cf-4f0a-a934-dbcf20ebc393",
"tier": "easy",
"dataset_id": "e6",
"status": "succee... | {
"score": {
"mean_loss": 5.590756416320801,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.18828829377889633
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.338226318359375,
"example_count": 60,
... |
ff94b36a-1372-400b-a382-20170b695218 | easy | dlyr3 | 2026-08-23 13:24:16.293698+00:00 | succeeded | cd4f77e3fccf3fc2c0ded48448ff4fc22f691904d2850a8f80d3eb9eebdad983 | 7,111 | null | """The M-layer of arXiv:2008.03936 applied to the repeated-squaring task.
M = B + T(U x) is built from the pooled prompt, exp(M) is taken with
torch.matrix_exp, and V + S exp(M) is read back out. The paper's activity
regularisation, lambda * ||exp(M)||_F^2, is returned as the auxiliary value and
added to the loss... | {
"id": "ff94b36a-1372-400b-a382-20170b695218",
"created_at": "2026-08-23 13:24:16.293698+00:00",
"db_md5": "43845100059e8bfa5a2c5daf8a93d474",
"submitter": "dlyr3",
"github_login": "dlyr3",
"run_id": "952426fc-a6ac-411b-b6dc-85627bde58e4",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 6.196731564053183,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009166666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.179819333713685,
"example_count": 600,... |
ff9e29e1-0cbf-40f1-9662-1419d47e8bd4 | easy | gauravmishra | 2026-08-10 08:26:10.622506+00:00 | succeeded | dc62c99929b34f6e345c8f3b1be855dd49562916df42971ce85ccaa01a868fad | 8,689 | null | """E43: historical K13-S385 with E09's Muon/AdamW partition.
Architecture parent SHA-256:
42dc862219390b8300e98260725c6c2621b27aae32f750f81902dea4bf95889d
Optimizer parent SHA-256:
350d876cd225f4280d9ab80241d4801218d0cb1f520cd062dea33a0a18df0719
"""
from __future__ import annotations
import torch
import torch.nn.fun... | {
"id": "ff9e29e1-0cbf-40f1-9662-1419d47e8bd4",
"created_at": "2026-08-10 08:26:10.622506+00:00",
"db_md5": "7237242bd86b3a08012ed9b9438422bf",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "f27e14b5-5119-45e5-b61e-044800af9b6b",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 8.505018419247538,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666666666668
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.237204012849405,
"example_count": 600,... |
ffa43c9b-df81-4496-927c-6e8ccc55961c | easy | aupadhyay | 2026-08-04 18:05:12.667846+00:00 | succeeded | 6cac905925ebbad461456e055795a6df65e0306ac1a9a64240d659334193d79d | 11,487 | null | """Phase 2: weight-tied looped architecture.
x^(2^T) mod N is one simple map applied T times. Instead of hoping a flat
transformer discovers composition (Phase 1 showed it only ever reaches the
composition ceiling), the loop is architectural:
1. Encode x into a value-space latent s0 via a digit binding basis
(s = ... | {
"id": "ffa43c9b-df81-4496-927c-6e8ccc55961c",
"created_at": "2026-08-04 18:05:12.667846+00:00",
"db_md5": "5a6a5e50fb2ed8774ea1fe33851197de",
"submitter": "Abhi Upadhyay",
"github_login": "aupadhyay",
"run_id": "b3c221e6-419e-438e-8775-2eb3b6820682",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 6.1511805057525635,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3933333307504654
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.803577423095703,
"example_count": 100,
... |
ffa4ce82-fa62-4fda-b398-ce1542b2fc3a | easy | DDanlov | 2026-08-26 01:12:06.236555+00:00 | succeeded | 42253786239236b46888154e8c59d283f34bb085eb13ff676b4aa295f989669c | 34,320 | null | import math
import time
from typing import Optional, Tuple, Any, Dict, List, Union
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except ImportError:
tr... | {
"id": "ffa4ce82-fa62-4fda-b398-ce1542b2fc3a",
"created_at": "2026-08-26 01:12:06.236555+00:00",
"db_md5": "5b8b5c04c45788b6a1f7876d30e5ec98",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "b2c72567-c7ca-4a7c-b5ac-5887a212cc31",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.3003181518219806,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010000000024835268
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.301273654929191,
"example_count": 600... |
ffa64c08-e66c-4f65-a9c1-9a8a9abda9ec | easy | Koenneker | 2026-08-05 09:21:40.981119+00:00 | succeeded | b53c966c65daa0fefab669ea8ac3fccae6aa9950425f85246c98b3e51facdc6c | 5,067 | null | r"""Weight-tied 2-loop LSTM: recall, deep supervision, label smoothing 0.2.
Winner of a ten-round local search over loop/TRM/LSTM hybrids. On E1 across
10 seeds at the full 60s budget it scores 7.52% +- 0.19 against the plain
LSTM's 6.77% +- 0.24 -- +0.75pp, 2.45 sigma, winning 9 of 10 seeds when
paired (t = 2.38, df ... | {
"id": "ffa64c08-e66c-4f65-a9c1-9a8a9abda9ec",
"created_at": "2026-08-05 09:21:40.981119+00:00",
"db_md5": "8d5539d7fb851b33ffbb640193f69cb2",
"submitter": "Ferdinand Könneker",
"github_login": "Koenneker",
"run_id": "0963ec9d-e6fb-480c-94bc-8c065d51e70f",
"tier": "easy",
"dataset_id": "e3",
"status"... | {
"score": {
"mean_loss": 3.225741605147216,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00874999980442226
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.280913008628495,
"example_count": 800,
... |
ffa76390-b6c6-4e45-9fbf-13fd44ff4336 | easy | andreamiele | 2026-08-06 09:04:29.506874+00:00 | succeeded | 609557feb40b6834af209ce9fccee790d6632feb193823e093d3b3f75558e93c | 22,804 | null | """Loop model with recall, composition consistency, and deep supervision.
Motivated by what the easy-to-hard / latent-reasoning literature converges on
and which earlier attempts here lacked:
* **Recall (input injection).** Deep Thinking (Schwarzschild et al. 2021;
Bansal et al. 2022) found iterating past the trai... | {
"id": "ffa76390-b6c6-4e45-9fbf-13fd44ff4336",
"created_at": "2026-08-06 09:04:29.506874+00:00",
"db_md5": "83f101d3d1703d13fc213532cf5decee",
"submitter": "Andrea Miele",
"github_login": "andreamiele",
"run_id": "8ccd93b8-d15a-4a46-ac3b-a4f59f7e3f90",
"tier": "easy",
"dataset_id": "e1",
"status": "s... | {
"score": {
"mean_loss": 2.3554973186091797,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.5000000087420146
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.79864275544437,
"example_count": 100,
... |
ffa7c7c0-d2d4-4dfb-a1ed-81fc23fbc77c | easy | KaustubhKumar05 | 2026-08-29 04:46:23.914974+00:00 | succeeded | c8d3923eb75c1eecba0eb8a07243f8e636cf211893f6fa66c68ea72498b75bf3 | 14,341 | null | """op2_t1w -- stock cell_op2 + T=1 loss upweighting (the scored rung).
Forked from cell_op2. ORIGINAL DOC BELOW.
cell_op2 -- ONE shared squaring operator with internal depth, applied T times.
WHY THIS SHAPE (verified on the real data, not assumed):
answer(T=t) == x squared t times mod N 20000 / 20000 e... | {
"id": "ffa7c7c0-d2d4-4dfb-a1ed-81fc23fbc77c",
"created_at": "2026-08-29 04:46:23.914974+00:00",
"db_md5": "f196f9e3e3d6de9df94e6a10679ea928",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "82770db6-d9c2-4cd4-b342-661dfdeae913",
"tier": "easy",
"dataset_id": "e5",
"status": "succee... | {
"score": {
"mean_loss": 3.0641888214815305,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333345750968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.681827371430504,
"example_count": 600... |
ffa8e59e-2e1b-4c35-9149-2b54782aaaa9 | easy | LauraGomezjurado | 2026-08-30 21:38:49.152422+00:00 | succeeded | 128c79dc238874bfb48c178b74c83587db56aada924b2be61310e64a31cf8c97 | 9,045 | null | """Looped depth-recurrent Transformer for repeated modular squaring.
Design notes
------------
The evaluator reads answers off the *last* ``target_len`` prompt positions, so
answers are right-aligned with the least-significant digit at the final valid
position. Digit significance is therefore a fixed offset from the e... | {
"id": "ffa8e59e-2e1b-4c35-9149-2b54782aaaa9",
"created_at": "2026-08-30 21:38:49.152422+00:00",
"db_md5": "02b848a8dc5b6e14c271c329e6e89f86",
"submitter": "LauraGomezjurado",
"github_login": "LauraGomezjurado",
"run_id": "72414c04-31e2-40a5-8b04-dd85264debfe",
"tier": "easy",
"dataset_id": "e7",
"st... | {
"score": {
"mean_loss": 1.9244479395094372,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.029010695997964252
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.921130657196045,
"example_count": 85,... |
ffa91639-d852-4c98-af74-f2da1937fe47 | easy | lzy54 | 2026-08-27 05:30:23.038843+00:00 | succeeded | cff143bebd9be0f8c903f1768d214d90c30d347351c5be2e8e6bbd3ee2c871fe | 6,885 | null | """B2 plus weak supervision at the sampled endpoint's penultimate state."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_... | {
"id": "ffa91639-d852-4c98-af74-f2da1937fe47",
"created_at": "2026-08-27 05:30:23.038843+00:00",
"db_md5": "46654244fbe3f2a394a320430400fefb",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "69557c54-89f3-41f6-a1fb-658f1869f34f",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.0553749799728394,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006756756920367479
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.097644329071045,
"example_count": 60,... |
ffab1276-260c-4b97-abae-a5f8d5c9d4f4 | easy | erdavis0 | 2026-08-23 09:27:04.688301+00:00 | succeeded | 5931d5cb503d697a8c8afbc967880b517d5e45ad3169886b085f7c747014d8fc | 21,320 | null | """Field-aligned convolutional sparse-coding transducer.
Public delimiters route opaque digit categories into separate right-aligned
N, X, and T tapes. Four parameter-disjoint sparse tapes perform five tied
learned analysis/synthesis residual updates. Interaction and carry receive
only X and T; N first enters the mo... | {
"id": "ffab1276-260c-4b97-abae-a5f8d5c9d4f4",
"created_at": "2026-08-23 09:27:04.688301+00:00",
"db_md5": "8afc3b0fe55be1106a80817c959b34ea",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "2d3e6492-eb95-4963-892a-567cbd2ce88e",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.31054458144503,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3690577675943416,
"example_count": 600,... |
ffab1cef-9340-4fb9-bab3-0a16b373edce | easy | DDanlov | 2026-08-13 01:30:19.821422+00:00 | failed | 24f2a2d3ec6651f373d2c49574c45da218b894c850195f115f64285d4d74ca69 | 27,540 | null | from __future__ import annotations
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
try:
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
except ... | {
"id": "ffab1cef-9340-4fb9-bab3-0a16b373edce",
"created_at": "2026-08-13 01:30:19.821422+00:00",
"db_md5": "9f210dcd062ee58a62ab637ff816509b",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "4a97a9e3-da36-4574-a4c7-2e13bfde79f8",
"tier": "easy",
"dataset_id": "e1",
"status": "failed",
... | null
|
ffabe54b-c4af-46d9-8dd5-da36158ae209 | easy | alirezashirvani-jr | 2026-08-29 05:40:55.485470+00:00 | succeeded | 52d3e8ea3655dd1b49603334ba1b505f43685cd8ec86c58dfe84c472ef64c8fc | 39,524 | null | """Adaptive learned discrete recurrent fabric for One Layer Deeper.
A gradient-trained, operation-free recurrent computational machine. The
public prompt grammar only routes literal digit tokens into canonical,
tail-aligned categorical registers. A reusable learned transition circuit
is applied serially, the public ... | {
"id": "ffabe54b-c4af-46d9-8dd5-da36158ae209",
"created_at": "2026-08-29 05:40:55.485470+00:00",
"db_md5": "038694af6bb07b235d91146bab29234b",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "ab1c8019-ecd8-4d00-b489-db1033631efd",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 2.9581757283157057,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04041666726271312
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.9633677005767822,
"example_count": 600... |
ffae0614-eabf-4a92-bf7f-b1daac05ff44 | easy | yashkant | 2026-08-16 00:18:34.006982+00:00 | succeeded | 1bc8df6a5e907b3a48e004b2a4c244e7b5b6835c95144812f8aba0bf9e5a6399 | 48,086 | null | """Round1805 logarithmic dyadic carry closure pair.
The exact Round1792 AdamW5 carrier is retained. Learned nonlinear carry
transfer curves are composed across right-aligned N/X token positions by a
binary-lifting closure. Parsed public T allocates only ``bit_length(T)``
ordered composition stages, with offsets cycl... | {
"id": "ffae0614-eabf-4a92-bf7f-b1daac05ff44",
"created_at": "2026-08-16 00:18:34.006982+00:00",
"db_md5": "378d71bd04ef53dc577c8314795f22fa",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "065ffde4-fe15-4b5d-bdd5-cfbe57044d02",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 2.2828125953674316,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.038333335891366005
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4989333152770996,
"example_count": 30... |
ffb0ecb6-e170-480f-b562-4d89b6de3158 | easy | alirezashirvani-jr | 2026-08-30 22:35:30.371613+00:00 | succeeded | 5e04560fff6d76c6350ed134dce34bbeb5ddf03992d210d9cdc6dda6ff09212c | 17,544 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PERIOD_MINIMUM = 11
PERIOD_MAXIMUM = 100
E... | {
"id": "ffb0ecb6-e170-480f-b562-4d89b6de3158",
"created_at": "2026-08-30 22:35:30.371613+00:00",
"db_md5": "cfec548a5587d3a2942b9c10220452bc",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "b51a6b80-efe8-4d25-a458-1b933f8d930e",
"tier": "easy",
"dataset_id": "e3",
... | {
"score": {
"mean_loss": 2.5511703491210938,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.19437500089406967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.0498576164245605,
"example_count": 800... |
ffb45397-96dd-4936-9c8e-bd85e1df69dd | easy | sapient-sapiens | 2026-08-26 11:28:50.845075+00:00 | failed | 64807cfdb3d8ec483f9ed8ec5fe983e7f9e995164ae8e13d045c643f268b9342 | 21,240 | null | """Standalone nested inner/outer recurrence candidate.
Representation is categorical only: ``E_token + E_group + E_leftrel``. No
field is decoded into a number, no field value sets depth, halting, or any
mixing weight, and the only supervised target is the final answer.
Structure
---------
One prompt encoder produce... | {
"id": "ffb45397-96dd-4936-9c8e-bd85e1df69dd",
"created_at": "2026-08-26 11:28:50.845075+00:00",
"db_md5": "eeca47d0f0d177f55103ed7cceef8502",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "a73d6e02-11f4-4601-8b77-04b03474ddbb",
"tier": "easy",
"dataset_id": "e5",
"status": "fa... | null
|
ffb57a61-b7fa-43a6-9c23-7e95b8d4db3d | easy | sapient-sapiens | 2026-08-19 19:35:17.323461+00:00 | succeeded | daa4fc1f1821b95b049e62d1602f10121592d66b9efbbb3e78e141bca4403b2b | 33,252 | null | """Depth-quantized recurrence with a mixture-over-depth objective.
The recurrent state is a bank of right-aligned digit slots. Every operator
application is followed by a soft quantization back onto the token simplex, and
the digit logits produced by that quantization are the answer logits at that
depth. There is no s... | {
"id": "ffb57a61-b7fa-43a6-9c23-7e95b8d4db3d",
"created_at": "2026-08-19 19:35:17.323461+00:00",
"db_md5": "2553f964fd1803486430c2690edf5c49",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "c51251ac-fa68-492b-a3dd-c9d9674cda79",
"tier": "easy",
"dataset_id": "e10",
"status": "s... | {
"score": {
"mean_loss": 3.5430408869034204,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0646262639061068
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2741074562072754,
"example_count": 125,... |
ffb96f1e-f9aa-43ed-b175-fe854b063925 | easy | khushidahi | 2026-08-06 04:14:47.481008+00:00 | succeeded | c33b2407289deccedecea96b05d9280013066fcfda217ae29679d568544d3be2 | 14,408 | null | """R6 structured decimal-workspace model for One Layer Deeper.
The model does not implement multiplication, modular reduction, or the public
recurrence. It uses the public tokenizer structure to build learned, right-
aligned decimal tapes for N, X, and T, then applies a shared neural transition
to a mutable work tape.... | {
"id": "ffb96f1e-f9aa-43ed-b175-fe854b063925",
"created_at": "2026-08-06 04:14:47.481008+00:00",
"db_md5": "bbbea86d8cf2206f3b410bebcb973d2e",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "910c3954-aeb9-4f6e-82c5-3275d511e0ad",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.323410153388977,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007083333563059568
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3173654079437256,
"example_count": 600... |
ffbe54b3-a204-4b71-85da-b1cdf435936d | easy | viridale | 2026-08-25 22:33:09.993457+00:00 | succeeded | c895501201f9a96557ac340c439d4c35011d7014f1c2b27f0bc5b3650c64bb43 | 28,716 | null | """Constraint-posterior bilinear Transformer program.
hypothesis: v912's seed-0 selector fits all 24 seen identities softly but its
final argmax transfers to only 17/24 OOD, proving the learned selector remains a
diffuse interpolation. Add a small entropy cost: it has zero gradient at the initial
uniform d... | {
"id": "ffbe54b3-a204-4b71-85da-b1cdf435936d",
"created_at": "2026-08-25 22:33:09.993457+00:00",
"db_md5": "2b3fa97360a121cb4f3674d8a7e5ce88",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "6fee0cb3-ca22-4daf-8bec-aab33bbf51f8",
"tier": "easy",
"dataset_id": "e4",
"status": "succe... | {
"score": {
"mean_loss": 0.0,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0,
"example_count": 1200,
"exact_accuracy": 1.0,
... |
ffcea55c-be4f-4ba3-a499-d1bef9a985e5 | easy | alirezashirvani-jr | 2026-08-31 00:53:14.794282+00:00 | succeeded | d30c15c4b819074a695cfe7f9ab43fd9530cdf8d2f612f55533983ffcf4e7dfb | 22,445 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PERIOD_MINIMUM = 2
PERIOD_MAXIMUM = 96
EXP... | {
"id": "ffcea55c-be4f-4ba3-a499-d1bef9a985e5",
"created_at": "2026-08-31 00:53:14.794282+00:00",
"db_md5": "a726edbe0af9de151bd033a13eebcee4",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "288f7174-262c-498c-b3d3-029fa6b20708",
"tier": "easy",
"dataset_id": "e10",
... | {
"score": {
"mean_loss": 0.0,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0,
"example_count": 125,
"exact_accuracy": 1.0,
"... |
ffd568a5-679b-4050-9ce7-f82c391a6f5b | easy | sapient-sapiens | 2026-09-01 02:46:56.475444+00:00 | succeeded | 9dfdc664abfaceb20c68b77ca4788a23c76a28fd282f3636abe72e28893b089e | 24,042 | null | """E5: a tested one-shot D6 square tied only across task time T.
The physical Transformer block is one complete learned square operator. It is
called once per square boundary and receives no raw N token: categorical N
only edits the shared weights and character table. Eight boundary calls are
fully differentiable; a... | {
"id": "ffd568a5-679b-4050-9ce7-f82c391a6f5b",
"created_at": "2026-09-01 02:46:56.475444+00:00",
"db_md5": "3f603c7cca05905638819b2583373ef9",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "0b288989-3edc-48e4-9d13-ceca359a02d5",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.159604737254063,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.161774514234654,
"example_count": 600,... |
ffdaefb6-d745-4170-9e22-c47a951b9ae1 | easy | DDanlov | 2026-08-20 22:48:06.232051+00:00 | failed | 55685715c4d1f18ff6c003d31c9c85037f58bac68498eeed2af9fccbf39e2129 | 21,253 | null | from __future__ import annotations
import math
import time
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except Imp... | {
"id": "ffdaefb6-d745-4170-9e22-c47a951b9ae1",
"created_at": "2026-08-20 22:48:06.232051+00:00",
"db_md5": "160baa3e28b1a2142fcd1c254a81d9d9",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "9845447a-fee2-4677-888a-ef57dde9306e",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
ffdeb091-b5ba-426f-887c-0caa329aebc4 | easy | rteehas | 2026-08-17 21:22:55.721933+00:00 | failed | db41662ecc75a2941dbaaf5c4b2e9eabb84fe419d56590530f9f6e92b84ee0d4 | 39,089 | null | """Anonymous marginal-program submission.
Mechanism
---------
The model is a differentiable probabilistic program. Its forward pass
computes, exactly, the answer-token predictions of a small structured latent
hypothesis space:
(prompt segmentation = which token types head the three fields)
x (field-role permutat... | {
"id": "ffdeb091-b5ba-426f-887c-0caa329aebc4",
"created_at": "2026-08-17 21:22:55.721933+00:00",
"db_md5": "539b855f58d22728daf24af83e3ef595",
"submitter": "Ryan Teehan",
"github_login": "rteehas",
"run_id": "1243d29d-8bb3-46e2-b7b0-3c69db952f4f",
"tier": "easy",
"dataset_id": "e1",
"status": "failed... | null
|
ffe1528f-6c9b-4eb4-966a-fd74ac4380be | easy | DDanlov | 2026-08-24 03:49:03.827386+00:00 | succeeded | 64c35cf9fa93a8b9cce0d450f3686b0678001292a949881117b1bd44b41b1f97 | 38,026 | null | """
Unified Hybrid Masked Deliberation Model (40.4M Parameters) - Pure Skipless, No Cross-Attention
Architecture:
- Scale: dim=1536, num_heads=24, d_ff=6144 (Exactly 40,437,888 parameters)
- Hybrid Prefix-Bidirectional + Causal Target Attention Mask
- Token Layout: [Prompt (L_p), 4 Pause Registers, Target Tokens (L_tgt... | {
"id": "ffe1528f-6c9b-4eb4-966a-fd74ac4380be",
"created_at": "2026-08-24 03:49:03.827386+00:00",
"db_md5": "84682b777d6ce302efbc9ed6ff6c3e09",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "d9b4f776-ac55-44f3-a82f-3145d8a061b2",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.3484784133126713,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0025
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3884983212423965,
"example_count": 600,
"... |
ffe31640-1c64-44e3-9431-e1a336e02516 | easy | yashkant | 2026-08-09 03:14:13.704580+00:00 | succeeded | 94cfd377427795b070d0eeb159e0d830759f0293f7e0e38a31a9f048d53446bf | 17,720 | null | """Round 143 with PyTorch add-zero attention sinks and raw EMA."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
from torch import... | {
"id": "ffe31640-1c64-44e3-9431-e1a336e02516",
"created_at": "2026-08-09 03:14:13.704580+00:00",
"db_md5": "5ccdaffccb47c973a0c700402d4812bc",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "796f82c8-2502-424e-9416-cd986b6ad7cb",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 3.526111937966939,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013750000009313226
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.522892760588534,
"example_count": 800,... |
ffe48ebf-4ac5-4f48-9bc0-91a759383a31 | easy | Mantissagithub | 2026-08-19 10:32:15.329588+00:00 | failed | 798ef29467515301da1a5588738159f5c0ca548ec61e0cce49fb19a82332bc42 | 29,316 | null | """attempt_035 + a contentless T field (depth enters only as iteration count).
ADDENDUM (attempt_042_gated): attempt_040_residual with the cross-iteration
carry scaled by a single learned scalar gate, sigmoid(carry_logit), init 0.5.
This is one scalar parameter (+1 element against the model-state ceiling) and
one mult... | {
"id": "ffe48ebf-4ac5-4f48-9bc0-91a759383a31",
"created_at": "2026-08-19 10:32:15.329588+00:00",
"db_md5": "bc0dc13b4322dd715063d1ecf7e1e81c",
"submitter": "Pradheep P",
"github_login": "Mantissagithub",
"run_id": "0c7a5891-50b3-422b-83de-88e519a460d5",
"tier": "easy",
"dataset_id": "e2",
"status": "... | null
|
fff2a4aa-9d1f-4db9-919e-39a9b866508f | easy | khushidahi | 2026-08-07 23:10:13.284051+00:00 | succeeded | 3f219c069c95ec53aa9feb35c0f54f4521ca6a3576a1e897b448e610ca5fd483 | 8,513 | null | """H0 conservative R3 Hard anchor.
This is the existing R3 architecture and AdamW recipe evaluated after exactly
3,000 optimizer steps. It provides the matched control for the boundary-aware
sequence-loss experiment.
"""
from __future__ import annotations
import random
import torch
import torch.nn.functional as F
f... | {
"id": "fff2a4aa-9d1f-4db9-919e-39a9b866508f",
"created_at": "2026-08-07 23:10:13.284051+00:00",
"db_md5": "e3494a095f26ba257f1ad6805cb81046",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "66937397-9287-422a-a853-ea46503dddc2",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 2.23199725151062,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004791666986420751
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2316629886627197,
"example_count": 300,... |
fff459ca-ca52-4b90-b4ce-05827cfda34d | easy | velocizapkar | 2026-08-16 07:36:35.288760+00:00 | succeeded | eacf048f1534dadee202ab543adcb41a2761caf702550e032e534a5d3eb8e079 | 20,154 | null | """Expected-whole-answer objective control for the field-relative workspace.
The model, initialization, field-product route, four tied decoder visits,
AdamW optimizer, learning-rate schedule, and batch sizes exactly match the
donor. For the first 33 percent of wall-clock training it also uses the
donor's sequence-bal... | {
"id": "fff459ca-ca52-4b90-b4ce-05827cfda34d",
"created_at": "2026-08-16 07:36:35.288760+00:00",
"db_md5": "28dfa30220795304bfba3a6e8b8b9786",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "87549a3c-24b7-4947-b76b-d13effa8b918",
"tier": "easy",
"dataset_id": "e6",
"statu... | {
"score": {
"mean_loss": 3.8269017934799194,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.086936941370368
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.396672010421753,
"example_count": 60,
... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.