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 |
|---|---|---|---|---|---|---|---|---|---|---|
463b1215-fd80-4564-b779-e67007792884 | easy | KaustubhKumar05 | 2026-08-27 11:44:16.461004+00:00 | succeeded | bf18b8bbad149b98f31317bc03168609b37b6275f551314ac0f91d65977f6226 | 10,748 | null | """abacus_reg — an explicit mutable numerical workspace, not a classifier.
§71. Every one of the twelve candidates measured on the r=0.938 tradeoff curve is a
CLASSIFIER OVER ANSWER DIGITS: read [N, x, T], mix it in a residual stream, emit logits.
None has a REGISTER -- no designated place where an intermediate value ... | {
"id": "463b1215-fd80-4564-b779-e67007792884",
"created_at": "2026-08-27 11:44:16.461004+00:00",
"db_md5": "839e71ec055e8f4220f8ee71408bebec",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "2bd1d9cf-3929-4eb0-b651-d513f7bd13bb",
"tier": "easy",
"dataset_id": "e3",
"status": "succee... | {
"score": {
"mean_loss": 3.7937675649959655,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013125000000000001
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.8147805753153414,
"example_count": 80... |
463c6665-38db-4808-a233-425be75a4614 | easy | shreyash-chonkie | 2026-09-01 00:25:29.524053+00:00 | succeeded | 767f3fcdcb086616a5eab4624a024f2ea6ddd95cab2ef661a1cdda7acf9129ef | 9,368 | null | """Through-time recurrent Transformer with checkpoint supervision."""
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,
assert... | {
"id": "463c6665-38db-4808-a233-425be75a4614",
"created_at": "2026-09-01 00:25:29.524053+00:00",
"db_md5": "9aee79e035750e6a588e1474b2a0938f",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "4c499130-6bb5-4e27-b49d-bebe1836cbbd",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 3.1812695264816284,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06999999657273293
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.379354953765869,
"example_count": 100,... |
463d8500-18b0-4b3f-b438-e32cf5cb4042 | easy | jordanrubin | 2026-08-07 06:13:09.084116+00:00 | succeeded | 6f17eb27b2bb1fbb5891f0392535694f326c447b00cd9f37136d01d78d2e30ec | 16,908 | null | """Autonomous digit-state recurrent Transformer for repeated modular squaring.
The model is deliberately organized around one learned transition:
p_0 = right_aligned_decimal_digits(x)
p_{k+1} = F_theta(p_k, decimal_digits(N))
The same two-block Transformer cell is applied exactly T times. T is used only
as ... | {
"id": "463d8500-18b0-4b3f-b438-e32cf5cb4042",
"created_at": "2026-08-07 06:13:09.084116+00:00",
"db_md5": "378594dd4459a9c7efa73ec8d9805eae",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "b901ac4a-254b-45ba-9505-fb3a79ac6183",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.317314170567644,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666433836022
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.315306098070915,
"example_count": 600,... |
463e0274-3e40-48c1-8573-8c57830db485 | easy | yashkant | 2026-08-14 15:41:05.603735+00:00 | succeeded | 825ba64e84b95097345ef8a807f119c2928876e4a11f4f77feefefcee0f49947 | 36,351 | null | """Round1819 learned ordered CP algebra inside Dykstra correction memory.
The exact Round1792 AdamW5 W8/D2 carrier is retained. At both existing
correction seams, one shared rank-four CP bilinear product composes the prior
correction with the new projection defect. Distinct learned left and right
factor maps make th... | {
"id": "463e0274-3e40-48c1-8573-8c57830db485",
"created_at": "2026-08-14 15:41:05.603735+00:00",
"db_md5": "6c53d27c898dcabb7df68af3fe9dec74",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "289f62ca-5dbd-4e29-99fe-b530484fcc6c",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 2.546064615249634,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04541666992008686
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.932450532913208,
"example_count": 300,
... |
46450fbe-b6f9-4200-9cce-7023cb1a7ba5 | easy | Dandandan | 2026-08-05 17:48:33.984980+00:00 | succeeded | 27df17b06b1a6ca3aeeb39a8f57bba0f8e263bd27485b49d90a6a622cfd706a5 | 36,821 | null | """End-to-end recursive function learner for One Layer Deeper.
A population of randomly initialized polynomial register networks is reused
exactly once per requested outer step. The evaluator trains every coefficient
and the evidence selector through its normal forward/backward/optimizer loop.
There is no recurrence-... | {
"id": "46450fbe-b6f9-4200-9cce-7023cb1a7ba5",
"created_at": "2026-08-05 17:48:33.984980+00:00",
"db_md5": "6113547b058dc12f42c9c04d3054cd5a",
"submitter": "Daniël Heres",
"github_login": "Dandandan",
"run_id": "18c43d88-d129-4e8a-9d44-c293257c52b9",
"tier": "easy",
"dataset_id": "e5",
"status": "suc... | {
"score": {
"mean_loss": 1.1348570754328067e-05,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.1365176222235958e-05,
"example_count": 600,
... |
464f0571-3e99-48dc-829c-0f4c9ad03d0d | easy | msnow0 | 2026-08-08 11:17:31.621844+00:00 | succeeded | 3625058685c685c0331c6ad85373dbaac929741f881a0f31e131ee5102f1b459 | 16,877 | null | """One Layer Deeper submission — mechanism 1 (see BRIEF.md §6, research/RULES.md).
Weight-tied recurrent operator over a discrete token tape. One outer iteration per
squaring step (loop count conditioned on the prompt's T, which is explicitly allowed);
between outer steps the latent is snapped back onto token embeddin... | {
"id": "464f0571-3e99-48dc-829c-0f4c9ad03d0d",
"created_at": "2026-08-08 11:17:31.621844+00:00",
"db_md5": "d90b481d36c8c66b9c23da5c39787a4d",
"submitter": "msnow0",
"github_login": "msnow0",
"run_id": "ddf02ada-5b03-40af-9c6e-fd90e1e6bba0",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.841646432876587,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.021666666492819786
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.495912790298462,
"example_count": 100,... |
465df8c6-e4d3-4ed1-9504-9de4cd0935dc | easy | satyaki-up | 2026-08-05 16:22:28.096248+00:00 | succeeded | 90212b36a17bc9b54976ab315954bc2ca30355dba1876dc624082ea3069c951a | 3,790 | null | """Basic single-pass Transformer with PyTorch AdamW."""
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
NUM_HEADS = 4
... | {
"id": "465df8c6-e4d3-4ed1-9504-9de4cd0935dc",
"created_at": "2026-08-05 16:22:28.096248+00:00",
"db_md5": "078579c99fa654c7d8ec681d30ab071f",
"submitter": "Satyaki",
"github_login": "satyaki-up",
"run_id": "5bc36897-a521-4f85-8e7d-ee7e7a994e4b",
"tier": "easy",
"dataset_id": "e1",
"status": "succeed... | {
"score": {
"mean_loss": 2.5948580503463745,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018333333544433117
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2644522190093994,
"example_count": 10... |
466de8c6-1543-45d0-9b72-efdb6187e3f0 | easy | khushidahi | 2026-08-07 04:45:10.796969+00:00 | succeeded | a1c5e9850ebbf8553cbc8168b8346237872848c4784640c0da82ac518ef619a0 | 7,292 | 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": "466de8c6-1543-45d0-9b72-efdb6187e3f0",
"created_at": "2026-08-07 04:45:10.796969+00:00",
"db_md5": "f863248d9618499f57047d3620e6f39c",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "58d53357-088c-4f79-bac5-9fd88f24f5c9",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.173213243484497,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000298023224
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.178095817565918,
"example_count": 600,... |
466f02d1-329b-4e99-aceb-a0a359e89d81 | easy | DDanlov | 2026-08-12 23:08:53.696894+00:00 | failed | 88cac49667d55d1dcececf8243a1a64244f82f24a9dadb35998a7baed8af1f2d | 28,966 | null | from __future__ import annotations
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as cp
from torch.optim.optimizer import Optimizer
try:
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
... | {
"id": "466f02d1-329b-4e99-aceb-a0a359e89d81",
"created_at": "2026-08-12 23:08:53.696894+00:00",
"db_md5": "b0cd394390a4839b67dc8fc2c7fa0cbc",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "cb5413da-1095-45ad-9ef5-17db02c4bcba",
"tier": "easy",
"dataset_id": "e4",
"status": "failed",
... | null
|
466f338d-5bc7-40c5-8e52-095d25e3dc0b | easy | Koenneker | 2026-08-05 09:08:21.005394+00:00 | succeeded | 7bf30e19d90e7f6cadb4ad67d0b2cd51634a4fd2d29a507be6f840a0dd78f8c1 | 3,628 | null | """Basic single-pass Transformer with PyTorch AdamW."""
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
NUM_HEADS = 4
... | {
"id": "466f338d-5bc7-40c5-8e52-095d25e3dc0b",
"created_at": "2026-08-05 09:08:21.005394+00:00",
"db_md5": "eaac69fe345f82973fcf392c504a40ff",
"submitter": "Ferdinand Könneker",
"github_login": "Koenneker",
"run_id": "642dbe4f-8f61-437e-a0ac-1f65d8daa895",
"tier": "easy",
"dataset_id": "e2",
"status"... | {
"score": {
"mean_loss": 2.5544973611831665,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01229166705161333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.7709851264953613,
"example_count": 300... |
466ffcc2-52af-42dc-9501-8e4b67a06c99 | easy | isaac0804 | 2026-08-04 17:26:14.123637+00:00 | succeeded | e150bfc799bbdeb702d126aed06816d9953c9ce4136dea1a005442383096f757 | 8,352 | null | """Weight-tied looped Transformer that reads T from the prompt to set its own depth.
The task is x_t = x_{t-1}^2 mod N for T steps. A single fixed-depth block has no
mechanism to represent "apply this operation T times" and cannot extrapolate to
depths unseen in training. This submission instead applies one shared blo... | {
"id": "466ffcc2-52af-42dc-9501-8e4b67a06c99",
"created_at": "2026-08-04 17:26:14.123637+00:00",
"db_md5": "fbe215c54c5741274101702dab493b79",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "90b7f17d-9ba1-474c-9b16-d1a1f66eb8a8",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 1.9280128479003906,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05833333171904087
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9340039491653442,
"example_count": 100... |
4675d1e7-80e4-439e-b817-c9bf7153104b | easy | sapient-sapiens | 2026-08-19 11:23:38.721484+00:00 | succeeded | a49fb11a941ad5b7b199ba49b7305d04fb0838b3cabead963bc7b2a9fd6a31e6 | 33,317 | 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": "4675d1e7-80e4-439e-b817-c9bf7153104b",
"created_at": "2026-08-19 11:23:38.721484+00:00",
"db_md5": "c5798ff07416602bd68622d18fe67528",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "d50c9086-8460-4bf6-8f37-3875464c3300",
"tier": "easy",
"dataset_id": "e9",
"status": "su... | {
"score": {
"mean_loss": 5.237375259399414,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.028888889588415623
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.437951564788818,
"example_count": 90,
... |
467748a0-6e30-4f47-8fde-7f44b84321ba | easy | sapient-sapiens | 2026-08-18 10:28:44.908643+00:00 | succeeded | 89c918ac93a5fc065ad7867e91776b20fd407ba988003d94183e75464eff5874 | 29,112 | null | """v40 lineage: T-free operator, N memory, PonderNet mix, no T in the loss.
Field-local encoder, T-free core, countable halt, answer-slot readout.
N is seeded and restored every step. T is never parsed in forward or loss.
"""
from __future__ import annotations
import math
import time
import torch
import torch.nn.fu... | {
"id": "467748a0-6e30-4f47-8fde-7f44b84321ba",
"created_at": "2026-08-18 10:28:44.908643+00:00",
"db_md5": "d3dc4665ec05e1ccf45ee5323608e8b4",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "94b12bb0-f022-4b21-aa06-75f7bd8434ae",
"tier": "easy",
"dataset_id": "e9",
"status": "su... | {
"score": {
"mean_loss": 1.8626393935542263,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013333333432674408
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8685052394866943,
"example_count": 90... |
4685a0f6-0eb2-4e73-a6b2-1faefe9e0467 | easy | newjordan | 2026-08-14 22:52:57.067202+00:00 | succeeded | 6da6e937508b55c0269fa90d648775dd5beecc4c8d7f4934fa089f933655e8a8 | 178,654 | 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": "4685a0f6-0eb2-4e73-a6b2-1faefe9e0467",
"created_at": "2026-08-14 22:52:57.067202+00:00",
"db_md5": "e6c32df21410d4316bc488cf172a8410",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "0fbfe584-f4ba-4106-b057-106efc989885",
"tier": "easy",
"dataset_id": "e9",
"status": "succeed... | {
"score": {
"mean_loss": 0.0029781069606542587,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0029874297324568033,
"example_count": 90,
... |
468869b5-e2ea-4db6-8217-ebd803cde34a | easy | oupadhyay | 2026-08-28 02:37:38.378738+00:00 | succeeded | 6ba82b5ffcbe4493f3d10e5ab3921b8ba3cd47382832b5f297b1b9da58d147be | 9,181 | null | """Compressed axial v10 with zero-initialized boundary-to-pair attention."""
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import OptimizerBundle, Submission, assert_model_state
PAD, N, X, T, ANS, DIGIT = 0, 2, 3, 4, 5, 7
MAX_STEPS, CHANNELS, STATE_ELEMENTS = 64, 64, 63146
class Con... | {
"id": "468869b5-e2ea-4db6-8217-ebd803cde34a",
"created_at": "2026-08-28 02:37:38.378738+00:00",
"db_md5": "49103cded34690ac50666ff67e4165a3",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "86a95de6-744c-4c23-8fb3-2f5e327cfd24",
"tier": "easy",
"dataset_id": "e6",
"status": "s... | {
"score": {
"mean_loss": 2.168792247772217,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010135135613381863
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.282790422439575,
"example_count": 60,
... |
468df304-151d-4142-a8ae-ec5adb59be40 | easy | khushidahi | 2026-08-03 21:31:18.259509+00:00 | succeeded | ae0f81fc9e06e0b7039acc6919b549ed0394e5e54b5e03aaf32bd3087114bca6 | 15,490 | 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": "468df304-151d-4142-a8ae-ec5adb59be40",
"created_at": "2026-08-03 21:31:18.259509+00:00",
"db_md5": "69d0bf3db42f338a7a7cfe798a67a588",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "411b8649-e908-4468-9783-276abe1c466c",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.206222891807556,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333767950535
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.200667142868042,
"example_count": 600,... |
4693d92c-1ddf-426c-a1d2-068e7a3c36b7 | easy | Mantissagithub | 2026-08-16 19:30:36.922467+00:00 | succeeded | 7bf30e19d90e7f6cadb4ad67d0b2cd51634a4fd2d29a507be6f840a0dd78f8c1 | 3,628 | null | """Basic single-pass Transformer with PyTorch AdamW."""
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
NUM_HEADS = 4
... | {
"id": "4693d92c-1ddf-426c-a1d2-068e7a3c36b7",
"created_at": "2026-08-16 19:30:36.922467+00:00",
"db_md5": "eaac69fe345f82973fcf392c504a40ff",
"submitter": "Pradheep P",
"github_login": "Mantissagithub",
"run_id": "161b0457-669e-4801-b3e8-922ec1c93c85",
"tier": "easy",
"dataset_id": "e6",
"status": "... | {
"score": {
"mean_loss": 4.039956331253052,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018468469381332397
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.647686958312988,
"example_count": 60,
... |
4696c0d4-941e-4db4-8916-217342d63277 | easy | gauravmishra | 2026-08-08 06:09:01.306646+00:00 | succeeded | f909f86b3b62d4a59bfa3d0a2d1d42ec6eb92b0601dd73c431cdd4d9431e8bda | 5,473 | null | """Four-block Transformer with four mutable per-forward memory tokens."""
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 = 1... | {
"id": "4696c0d4-941e-4db4-8916-217342d63277",
"created_at": "2026-08-08 06:09:01.306646+00:00",
"db_md5": "ceae26ba231943df422f344bbdf6c0fa",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "baee1810-282f-4c0a-8667-b18ba7dd5a78",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 9.377906147964154,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003750000024835269
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.004785238360075,
"example_count": 600... |
469746f2-7705-4b4b-85f4-125cb41af3e8 | easy | khushidahi | 2026-08-08 02:45:30.324541+00:00 | succeeded | c482d2475a1901b2e6b16caf68e93090a984009702021b5ad8b07d85a241070b | 15,586 | null | """R42 Fourier-number representation experiment.
All input-dependent numeric parsing stays inside differentiable GPU tensor
operations. This is a learned-model representation, not a task solver: it does
not compute modular squaring or create intermediate labels.
"""
from __future__ import annotations
import math
impor... | {
"id": "469746f2-7705-4b4b-85f4-125cb41af3e8",
"created_at": "2026-08-08 02:45:30.324541+00:00",
"db_md5": "0a28e7efe47d37b92b62c394122f105c",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "c0a34d17-578b-4ca9-973d-5fb5b3cca465",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.165603171564169,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333353511989
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1702184677124023,
"example_count": 600... |
46987a98-f9f9-4f93-9181-e2960abc0c2b | easy | Mantissagithub | 2026-08-19 10:51:12.494238+00:00 | succeeded | 0fbedfd82fdc87e3ec1dcc0108da95d23adce1a5983f308a33ee363f3f5d49a7 | 29,614 | 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": "46987a98-f9f9-4f93-9181-e2960abc0c2b",
"created_at": "2026-08-19 10:51:12.494238+00:00",
"db_md5": "d3173b25a9db031a083e9ed309455722",
"submitter": "Pradheep P",
"github_login": "Mantissagithub",
"run_id": "7c3e4f85-6f98-4d79-92ce-cbf0b3649b45",
"tier": "easy",
"dataset_id": "e2",
"status": "... | {
"score": {
"mean_loss": 7.006813168525696,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.28562501398846507
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 11.587806701660156,
"example_count": 300,... |
46a873b5-b41d-4dba-8e1e-63198b0ca035 | easy | sapient-sapiens | 2026-08-31 22:22:03.872774+00:00 | succeeded | 96aa702b5909a147a59c8ec19a28c3e70c1f7c86c97c8cc6fee60ccd497a5c9f | 16,057 | null | """One-block nested E5 model: four learned Horner calls per square, eight squares."""
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_mod... | {
"id": "46a873b5-b41d-4dba-8e1e-63198b0ca035",
"created_at": "2026-08-31 22:22:03.872774+00:00",
"db_md5": "2451c806f02bd1faa9eb07c5c2bc974b",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "31468068-08b6-4059-aa14-774cded549b5",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.1524351945334512,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004583333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.15439102040278,
"example_count": 600,... |
46bc7b76-5ea4-4b37-8625-8ed988d51d73 | easy | karanganesan | 2026-08-06 05:04:32.481043+00:00 | succeeded | 80e70bb3f53e19fa5704064ef0eb82d444c96b20df15f01392f739a38444341c | 25,611 | null | """Parametric looped-transformer family (P1).
One weight-tied transformer block applied k times in latent space. Config
flags cover four P1 families with one file:
looped recall=0 gated=0 tfilm=0 plain weight-tied loop
looped-recall recall=1 re-inject the input embedding each
... | {
"id": "46bc7b76-5ea4-4b37-8625-8ed988d51d73",
"created_at": "2026-08-06 05:04:32.481043+00:00",
"db_md5": "a98a5f6a373ca3d3c05a258b03bfa2a0",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "e8aa767b-2522-4731-a0b4-3c0a086c6181",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.405458927154541,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666716337204
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3720717430114746,
"example_count": 100... |
46bdbfbb-c813-45e4-82ff-4ae1bcd72618 | easy | khushidahi | 2026-08-05 23:28:18.111610+00:00 | succeeded | e5dcd93cd3ee08829b364a7fbd6c1c762d577ba0ed38938b20158c6902a53790 | 17,392 | 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": "46bdbfbb-c813-45e4-82ff-4ae1bcd72618",
"created_at": "2026-08-05 23:28:18.111610+00:00",
"db_md5": "4f396632098b1049bb54451f135d2f7b",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "d58f43ad-58ba-4f8e-a229-7da1869e9ad2",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.2111889123916626,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666977107525
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2088539600372314,
"example_count": 60... |
46bf5127-ba90-4036-a8af-22e06b7ace34 | easy | erdavis0 | 2026-08-29 00:42:03.367384+00:00 | succeeded | 6dd63d7fbd57f1dfcb99a68f4eb1c3004d5da2334182db359d857967577b9df9 | 20,131 | null | """Curvature table with a compact gentle modulus-relative residual."""
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,
TokenLossBatc... | {
"id": "46bf5127-ba90-4036-a8af-22e06b7ace34",
"created_at": "2026-08-29 00:42:03.367384+00:00",
"db_md5": "02d94eaf51f2ef55a1f2bce028a4c985",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "e4b620a3-e714-4000-b641-7e0b26ae9e58",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",... | {
"score": {
"mean_loss": 0.00023994335788302124,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.00016433870769105852,
"example_count": 60,
... |
46c92274-496b-423f-ae77-d062dd9d0269 | easy | erdavis0 | 2026-08-27 03:58:19.994565+00:00 | succeeded | ca6f1b5c93401e93c705f4189e451ef833a9103266142696e3b482501ff8d97a | 31,305 | null | """Typed-prefix reversible transition B1: CE plus selector-oracle KL.
Public delimiter syntax routes opaque N, X, and T fields into three strict
functional roles. X initializes only an evolving auxiliary tape, N initializes
only an invariant environment lane, and T reaches only a learned monotone
prefix controller. ... | {
"id": "46c92274-496b-423f-ae77-d062dd9d0269",
"created_at": "2026-08-27 03:58:19.994565+00:00",
"db_md5": "3f9846560514c463a6ae282585eac958",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "4fb7a5c6-ebfe-404d-a591-66019c83f621",
"tier": "easy",
"dataset_id": "e7",
"status": "succeeded",... | {
"score": {
"mean_loss": 6.153851747512817,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.11564171314239502
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.116339683532715,
"example_count": 85,
... |
46cc7541-f09d-41be-b002-9a3a684f9902 | easy | DDanlov | 2026-08-27 21:30:02.037468+00:00 | succeeded | 8d415cc1f63d080585d0c824893944ca92ab4ec1524d3cac5eb52ded6b075276 | 38,194 | 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": "46cc7541-f09d-41be-b002-9a3a684f9902",
"created_at": "2026-08-27 21:30:02.037468+00:00",
"db_md5": "ce30cec8ff4f02127ec4ecf3ec8fa378",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "15973ee1-af82-45da-8a7f-87c0f2ba8494",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.4779702706916913,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00499999995653828
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4889447689056396,
"example_count": 600... |
46ceb3fd-cc03-4b21-97e3-f3cd508391f9 | easy | sankalp1999 | 2026-08-04 12:07:42.052517+00:00 | succeeded | 5962733f959df02b97efe17038be4d172bcd59e1804ac7a621e583058c1c8c13 | 7,814 | null | """Latent recurrence: encode once, apply a shared squaring map T times, decode.
v4 put the loop inside a transformer and underfit badly (train acc 4% after
4,432 steps) because attention cost ~7ms/step at 13 tokens and bought nothing.
v2 fit the train set perfectly with a cheap flattened encoder but memorized,
because... | {
"id": "46ceb3fd-cc03-4b21-97e3-f3cd508391f9",
"created_at": "2026-08-04 12:07:42.052517+00:00",
"db_md5": "2df1e85a69d51f54e4aaf8888a56f15b",
"submitter": "Sankalp ",
"github_login": "sankalp1999",
"run_id": "06f0f896-25af-4db8-8a44-7588dfd8b215",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 5.7700419425964355,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.053333332762122154
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.509758472442627,
"example_count": 100... |
46d34e26-78e5-4e9d-9716-60c5a2d639a7 | easy | liam-gb | 2026-08-13 00:55:34.587001+00:00 | succeeded | eb2504bb7d0157e632cf799a1f18f53674416b6d2c2037173c708bf726abbc5b | 14,358 | null | """rns_pm_gather_ct_widet: dataset-sized digit head, dead loop tail trimmed.
The wide readout is what makes long answers reachable; the trim is what pays
for it, by not running loop iterations whose update is multiplied by zero.
Base docstring:
Single change vs rns_pm_gather_ct: the readout carries one
distance-from... | {
"id": "46d34e26-78e5-4e9d-9716-60c5a2d639a7",
"created_at": "2026-08-13 00:55:34.587001+00:00",
"db_md5": "1a548bea555b987cf0a175795dc3aa83",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "83b522e3-2303-48c8-9506-49ba16180bd4",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 4.265423225725556,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02750000009313226
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.531272888183594,
"example_count": 600,
... |
46dead24-1e46-4fe6-b098-ff1b6778da7f | easy | DDanlov | 2026-08-31 20:44:02.390069+00:00 | succeeded | 4793b22411a5630f589d206194fdec43a6ad51b798935c6cb7fc26f912209299 | 32,407 | null | """
Unified Continuous Hopfield Equilibrium Propagation Deliberation Engine
=======================================================================
Architecture:
- Deliberation: Continuous Hopfield Multi-Layer Energy System (L=4, D=768, d_ffn=2304, H=16)
- Integration: In-Place Strang-Split Conformal Symplectic Velocit... | {
"id": "46dead24-1e46-4fe6-b098-ff1b6778da7f",
"created_at": "2026-08-31 20:44:02.390069+00:00",
"db_md5": "9744d76628aef381d04c558e15eb4fd3",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "2c56a432-704f-4733-8f9b-2c9bade81fdb",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.0748478725441135,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0022916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.0942932369700697,
"example_count": 1... |
46ef96b8-2fa2-4aad-aa7d-9bc455d14d4f | easy | LauraGomezjurado | 2026-08-30 22:17:18.696642+00:00 | succeeded | 048c5332dab3f345d86e0a32e260258ea84b61994994d31eb1a35dbf028c94cc | 11,853 | 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": "46ef96b8-2fa2-4aad-aa7d-9bc455d14d4f",
"created_at": "2026-08-30 22:17:18.696642+00:00",
"db_md5": "0feaae9a00abd4574202a30d72615940",
"submitter": "LauraGomezjurado",
"github_login": "LauraGomezjurado",
"run_id": "8df2b0ce-9205-46b9-9799-99be3e659780",
"tier": "easy",
"dataset_id": "e7",
"st... | {
"score": {
"mean_loss": 2.8697301325297753,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05592691644016859
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.052445204665021,
"example_count": 85,
... |
46f2180e-54bc-41bc-bbb1-1c130bf6e129 | easy | eig1n | 2026-09-01 03:43:04.425545+00:00 | succeeded | 7c76dde4cbe694d52eabf4d92d4b46b3c98e4df50bdb57779b05b4793526accf | 8,750 | null | """Learned finite-state operator: a row-stochastic transition applied by matrix power.
Rule notes
----------
* Rule 6: every trainable tensor uses a PyTorch default random initialisation
(the sole exception is ``RMSNorm.weight = ones``, as in the organisers' own
``submissions/baseline_adamw`` reference) and is upd... | {
"id": "46f2180e-54bc-41bc-bbb1-1c130bf6e129",
"created_at": "2026-09-01 03:43:04.425545+00:00",
"db_md5": "73d5ff38893c88e502428c208939a243",
"submitter": "eig1n",
"github_login": "eig1n",
"run_id": "9be3aca3-c09c-4aff-9c6b-655f6179d60e",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
... | {
"score": {
"mean_loss": 7.840900659561157,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03999999910593033
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.4856390953063965,
"example_count": 100,... |
46f46c27-045c-4e08-a1f2-7e3885a5883c | easy | epicosy | 2026-08-22 23:01:06.343371+00:00 | succeeded | b44428c2eb050922fc23594e21180e85d23fcd3ac6f914066b89fd966b9a12e6 | 5,430 | null | """Latent-state Universal Transformer baseline for the Easy tier.
The prompt stream is encoded once and kept fixed. A single shared block
recurrently updates learned latent/query tokens through cross-attention to the
prompt. The final logits remain aligned with the original input sequence, as
required by the public su... | {
"id": "46f46c27-045c-4e08-a1f2-7e3885a5883c",
"created_at": "2026-08-22 23:01:06.343371+00:00",
"db_md5": "b553ea8294865682d9c0b34d689a0a79",
"submitter": "Eduard Pinconschi",
"github_login": "epicosy",
"run_id": "dbf66796-0fa5-4a8e-ac24-058db60c4c33",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 2.3809860944747925,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.054999999701976776
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4872772693634033,
"example_count": 10... |
46f58fa2-0b3e-4873-b1e6-a673a7f5abab | easy | repst | 2026-08-22 20:36:55.086884+00:00 | failed | 034ca36f95a6236d2e3408d52819374a7e698d85377ef83480ef0ee57ed04cf4 | 18,615 | null | # R521: published-parser sampled-search recurrent program machine.
#
# LEARNED: anonymous numeric-field route, program length, recurrent instruction
# policy, selected modular transition, and refinement of the digit lexicon.
# Training explores an MDL-stratified 96-of-595 semantic vocabulary; evaluation
# greedily exec... | {
"id": "46f58fa2-0b3e-4873-b1e6-a673a7f5abab",
"created_at": "2026-08-22 20:36:55.086884+00:00",
"db_md5": "debe287ece6f5f8a477931d80dff3e39",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "d4dfcaae-0e69-4554-bb35-92c619776f4a",
"tier": "easy",
"dataset_id": "e1",
"status": "failed"... | null
|
46fb559a-1694-4933-a818-7bd49f62ea73 | easy | chuk-uzowihe | 2026-08-12 17:33:11.721857+00:00 | succeeded | 9f686561e57745d74a462fbc4c39913f399f7dd8f5281e8cee274372fae6ab4b | 23,007 | null | """v10: dense fixed-loop recurrence with deep supervision.
Lineage: FPRM structure (alpha/beta-tied residual scaling, pre-rms_norm,
input injection every iteration) + the v8 2D digit lattice / RoPE task
encoding. Three deliberate departures from the v9 MoEUT line:
1. **Full BPTT.** The loop runs a fixed N_ITER iterat... | {
"id": "46fb559a-1694-4933-a818-7bd49f62ea73",
"created_at": "2026-08-12 17:33:11.721857+00:00",
"db_md5": "89e4f9bdd9c53aaa1325862799c73ce6",
"submitter": "Chuk Uzowihe",
"github_login": "chuk-uzowihe",
"run_id": "f4a0a3f5-df5f-4414-8d93-27ded7a400a0",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 24.25483226776123,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07999999821186066
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 20.951358795166016,
"example_count": 100,... |
47072619-392d-4b76-b05b-186401ddd9af | easy | chad-atexpedient | 2026-08-10 02:13:40.501253+00:00 | succeeded | 6e8d9f59f3296741ad218f7dc41c6ecd5df52fd1c4ec29b482008e72ff37431f | 10,942 | null | """BL8C: width-64 final-readout control."""
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
IMMUTABLE_LAN... | {
"id": "47072619-392d-4b76-b05b-186401ddd9af",
"created_at": "2026-08-10 02:13:40.501253+00:00",
"db_md5": "778e503a475e99b26eac5365c1c587d0",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "36096061-1fc7-4727-81ee-3360306e6b00",
"tier": "easy",
"dataset_id": "e4",
"st... | {
"score": {
"mean_loss": 4.647446177893486,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0027083333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.666327064525564,
"example_count": 120... |
470a130a-376b-4fdb-afcc-ee0bcdb4f776 | easy | rteehas | 2026-08-19 11:41:15.718655+00:00 | succeeded | 4ddc57b8edc46a40f195d9c15a57dcc08b717ffa69ec49c0dca5359e443401b2 | 53,669 | 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": "470a130a-376b-4fdb-afcc-ee0bcdb4f776",
"created_at": "2026-08-19 11:41:15.718655+00:00",
"db_md5": "6fa6d743b6a8bcd3373b192d9b2902f5",
"submitter": "Ryan Teehan",
"github_login": "rteehas",
"run_id": "64b3a6f6-a816-4eee-b47f-1564b2937a67",
"tier": "easy",
"dataset_id": "e2",
"status": "succee... | {
"score": {
"mean_loss": 2.833213448524475,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8332133293151855,
"example_count": 300,
"exac... |
470a62e3-ba2b-4810-a345-233343d3f35f | easy | jordanrubin | 2026-08-12 04:49:17.181295+00:00 | succeeded | 5e67d1a282bda9fab14fa7d6e3200be9c6aac5fca5eb9b7f7a229db9bcf84720 | 22,494 | 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": "470a62e3-ba2b-4810-a345-233343d3f35f",
"created_at": "2026-08-12 04:49:17.181295+00:00",
"db_md5": "38f0396c0fc03a18e69ddda1edcd8120",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "53c65e18-1d61-4f48-94a1-0a8911acbddb",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.5712389426625473,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013333333035310108
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1615587398075737,
"example_count": 60... |
4711f9fe-aa94-4554-97a3-f9d67290ec48 | easy | rougee | 2026-08-23 16:30:16.721461+00:00 | failed | 0bccbf839e66f5dccd3c3d6eb9c2e46584c33809c7c6bc73b1d8f69611a1cd14 | 12,923 | null | """Official Submittable Champion for One Layer Deeper.
Hardware & Algebraic Co-Design:
1. Dyadic Harmonic Phasor Prelude: Power-of-2 spectrum omega_k = 2^k * pi on (S^1)^16 (pointwise, zero dim=1 reduction)
2. Matrix-Exponential Lie Group SO(d): W = exp(S) where S = A - A^T (native GPU kernel, zero linalg.solve crashe... | {
"id": "4711f9fe-aa94-4554-97a3-f9d67290ec48",
"created_at": "2026-08-23 16:30:16.721461+00:00",
"db_md5": "7872026f91b15d2655c631151adea908",
"submitter": "Rougee",
"github_login": "rougee",
"run_id": "20ed63c6-a17b-4332-bbf0-be05f6341892",
"tier": "easy",
"dataset_id": "e1",
"status": "failed",
"... | null
|
471462a8-9a2b-4d3c-943d-334078d20a46 | easy | priteshraj10 | 2026-08-10 23:48:56.332009+00:00 | succeeded | 3852dd49a984e8b230ae1ab9f59b24aa8034d8e392cca556e56b960249a98b68 | 6,973 | null | """Recurrent Universal Transformer with SwiGLU FFN, Cosine Warmup, Batch Reuse, and Dropout Regularization."""
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,
Su... | {
"id": "471462a8-9a2b-4d3c-943d-334078d20a46",
"created_at": "2026-08-10 23:48:56.332009+00:00",
"db_md5": "74412b6c5eba7d53d0bfd20f31992742",
"submitter": "Pritesh raj ",
"github_login": "priteshraj10",
"run_id": "53d081a8-246f-4b3b-a4b7-b93c0cfcc455",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 1.9759219884872437,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06666666455566883
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8952438831329346,
"example_count": 100... |
472085b1-d6f8-4b4e-b846-f7a1a7621716 | easy | sapient-sapiens | 2026-08-25 09:10:18.283698+00:00 | succeeded | 68f6b35118e4eb708459b2d1bf0dc2dcbf61e723da023323b80503453419e658 | 18,317 | 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": "472085b1-d6f8-4b4e-b846-f7a1a7621716",
"created_at": "2026-08-25 09:10:18.283698+00:00",
"db_md5": "3bced06c4cbeeffe968ad51451c4cb56",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "6b806a89-d89f-4e92-87bd-f575007e771b",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 10.40665523334209,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013333333333333332
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.24973361695294,
"example_count": 600,... |
4724ad6c-a609-4c8a-81f4-98860e91f9eb | easy | newjordan | 2026-08-09 20:15:22.209103+00:00 | succeeded | a729f6bb0df1dfe48b68658b6d747e9d765131484ce839d06a15dac32399a53d | 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": "4724ad6c-a609-4c8a-81f4-98860e91f9eb",
"created_at": "2026-08-09 20:15:22.209103+00:00",
"db_md5": "c3aa11c6494d5acd02af87ecf0d026d9",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "33612a55-a129-4e1c-b116-26a950170d71",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 3.822435792244505e-05,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.830185960396193e-05,
"example_count": 300,
... |
47278c0f-42a3-4d2d-8d7f-2a624caf20fe | easy | Koenneker | 2026-08-05 09:30:31.726374+00:00 | succeeded | fee387cca6c7e0403d800583b2caf6f881aba8f546601e30ae71e799cb4aa59e | 3,194 | null | """Vanilla single-layer LSTM with PyTorch AdamW at batch size 200.
Identical to `submissions/baseline_lstm` except for `batch_size=200`.
E1 has 600 training examples and the manifest sets `drop_last: true`, so
the manifest default of 512 uses one batch per epoch and discards 88
examples (15%) every epoch. 200 divides ... | {
"id": "47278c0f-42a3-4d2d-8d7f-2a624caf20fe",
"created_at": "2026-08-05 09:30:31.726374+00:00",
"db_md5": "e103231cbe922a8a1845a6e6dc7b1f3e",
"submitter": "Ferdinand Könneker",
"github_login": "Koenneker",
"run_id": "94d7f901-4aaa-4be9-91c4-df8f41f811e2",
"tier": "easy",
"dataset_id": "e4",
"status"... | {
"score": {
"mean_loss": 6.767419822497306,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003020833265812447
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.451087948215144,
"example_count": 1200... |
47302fcc-4e36-47ed-82f9-cee85c8beaa3 | easy | lpbb | 2026-08-14 10:21:25.122546+00:00 | succeeded | c5c02be597ae5929749de3bbfa8074c2581d90492a2daf440b4dcb675be99d8b | 4,165 | null | """CHEAT L0 — oracle: computes x^(2^T) mod N with integer ops inside forward and
emits the answer directly as logits. Not a submission; a plumbing + ceiling probe:
if this does not score ~100%, our understanding of the data format is wrong."""
from __future__ import annotations
import torch
from torch import Tensor, ... | {
"id": "47302fcc-4e36-47ed-82f9-cee85c8beaa3",
"created_at": "2026-08-14 10:21:25.122546+00:00",
"db_md5": "0c78764713fb54b16baa9538cb649f15",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "dcf7c09f-7433-418a-afb7-7ac03c4a130f",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",
"s... | {
"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": 800,
"exact_accuracy": 1.0,
"... |
4731af4a-0d2b-49e4-a400-c2b1b485208f | easy | briheet | 2026-08-05 07:14:14.627678+00:00 | succeeded | 58e47852c66b2d17be7b3c4ff4d3590b844931eee832ac8ee70a5ae92b6f6150 | 3,629 | null | """Basic single-pass Transformer with PyTorch AdamW."""
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 = 96
NUM_HEADS = 4
... | {
"id": "4731af4a-0d2b-49e4-a400-c2b1b485208f",
"created_at": "2026-08-05 07:14:14.627678+00:00",
"db_md5": "be0464d036b96770f3bb4da2ed5e698c",
"submitter": "Aizenn",
"github_login": "briheet",
"run_id": "b43bef1f-5b79-4bd9-9f5e-966a6c45b634",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",... | {
"score": {
"mean_loss": 3.0403127670288086,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999828636646
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2239017486572266,
"example_count": 100... |
47388805-f73f-453f-962c-fa5bc75c56ee | easy | KaustubhKumar05 | 2026-08-27 12:08:18.789437+00:00 | succeeded | 19134227419598e0e4d80e3119822e122978fadfbc99eccee4b6fb731792f779 | 12,508 | null | """abacus_nodecay — CONSTANT LR for the whole run. No decay, no floor.\n\nMuon runs for the WHOLE run and the LR schedule is the ORIGINAL hold=0.85 /\nfloor=0.02. This is the configuration the two collapsed m4 runs used. It exists\nso the 2x2 has a measured control on every dataset, not just m4.\n\nPair with abacus_fre... | {
"id": "47388805-f73f-453f-962c-fa5bc75c56ee",
"created_at": "2026-08-27 12:08:18.789437+00:00",
"db_md5": "7a0349936d2d493300b6ac12c0346127",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "ae317bc0-00ea-4c67-b032-9c717ed75e0c",
"tier": "easy",
"dataset_id": "e3",
"status": "succee... | {
"score": {
"mean_loss": 4.537202601288634,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006875000000000001
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.5459360163604,
"example_count": 800,
... |
4739a66b-d947-42fe-952e-f9afff36d92d | easy | viridale | 2026-08-02 16:06:45.337177+00:00 | succeeded | e9a687910975e81e9c2058c04329436cf50c7a4a8ef6b8d83bcd5f31eb7ca03e | 55,786 | null | """Physically isolated evidence-triggered five-monomial expansion.
hypothesis: v108 correctly skips family-7 computation on practice, but sharing
one 16.9M rule parameter still materializes the unused tail's gradient and
yields only 2,862 official steps. Store the v102 pool and family 7 in two
learned par... | {
"id": "4739a66b-d947-42fe-952e-f9afff36d92d",
"created_at": "2026-08-02 16:06:45.337177+00:00",
"db_md5": "1ba1b7bf2246a8d9fc1b9a3d48fdfed2",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "6980c1dd-ca89-4608-a922-81ad2b0f2c78",
"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,
"... |
4742cdae-a78a-4720-9e1d-e52e5002e921 | easy | sapient-sapiens | 2026-08-29 21:59:33.032612+00:00 | failed | b090a8a21bd29c49903bc30cf5024404461d7a5cd1857596ada82c07874e07cd | 15,384 | null | """Three-loop one-layer Transformer for iterated modular squaring.
The learned arithmetic operator is one tied pre-norm Transformer block. Every
square makes four Horner-digit calls and every digit makes nine micro-calls.
N and the complete operand are immutable during a square; scratch resets at a
digit boundary; an... | {
"id": "4742cdae-a78a-4720-9e1d-e52e5002e921",
"created_at": "2026-08-29 21:59:33.032612+00:00",
"db_md5": "180b6194d45fd5733c733ae119bbd2d5",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "3858e198-f907-4558-93da-9add3de07834",
"tier": "easy",
"dataset_id": "e5",
"status": "fa... | null
|
47466439-40db-47b1-8cf0-a82992d5fb3e | easy | repst | 2026-08-26 15:11:46.488369+00:00 | failed | 9edae9166abb83aaa02600854e4db228892812bf687ede9c9e95f313dc5518ac | 131,861 | null | # R737: immutable-input-readable recurrent instruction vocabulary.
#
# R736 could learn one shared complete program from complete-answer likelihood,
# but its recurrent STEP was unary: after SETUP it could see only evolving state
# Z. R737 makes the machine interface-complete for immutable input X by adding
# every sem... | {
"id": "47466439-40db-47b1-8cf0-a82992d5fb3e",
"created_at": "2026-08-26 15:11:46.488369+00:00",
"db_md5": "6c2e9da10608f8b9e5735d1fe201d1ae",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "ab7fa6f5-ac26-4dfa-99e7-ede2297980f9",
"tier": "easy",
"dataset_id": "e1",
"status": "failed"... | null
|
47575f18-93eb-427c-a095-5a6388034cb0 | easy | KaustubhKumar05 | 2026-08-26 11:23:27.243252+00:00 | succeeded | 3306060499953b1be422ba0d3cc1880149c86823de34ad66391d530b271f2538 | 12,488 | null | """stationary — candidate C: force a behaviourally stationary operator.
E7: the accuracy cliff sits EXACTLY at max(train T) in every row of the sweep
(train to 4 -> dead at 8; train to 16 -> dead at 32). Training with more recurrence
only relocates the cliff. So depth is being memorised, not iterated.
E8: measured wh... | {
"id": "47575f18-93eb-427c-a095-5a6388034cb0",
"created_at": "2026-08-26 11:23:27.243252+00:00",
"db_md5": "f362e4005c2b526cabfc297228e5fa6e",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "fe709ad7-dfbe-4667-8012-73a1899a03f4",
"tier": "easy",
"dataset_id": "e5",
"status": "succee... | {
"score": {
"mean_loss": 4.0606120866845075,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004166666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.090454083119807,
"example_count": 600... |
475a8bab-c475-4ff7-acdb-97e29afc9a31 | easy | shirvani-jr | 2026-08-30 23:55:21.276030+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": "475a8bab-c475-4ff7-acdb-97e29afc9a31",
"created_at": "2026-08-30 23:55:21.276030+00:00",
"db_md5": "cfec548a5587d3a2942b9c10220452bc",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "279b2c87-8ef7-464e-a15a-d5d627ac8b9d",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.8210963855581963,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.1255208324516813
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.369885815039284,
"example_count": 1200,... |
476588c1-4fd8-4cf3-a091-e538aeea4c15 | easy | jordanrubin | 2026-08-07 01:29:26.786773+00:00 | succeeded | 0eb249f609a0101d70cb07bbf9b0f2aa4cc7c5f7465d3e5e29bc4c3e68abecc6 | 14,786 | null | """Autonomous digit-state recurrent Transformer for repeated modular squaring.
The model is deliberately organized around one learned transition:
p_0 = right_aligned_decimal_digits(x)
p_{k+1} = F_theta(p_k, decimal_digits(N))
The same two-block Transformer cell is applied exactly T times. T is used only
as ... | {
"id": "476588c1-4fd8-4cf3-a091-e538aeea4c15",
"created_at": "2026-08-07 01:29:26.786773+00:00",
"db_md5": "497d08edf33fa52921753c6fa8942084",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "2fa1ab78-13ea-47c0-afc8-f9e732f235e0",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.080394816234337,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009999999776482582
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.109641040788103,
"example_count": 600,... |
476b2cac-3615-475f-97ec-d199559d3d51 | easy | jordanrubin | 2026-08-06 19:26:15.138288+00:00 | succeeded | 2214473aabeee3e5a3252928aedd50993436405bf900fc579ecc4ef88811f4f4 | 14,788 | null | """Autonomous digit-state recurrent Transformer for repeated modular squaring.
The model is deliberately organized around one learned transition:
p_0 = right_aligned_decimal_digits(x)
p_{k+1} = F_theta(p_k, decimal_digits(N))
The same two-block Transformer cell is applied exactly T times. T is used only
as ... | {
"id": "476b2cac-3615-475f-97ec-d199559d3d51",
"created_at": "2026-08-06 19:26:15.138288+00:00",
"db_md5": "f917ba4387feee357d1f6822e4a7825f",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "6b13ca4c-01aa-4be5-9eca-57e6aebf3f92",
"tier": "easy",
"dataset_id": "e3",
"status": "s... | {
"score": {
"mean_loss": 2.1217712641711164,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003124999930150807
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.122269399937432,
"example_count": 800... |
476cd3fe-bf40-466f-bc7b-0ba3b14906be | easy | c-nnr | 2026-08-04 13:17:14.340062+00:00 | succeeded | 838cd0e4ee6591b873a984fe556ed6917f4de0223405b232e838809c864abe37 | 6,469 | null | """One Layer Deeper E1 final-15 experiment: Embedding-scale sweep above 0.1 to retain stronger token geometry without default-scale overconfidence."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
Op... | {
"id": "476cd3fe-bf40-466f-bc7b-0ba3b14906be",
"created_at": "2026-08-04 13:17:14.340062+00:00",
"db_md5": "ce9a8f9f3496d329e605481620f178e9",
"submitter": "c-nnr",
"github_login": "c-nnr",
"run_id": "b91331be-151b-419f-91a6-27c12c131c94",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
... | {
"score": {
"mean_loss": 6.979811906814575,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.024999999441206455
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.046462535858154,
"example_count": 100,... |
476e2c2d-de6a-4a5d-9b51-a0af30bf44ff | easy | sirish-gambhira | 2026-08-10 01:12:06.535369+00:00 | succeeded | b45c3a7c939775427efda0073a2e3f248c0456549c013236b1fe60df5e023870 | 7,521 | null | """Decimal-structure-aware bidirectional Transformer for token arithmetic."""
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_mode... | {
"id": "476e2c2d-de6a-4a5d-9b51-a0af30bf44ff",
"created_at": "2026-08-10 01:12:06.535369+00:00",
"db_md5": "72425dc98b937d28806a502c247ab46a",
"submitter": "Sirish Gambhira",
"github_login": "sirish-gambhira",
"run_id": "44132df3-ebdc-493f-85df-80638e9fe70a",
"tier": "easy",
"dataset_id": "e2",
"stat... | {
"score": {
"mean_loss": 3.1420058012008667,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333972841501
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.126713275909424,
"example_count": 300... |
4772b564-2645-490e-8c7c-ccd98445cd62 | easy | oupadhyay | 2026-08-24 00:37:32.757897+00:00 | succeeded | 167b61ed5d81b7ed3bef6a948e0f4b57cfb52dfef8646cd94c369ce3d06f60e5 | 4,151 | null | """Pairwise digit-relation scan for the Easy tier."""
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=4,48,96
class C:
def __init__(s,vocab_size,max_seq_... | {
"id": "4772b564-2645-490e-8c7c-ccd98445cd62",
"created_at": "2026-08-24 00:37:32.757897+00:00",
"db_md5": "59239a510be669c7a35cc48882f8e889",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "7a9ba1a5-7ad7-4393-a5c8-28eba3e315e2",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.150695824257818,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010000000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1415912581131598,
"example_count": 600... |
4784249a-d1d9-4bd2-bf7b-658a71382191 | easy | yashkant | 2026-08-17 01:10:56.226166+00:00 | succeeded | 46d961099bd6474534a8967381bf85854be30ba743230227329ef816cace678d | 43,555 | null | """Round2190 removes cross-step Dykstra correction carry.
The retained official carrier, primal recurrence, two learned projections,
antisymmetric set codes, parsed-T allocation, optimizer, and EMA are fixed.
The candidate reuses the learned initial A/B correction offsets on every
cycle instead of carrying hand-writte... | {
"id": "4784249a-d1d9-4bd2-bf7b-658a71382191",
"created_at": "2026-08-17 01:10:56.226166+00:00",
"db_md5": "e215d83c1cec4ff789eda5276fca768c",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "9ad4afac-3c7d-4c35-8f4c-b5103da26b3a",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 2.15193199400705,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333345750967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1569174790061645,
"example_count": 600,... |
4787564c-9e3d-440b-bd6c-1067ed96be35 | easy | pepinu | 2026-08-26 17:37:10.611100+00:00 | succeeded | 74307ba76d035d71e8330e51db839434ca24ae8d3efbab3f4b12dd2ec11dd273 | 46,270 | null | """Squaring mod N as a learned permutation on CRT components.
WHY THIS SHAPE. (Z_N, x) is a commutative monoid, so every finite-dimensional
linear representation splits into multiplicative characters and squaring is
diagonal there -- the component basis is the only one in which this operation is
shallow. Two consequen... | {
"id": "4787564c-9e3d-440b-bd6c-1067ed96be35",
"created_at": "2026-08-26 17:37:10.611100+00:00",
"db_md5": "d7f34595d7f996ce3ab1db374501668a",
"submitter": "pepinu",
"github_login": "pepinu",
"run_id": "dbe1d758-c20e-4359-b8ce-e9ea64c8ce57",
"tier": "easy",
"dataset_id": "e2",
"status": "succeeded",
... | {
"score": {
"mean_loss": 0.014268925413489342,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.014290287159383297,
"example_count": 300,
... |
4787d25d-2f8f-4ea6-96aa-7c5c97c1707e | easy | adambrzosko | 2026-08-30 16:57:16.245501+00:00 | succeeded | d8c14782c338ec1fe31cc37263030830ef238cf0bef0f072f479d27541724344 | 14,140 | null | """One Layer Deeper — digit-slot recurrent machine ("dm").
The prompt is [N] digits(N) [X] digits(x) [T] digits(T) and the label is the
decimal expansion of x^(2^T) mod N, tail-aligned: the last prompt position
carries the units digit, the one before it the tens digit, and so on.
Architecture. The model parses the ... | {
"id": "4787d25d-2f8f-4ea6-96aa-7c5c97c1707e",
"created_at": "2026-08-30 16:57:16.245501+00:00",
"db_md5": "223a2a05a1dd04c43a6f6fb6242202f9",
"submitter": "Adam Brzosko",
"github_login": "adambrzosko",
"run_id": "52f93fe9-f8fc-44e9-91ae-8b1685d5c900",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 3.3907331984553517,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666679084302
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2662241833092387,
"example_count": 60... |
478a94d5-5783-46f3-911b-6afe57ade53a | easy | dlyr3 | 2026-08-23 19:59:42.694936+00:00 | failed | f94b635f6dd2b70c17e8a3d599ceaea226881d1a6201c012187cb986cfacb760 | 7,386 | 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 is deliberately NOT applied: measured on e1,
any nonzero lambda drives ||exp(M)||_F... | {
"id": "478a94d5-5783-46f3-911b-6afe57ade53a",
"created_at": "2026-08-23 19:59:42.694936+00:00",
"db_md5": "3f569f9edd87fb3629c5fd5274b314cf",
"submitter": "dlyr3",
"github_login": "dlyr3",
"run_id": "541899e5-cf06-4d96-a148-1b314f4582cd",
"tier": "easy",
"dataset_id": "e1",
"status": "failed",
"st... | null
|
47906cec-dfb9-4354-aa42-bdf562baf0bb | easy | amirmeisami | 2026-08-16 17:17:30.106403+00:00 | succeeded | e1e6a8f27504f9f087b91e90894d5a57cced05d23bd63eae47a9a85e4190952f | 21,071 | null | """E011-A1: compact closed-state digit transition."""
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,
assert_model_state,
)
... | {
"id": "47906cec-dfb9-4354-aa42-bdf562baf0bb",
"created_at": "2026-08-16 17:17:30.106403+00:00",
"db_md5": "fd1f3739d553b08f714ea275cd755874",
"submitter": "Amir Meisami",
"github_login": "amirmeisami",
"run_id": "48e9e87b-7a48-44c0-8fa9-eed2e3926e18",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.7327165840152166,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00625
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.737352867030242,
"example_count": 600,
"... |
4790eb11-e49c-48ae-bd32-65b2b8f7a033 | easy | repst | 2026-08-28 18:46:41.879636+00:00 | failed | 5790bdb050b10430572d9154ce03d9f23ff3139c219ecaee16c46b7b535c6087 | 32,634 | null | """
R783: learned primitive interface programs with recurrent neural execution.
Authored architecture, analogous to an LM vocabulary and execution substrate:
* the public decimal N/X/T token ABI and exact field parser;
* a minimum-description-length syntax of at most two SSA instructions per
setup/recurrent/readout... | {
"id": "4790eb11-e49c-48ae-bd32-65b2b8f7a033",
"created_at": "2026-08-28 18:46:41.879636+00:00",
"db_md5": "aff832343e1256b61e9bfe7d6657edd4",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "36e01048-6470-4e85-bb86-618ae4b90289",
"tier": "easy",
"dataset_id": "e1",
"status": "failed"... | null
|
4793cb50-c012-4efd-a57e-81afeccff919 | easy | shan23chen | 2026-08-08 01:53:00.898754+00:00 | succeeded | fac9a6ba626b775a45c79137ae81da85e5482292515a695407c955cf1237de73 | 20,626 | null | """Generic recurrent Transformer with normalized numeric and CRF objectives.
The model is deliberately task-generic: it learns a transition over prompt and
workspace tokens instead of embedding the public modular-squaring algorithm.
A learned linear-chain CRF models dependencies among adjacent output digits;
its endpo... | {
"id": "4793cb50-c012-4efd-a57e-81afeccff919",
"created_at": "2026-08-08 01:53:00.898754+00:00",
"db_md5": "5903e6a4e42aab5827fb69b8f504a3d2",
"submitter": "Shan Chen",
"github_login": "shan23chen",
"run_id": "7c46be54-937b-4ea7-883c-4ab4dfb53b8c",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 3.2770293951034546,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666567325592
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2375378608703613,
"example_count": 60... |
479a520c-9c59-4ce6-9727-2fdeedfa4a2d | easy | oupadhyay | 2026-08-25 08:37:46.233190+00:00 | succeeded | 7eb46a8a01a756636d64ba5b364018acc3090fb15123d413d1de58393842ba12 | 4,300 | null | """Tiny global quadratic state with a shared dynamic-width decoder."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import OptimizerBundle,Submission,TokenLossBatch,assert_model_state
E,H,K=32,16,64
class C:
def __init__(s,vocab_size,max_seq_len):s... | {
"id": "479a520c-9c59-4ce6-9727-2fdeedfa4a2d",
"created_at": "2026-08-25 08:37:46.233190+00:00",
"db_md5": "76fa89a21319de6152cdc31c8674f89a",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "90f0c9b3-500b-4131-9ac2-32334642fd53",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 3.3516250854612495,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004166666716337204
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.534492084263686,
"example_count": 600... |
47a49539-b4d3-4e31-95aa-191aac5a1a12 | easy | 0Chris5R | 2026-08-07 12:53:58.399813+00:00 | succeeded | 1c05ebdeeef849b2a953975b8032673e1ba410a57e4d9afa154d659f4f94db9b | 30,818 | null | """Shared-digit Neural GPU with late full-objective EMA evaluation.
The model aligns decimal digits by significance, forms a learned low-rank
second-order feature map of the current state, and evolves a two-dimensional
workspace with the original convolutional-GRU equations. Each arithmetic
transition applies the sam... | {
"id": "47a49539-b4d3-4e31-95aa-191aac5a1a12",
"created_at": "2026-08-07 12:53:58.399813+00:00",
"db_md5": "0a960a736b01d2b34694614a5c49cbd6",
"submitter": "Chris ",
"github_login": "0Chris5R",
"run_id": "125d619d-9948-49b9-af17-7e458cd9fe49",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.1919031294815756,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.012916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.513621065114111,
"example_count": 600... |
47a731cc-ee76-43d5-a9fc-b27842b8b971 | easy | khushidahi | 2026-08-17 12:55:47.220639+00:00 | succeeded | 3280aa6f72ddea278fd9f380882f8bd7762345d600fdef09d21c62b06db7ddfe | 20,577 | null | """R190: scale-matched full-schedule dual-state TRM transition.
This submission tests the strongest rules-safe generic baseline suggested by
the public competition discussion. It does not implement squaring, modular
reduction, or any other task solver. Numeric prompt fields are only rearranged
onto equal-length, rig... | {
"id": "47a731cc-ee76-43d5-a9fc-b27842b8b971",
"created_at": "2026-08-17 12:55:47.220639+00:00",
"db_md5": "3c209785592d27f615bbd293d8a45807",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "6bf4c687-358b-4978-b2df-217328dc7d68",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.714842097872205,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.262672653112711,
"example_count": 600,... |
47acdf16-1436-4cb1-bb4d-bfa0363872ec | easy | shirvani-jr | 2026-08-29 02:50:29.185601+00:00 | succeeded | 1bda53a65f2a2dc3e1e2a760c49b4d1c32a1aab6afe296cf11bb5e4fffc2528e | 48,940 | null | """Adaptive learned discrete recurrent fabric for One Layer Deeper.
A gradient-trained, operation-free recurrent computational machine. The
model learns, from the competition's endpoint supervision: how to parse the
prompt into a bank of categorical digit registers, a reusable transition
circuit applied serially, a l... | {
"id": "47acdf16-1436-4cb1-bb4d-bfa0363872ec",
"created_at": "2026-08-29 02:50:29.185601+00:00",
"db_md5": "f612e7a75854a9e696365ae7ed90fe53",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "496146ba-dd0b-45ec-b895-0f1ca17e225e",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.0139869451522827,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666716337204
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.028477430343628,
"example_count": 100... |
47c3dda8-ea08-4a5e-a779-613afb9e4924 | easy | AFeinfield | 2026-08-19 22:31:36.489046+00:00 | succeeded | 4c828c8742abaa200b3fcc01a2875473d35d03c453b690ba829a10c3d95e2492 | 17,102 | null | """077: ACTIVE-ROW trajectory loop (compute optimization of 073).
Identical model, optimizer and training recipe as 073; the only change is
how ``Model.forward`` runs the T-indexed trajectory loop.
073 applied the block to the FULL batch on every one of ``max_T`` steps,
even though a row whose ``t_values[row] < step`... | {
"id": "47c3dda8-ea08-4a5e-a779-613afb9e4924",
"created_at": "2026-08-19 22:31:36.489046+00:00",
"db_md5": "ea674820dce43683fa9b0d6a606bab09",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "c62228b9-6af3-4e06-a82e-e04b95931fbb",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 10.04990291595459,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008750000270083547
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 11.687721252441406,
"example_count": 300... |
47c4785a-a1d3-4fae-aa2a-5cc5eff71e1d | easy | ehonig | 2026-08-20 10:06:08.408156+00:00 | succeeded | 59c0d05192cbf8d6ec53bf121bde8656c14fef1cc4ec27183faa5bc245f4393d | 24,999 | null | """A plain Transformer for repeated modular squaring, and nothing else yet.
This is a deliberate restart. The previous submission accumulated a field
parser, a carry scan, a learned reciprocal, periodic features, a digit
bottleneck, a depth selector and two cell types, most of them tuned around a
memorisation table th... | {
"id": "47c4785a-a1d3-4fae-aa2a-5cc5eff71e1d",
"created_at": "2026-08-20 10:06:08.408156+00:00",
"db_md5": "3014552377a7c60e23fdbe110d4918cd",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "7d27a99b-91d0-4734-9f74-0dc59d78f248",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 1.8708087801933289,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.026666666381061077
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.822935700416565,
"example_count": 100... |
47c5b2fc-f62c-4b36-af33-2d43d8dfd032 | easy | DDanlov | 2026-08-15 14:27:48.897878+00:00 | succeeded | ba423828fd0987a5025905915a20d4fcf15e7a8dcc87eaaaf57a1a088dbeadc6 | 19,086 | null | """
Official Submission for One Layer Deeper Challenge
Architecture: Pure No-Norm, No-Skip DKS Recurrent Core with Double Embedding (dim=512),
Input RMS Normalization only, 64-Channel Harmonic+Dyadic Torus, Power-of-10 Alignment, and RohanShampoo.
"""
from __future__ import annotations
import math
from typing import O... | {
"id": "47c5b2fc-f62c-4b36-af33-2d43d8dfd032",
"created_at": "2026-08-15 14:27:48.897878+00:00",
"db_md5": "511dfeca7884fd8044024a2c27f2057f",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "a93d3cbc-5db3-44df-b361-5e3b9fe10f94",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 8.515925790319649,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04500000017384688
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.581276406103106,
"example_count": 100,
... |
47c695cb-e37b-4ef5-a37c-8b5ce5346fae | easy | khushidahi | 2026-08-20 04:10:29.692050+00:00 | succeeded | 115e416ffc846c0991d15df2f9b94779e8f9c9436d47bc032239d98d0f19416c | 34,872 | null | """R305: serial-product semantic-factorized functional root.
This candidate is deliberately diagnostic. It trains on evaluator-owned final
labels at several depths so that one shared transition must compose. The model
never converts N or X to native numbers and never implements multiplication or
modular reduction.
... | {
"id": "47c695cb-e37b-4ef5-a37c-8b5ce5346fae",
"created_at": "2026-08-20 04:10:29.692050+00:00",
"db_md5": "9e83f60db4f3411321f3bdfe1cd2e318",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "cc439a0b-f0f9-43f6-adb2-a61a263a9f5f",
"tier": "easy",
"dataset_id": "e3",
"status": "succ... | {
"score": {
"mean_loss": 2.2415287664254393,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0056250000605359675
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2249080458758606,
"example_count": 8... |
47c7085a-1741-4d15-a955-1af935b8d2b1 | easy | TZZheng | 2026-08-06 17:36:56.253761+00:00 | succeeded | 2fad6b4e7cd8cea1bc39733cdefe9820ee5cc589a8e69804d8b1ff041e2cec0c | 7,301 | null | """Self-contained standard Transformer scaling-sweep submission."""
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,
)
VARIANT = {'block... | {
"id": "47c7085a-1741-4d15-a955-1af935b8d2b1",
"created_at": "2026-08-06 17:36:56.253761+00:00",
"db_md5": "f3ad307516b4651a9f686c1ab6ab0b3f",
"submitter": "TZZheng",
"github_login": "TZZheng",
"run_id": "0457058b-6bed-448a-ba27-22f36dafb5b4",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 9.108155250549316,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01833333307877183
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.903218269348145,
"example_count": 100,
... |
47c8f261-3d33-4934-82c2-a27086217d93 | easy | chuk-uzowihe | 2026-08-09 20:51:51.890380+00:00 | succeeded | 183a8ba5a5db2f1ba4c5eee1f64b2da4440a1dbeed178214484f539061691759 | 38,838 | 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": "47c8f261-3d33-4934-82c2-a27086217d93",
"created_at": "2026-08-09 20:51:51.890380+00:00",
"db_md5": "f09f573d27cd9907357bc35ee69dd108",
"submitter": "Chuk Uzowihe",
"github_login": "chuk-uzowihe",
"run_id": "c26577c2-87de-4408-9293-7375d431969b",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 5.431775331497192,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06666666641831398
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.040829658508301,
"example_count": 100,
... |
47cafbeb-817d-41a2-89f3-e4fade8e638b | easy | sirish-gambhira | 2026-08-10 00:51:44.889719+00:00 | succeeded | b45c3a7c939775427efda0073a2e3f248c0456549c013236b1fe60df5e023870 | 7,521 | null | """Decimal-structure-aware bidirectional Transformer for token arithmetic."""
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_mode... | {
"id": "47cafbeb-817d-41a2-89f3-e4fade8e638b",
"created_at": "2026-08-10 00:51:44.889719+00:00",
"db_md5": "72425dc98b937d28806a502c247ab46a",
"submitter": "Sirish Gambhira",
"github_login": "sirish-gambhira",
"run_id": "89eeb9e5-0c09-4955-84c5-e05832165846",
"tier": "easy",
"dataset_id": "e5",
"stat... | {
"score": {
"mean_loss": 6.034815788269043,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333507180214
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.965572834014893,
"example_count": 600,... |
47cf18c6-6be0-47f1-80b5-672a95bd6e71 | easy | isaac0804 | 2026-08-06 15:22:47.512731+00:00 | succeeded | fd2a65396b3e5fdd633fe62ee304e59279201d7144801e7755e44d61b1accf92 | 6,603 | null | """E3 candidate #6: narrow continuous information bottleneck (128 -> 8 -> 128
squeeze), mid-stack, on every token position -- fixes a capacity-escape flaw
found in e3_vq_bottleneck.
e3_vq_bottleneck quantized every one of ~13 sequence positions independently
against a 64-entry codebook. That is not a real bottleneck: ... | {
"id": "47cf18c6-6be0-47f1-80b5-672a95bd6e71",
"created_at": "2026-08-06 15:22:47.512731+00:00",
"db_md5": "5f3039987f102e4036603b4a74dfe9eb",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "b90a2ad3-4a2d-490e-ad1f-7fdf8aa56b3a",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.2096208148185372,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008125000153668224
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2597025540078644,
"example_count": 80... |
47d31f00-0bdf-47a9-ae53-7471991a9016 | easy | shirvani-jr | 2026-08-31 02:18:20.972857+00:00 | succeeded | 13a1a49b45b9bd9caa23db4129202ddb3f944ec16317acc5803307c0d2e8885e | 20,295 | 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 = 3
PERIOD_MAXIMUM = 100
EX... | {
"id": "47d31f00-0bdf-47a9-ae53-7471991a9016",
"created_at": "2026-08-31 02:18:20.972857+00:00",
"db_md5": "3ead2625f5baf79031af4dabfaa47097",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "5fd0ab11-3542-4cef-b207-9c0e79c76f5d",
"tier": "easy",
"dataset_id": "e9",
"status": "succeeded"... | {
"score": {
"mean_loss": 1.9117134809494019,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.027777778916060925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9281668663024902,
"example_count": 90... |
47d602e1-79c2-4ae7-90d8-37131d20d337 | easy | Bananafly | 2026-08-11 18:28:04.000579+00:00 | succeeded | 94d832edca3d2bc0c0d39fdcdf0002d6ebc2729786b2873de1fb5b7758f22fbf | 10,217 | null | """Gated recurrent Transformer with latent scratch space."""
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,
ass... | {
"id": "47d602e1-79c2-4ae7-90d8-37131d20d337",
"created_at": "2026-08-11 18:28:04.000579+00:00",
"db_md5": "aee0d19c2285d65bb77413a5fa55a42a",
"submitter": "Andre Kreidemann",
"github_login": "Bananafly",
"run_id": "f4210dd0-a580-4be9-b6f6-b580b36f005f",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 3.19152569770813,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.014999999664723873
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.053365707397461,
"example_count": 100,
... |
47e28da3-2481-49c1-a623-4a7dd9e794d9 | easy | gauravmishra | 2026-08-09 22:02:20.909692+00:00 | succeeded | b0979291365095e306a4f792207dd9c45c6c4b4543405555f7cfab816652b42f | 4,249 | null | """Basic single-pass Transformer with PyTorch AdamW."""
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_state,
)
D_MODEL = ... | {
"id": "47e28da3-2481-49c1-a623-4a7dd9e794d9",
"created_at": "2026-08-09 22:02:20.909692+00:00",
"db_md5": "ff141e9d63f34f2c2e34713a993447aa",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "4f7f42bd-2845-45c0-9ee9-125e57a90c82",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 3.765537404166099,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.381876496456129,
"example_count": 600,... |
47e6a849-0e08-4835-bfec-617d2d7ec0f5 | easy | ehonig | 2026-08-13 22:44:19.302959+00:00 | succeeded | 1494e9e05a05eb88051105bb4d678c8c98382139bb115c1fe8b37e5adaa927cb | 44,391 | null | """DigitLoop — a place-value recurrent model for repeated modular squaring.
The task is ``y = x^(2^T) mod N`` with the answer read back tail aligned, one
decimal place value per token position from the right. A stack of attention and
MLP layers has no primitive for the two operations this needs — a partial-product
co... | {
"id": "47e6a849-0e08-4835-bfec-617d2d7ec0f5",
"created_at": "2026-08-13 22:44:19.302959+00:00",
"db_md5": "c9873abd510454aa95936aa69a587508",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "a9b4b4c6-dc24-4151-bdb1-b9554a805132",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 4.919789791107178,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0716666653752327
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.239845275878906,
"example_count": 100,
... |
47ef3d84-e1b0-4400-b1e2-f701a908cdf2 | easy | oupadhyay | 2026-08-28 02:18:03.093394+00:00 | succeeded | 73045a9d2a6eb36d5478b8fefaef7789792d16753dcf244d9535c98a81a3e872 | 4,317 | null | """Conditional structured-loss phase 5: hard35."""
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=4,64,32
class C:
def __init__(s,vocab_size,max... | {
"id": "47ef3d84-e1b0-4400-b1e2-f701a908cdf2",
"created_at": "2026-08-28 02:18:03.093394+00:00",
"db_md5": "44606caf57007602b7f46df9a71ce51e",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "d4c55e9c-8069-457f-b655-56e737ef19c9",
"tier": "easy",
"dataset_id": "e6",
"status": "s... | {
"score": {
"mean_loss": 1.8842089176177979,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.11869369633495808
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.7483017444610596,
"example_count": 60,... |
47fa77a9-bc7d-42ad-94f1-f64822d4afe3 | easy | DDanlov | 2026-08-22 22:00:47.706123+00:00 | succeeded | d63f531ca3673e316979da578d66d2cccbcce9d1f5a832a2a06d40ac7175438d | 26,884 | null | """
Unified Hybrid Masked Deliberation Model (75.8M Parameters) - Variant 2 (Pure Skipless, No Cross-Attention)
Architecture:
- Scale: dim=1536, num_heads=24, d_ff=6144 (Exactly 75,826,176 parameters)
- Hybrid Prefix-Bidirectional + Causal Target Attention Mask
- Token Layout: [Prompt (L_p), 4 Pause Registers, Target T... | {
"id": "47fa77a9-bc7d-42ad-94f1-f64822d4afe3",
"created_at": "2026-08-22 22:00:47.706123+00:00",
"db_md5": "ad79263cbcb88cc085818a3b6a5414f8",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "b92b6f59-0b15-4c12-9a81-f4028fac96ea",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.8332136022346512,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8332135677337646,
"example_count": 600,
"exa... |
4801bb74-e928-4de3-ac4a-b9e4b35f2a52 | easy | erdavis0 | 2026-08-19 10:01:52.494529+00:00 | succeeded | 3a470214c54e81a6fdbb5d2db7b891c47f79af6467c0efdd358d5f25ff77111a | 10,596 | null | """Learned-slot low-rank operator with calibrated endpoint supervision."""
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,
TokenLoss... | {
"id": "4801bb74-e928-4de3-ac4a-b9e4b35f2a52",
"created_at": "2026-08-19 10:01:52.494529+00:00",
"db_md5": "95e07760f3e09e3319abe1e605d1375c",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "da571b51-94bf-4c33-96d2-d674390e5142",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.0426560044288635,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.053603606298565865
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9877299070358276,
"example_count": 60... |
48031b1f-54cd-426d-a28d-844ea2890e59 | easy | DDanlov | 2026-08-29 11:44:57.282328+00:00 | succeeded | 49347fbc8b4d617e593f74ef95750161060c4492f2d5fa63a821beabffca6e0e | 15,094 | null | """
submission_eqprop_deep_e5.py
Modernized 2-Pass Continuous Hopfield Energy Deliberator (adapted from eqprop_deep)
for the One Layer Deeper API (Tier Easy, Dataset E5).
Mathematical Principles:
1. Input Representation: Task-agnostic token + sinusoidal positional embeddings with learned mask placeholder.
2. Continuou... | {
"id": "48031b1f-54cd-426d-a28d-844ea2890e59",
"created_at": "2026-08-29 11:44:57.282328+00:00",
"db_md5": "62bf27eb65b1c608ef4de58689d71a24",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "1e6850f6-3880-4639-8ea0-64d4e9e34935",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.2451939113159405,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333333333335
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2487411861195157,
"example_count": 6... |
4816d86a-d0c2-4666-8e5d-3e853e052012 | easy | sapient-sapiens | 2026-08-31 02:43:46.798136+00:00 | succeeded | d00e44a4b1f614417b8b0c6848a88215c27a1d2a0346c9524fced7a634cc446b | 17,770 | null | """Eligible one-call direct-logit E5 operator with a small residual init.
This is the matched direct-centered-logit sibling whose only optimization
change is initializing the residual-head weight with standard deviation
1e-3 instead of exactly zero. One literal pre-norm Transformer block is
reused at all 4 x 8 Horner... | {
"id": "4816d86a-d0c2-4666-8e5d-3e853e052012",
"created_at": "2026-08-31 02:43:46.798136+00:00",
"db_md5": "45dc58e5bb9444f2c789f1a919128fd9",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "07a5e86d-e08e-495f-98f0-f85ca1f0f91a",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.163569024608875,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000037252903
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.159581711474021,
"example_count": 600,... |
48172e80-df18-4050-9618-d7a3bd78823e | easy | jordanrubin | 2026-08-12 07:10:26.233536+00:00 | succeeded | ab34a11198c302dd2edc317b1b463ccffb20bd9019d78abed718f3a88cc5d938 | 24,918 | 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": "48172e80-df18-4050-9618-d7a3bd78823e",
"created_at": "2026-08-12 07:10:26.233536+00:00",
"db_md5": "19d38540224b58e35ee3887ac55c96ea",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "12c481b2-ce3e-49b8-8a0c-476516f3d7e0",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.8360050231337532,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02708333257275323
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2104182250029303,
"example_count": 600... |
4819eb62-2c60-4fa1-822b-917ca1043b38 | easy | khushidahi | 2026-08-21 20:14:08.798911+00:00 | succeeded | 8c58e9bb5dd1eefa603f6d989dd843d1a6fc5a7ad150abe5e7fbd393dfe23610 | 29,817 | null | """R352: final-only-loss K8 variable-modulus control.
This submission tests the strongest rules-safe generic baseline suggested by
the public competition discussion. It does not implement squaring, modular
reduction, or any other task solver. Numeric prompt fields are only rearranged
onto equal-length, right-aligned... | {
"id": "4819eb62-2c60-4fa1-822b-917ca1043b38",
"created_at": "2026-08-21 20:14:08.798911+00:00",
"db_md5": "631fb14433cede38a273750a82c4954d",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "19821b81-df32-4683-be84-c74198b60482",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.2815137097115903,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0037500000124176345
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2793815125264394,
"example_count": 6... |
481cbbfa-96d0-48a4-9b21-590082141211 | easy | khushidahi | 2026-08-25 13:00:55.230699+00:00 | succeeded | 8c0c1b6f7e351df3922fa74773e6bb838b661745036e913e98537e703af66351 | 26,597 | null | """CSMM P0 implementation reconciled to CSMM_EXECUTABLE_SPEC.md.
This file implements P0-Adam. N compiles a generic eight-slot microprogram.
X initializes the tape. T exports only one scalar monotone apply/hold budget.
"""
from __future__ import annotations
import math
import time
import torch
import torch.nn.funct... | {
"id": "481cbbfa-96d0-48a4-9b21-590082141211",
"created_at": "2026-08-25 13:00:55.230699+00:00",
"db_md5": "e668b2017e39c10c6d29e622a764479b",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "76d60dab-cf62-4e3b-bdf2-13b35e8d65cd",
"tier": "easy",
"dataset_id": "e3",
"status": "succ... | {
"score": {
"mean_loss": 2.236435138422477,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016250000046566127
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2312779175818944,
"example_count": 800... |
48285e72-5a96-4db8-bd3d-d23ab2709086 | easy | khushidahi | 2026-08-03 22:40:50.774201+00:00 | succeeded | d7797656025009174779ba680cbbb40b4f64bf5f8a953be639300ef590507548 | 15,555 | 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": "48285e72-5a96-4db8-bd3d-d23ab2709086",
"created_at": "2026-08-03 22:40:50.774201+00:00",
"db_md5": "2d746f6a3ef1f5cd9a13e51f82744c29",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "2bd48d18-0948-4a88-a5fa-4da84d26df71",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.197646379470825,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333302289248
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.193133592605591,
"example_count": 600,... |
482cf6cc-58c6-40f8-b767-bd1b9df626f5 | easy | karanganesan | 2026-08-07 08:27:03.293000+00:00 | succeeded | b1c604f828e446c90aa6b77faa4aa7808a3bccad4870a7d0c6aabb800b1c6cfe | 25,600 | null | """Parametric looped-transformer family (P1).
One weight-tied transformer block applied k times in latent space. Config
flags cover four P1 families with one file:
looped recall=0 gated=0 tfilm=0 plain weight-tied loop
looped-recall recall=1 re-inject the input embedding each
... | {
"id": "482cf6cc-58c6-40f8-b767-bd1b9df626f5",
"created_at": "2026-08-07 08:27:03.293000+00:00",
"db_md5": "5b7809258b635ad8bba9e679ea5c94c0",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "7d04afdd-4e56-44f6-8a7f-1865ea59bcf7",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 3.822319507598877,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0054166666232049465
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.3480334281921387,
"example_count": 60... |
4836b541-66e3-44ba-b061-778c77d34b5c | easy | DESU-CLUB | 2026-08-30 19:12:00.701534+00:00 | succeeded | 491b2aca7c985d854d715854c509d68ee45805d19929ee5a337784132638dc4f | 15,215 | null | """Discrete-state loop over a place-value tape, trained with orthogonalize-only Muon.
The recurrent state carried between squarings is the *digit string*, not a
residual stream. Each outer iteration reads (N, x) off an LSB-first place-value
tape, emits digit logits for one squaring, hard-argmaxes them (straight-throu... | {
"id": "4836b541-66e3-44ba-b061-778c77d34b5c",
"created_at": "2026-08-30 19:12:00.701534+00:00",
"db_md5": "072a34ff9b8764ce77ef729bda902ca4",
"submitter": "Low Keng Hoong, Warren",
"github_login": "DESU-CLUB",
"run_id": "b2795909-5954-4109-9c9d-59d55a9a507b",
"tier": "easy",
"dataset_id": "e6",
"sta... | {
"score": {
"mean_loss": 1.1078858081608602,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.6889639770662462
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.7938528060913086,
"example_count": 60,
... |
483efa8e-2e6d-4ec9-8715-0a29a7990b35 | easy | int3 | 2026-08-18 10:40:32.221581+00:00 | succeeded | 96af75d5cfaf9658670ade7afa50c1dc6f64a4f25caf46a60e307889ff28cfcc | 6,983 | null | """Tied-weight recurrent transformer: iterate a learned squaring step T times.
Family A (depth extrapolation by construction): an encoder parses the prompt,
one weight-tied core block is applied T times (T read from the prompt's digit
tokens, per-example halting masks keep everything batched), and a decoder emits
the ... | {
"id": "483efa8e-2e6d-4ec9-8715-0a29a7990b35",
"created_at": "2026-08-18 10:40:32.221581+00:00",
"db_md5": "8bd840de53cbfd580a69b830672e79d9",
"submitter": "Jez Ng",
"github_login": "int3",
"run_id": "752f6e2e-f3a9-4bcf-86a8-d87211bb9ca2",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
... | {
"score": {
"mean_loss": 6.634246110916138,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0283333333209157
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.298418045043945,
"example_count": 100,
... |
48451b59-1c38-4606-bc05-084e5bd6a523 | easy | AFeinfield | 2026-08-12 23:41:58.883320+00:00 | succeeded | 206a0daa6ab4b756595800b8f19ef1330be2aae13154d855d4c58ed099d01eeb | 4,661 | null | """Looped (weight-tied) bidirectional Transformer, d=256, 8 iterations.
002: same as 001 but LR schedule horizon matched to measured ~260 server steps.
prior. Same parameter count order as a 1-block model; depth comes from reuse.
Warmup + cosine LR. No compile (60s budget).
"""
from __future__ import annotations
imp... | {
"id": "48451b59-1c38-4606-bc05-084e5bd6a523",
"created_at": "2026-08-12 23:41:58.883320+00:00",
"db_md5": "138d05c8519c6c96602f0a09e85e6d52",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "99e61690-607a-4f9a-a502-fb5f79d04ba6",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 1.9222792387008667,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.023333333432674408
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8817614316940308,
"example_count": 10... |
484e2bb1-08e5-4c0a-b360-2d4eb8f2281b | easy | adambrzosko | 2026-08-30 08:49:50.929111+00:00 | succeeded | 5ea1d3951e5ceb4c2d47ebad1b3f486b510d1b0b2d22108ffc16805a136d8df8 | 10,793 | null | """One Layer Deeper — is model size free?
Batch held at 128 (645 steps last time), model ~8x larger. If steps stay
near 645, the per-example cost is not our compute and the 500M state
budget is available for free. Read completed_training_steps, not score.
Only BATCH_SIZE differs across the three probe variants. MAX_S... | {
"id": "484e2bb1-08e5-4c0a-b360-2d4eb8f2281b",
"created_at": "2026-08-30 08:49:50.929111+00:00",
"db_md5": "a9f0da203086da441c99d4f88c395fae",
"submitter": "Adam Brzosko",
"github_login": "adambrzosko",
"run_id": "4765f9ba-7de3-4099-ac1d-69d00d49090f",
"tier": "easy",
"dataset_id": "e6",
"status": "s... | {
"score": {
"mean_loss": 2.161551594734192,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013513513840734959
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.216083526611328,
"example_count": 60,
... |
48520991-8c37-4964-b204-9d2e384d1255 | easy | aditya-ramabadran | 2026-08-28 23:51:50.700382+00:00 | succeeded | c47dc9ed0146073f1cca66a293b50f5b48bef7bdad73cc2b9f94fcd4473f5885 | 22,495 | null | """Anonymous phase host over a right-aligned token tape.
Contest-contract port of the strongest published direct modular-squaring
architecture in this lab: a small set of output-aligned slots, anonymous
untied phases, content-*independent* learned positional routing, plain
residuals, a deliberately weak digit head, cr... | {
"id": "48520991-8c37-4964-b204-9d2e384d1255",
"created_at": "2026-08-28 23:51:50.700382+00:00",
"db_md5": "5e7e19ebe82ae7e5a7e36f97cc55973f",
"submitter": "aditya test",
"github_login": "aditya-ramabadran",
"run_id": "82acd13e-4f3f-4199-b43e-15e6243befe9",
"tier": "easy",
"dataset_id": "e1",
"status... | {
"score": {
"mean_loss": 2.825440287590027,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.029999999329447746
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.983518123626709,
"example_count": 100,... |
485c1a2a-2905-4f93-be3f-8c4fa957e2a3 | easy | Avyukth | 2026-08-12 20:04:01.587882+00:00 | succeeded | d4307e8f2da278d8c0944bddc54ae2ee8d29e381b27d0de6769af82d63813193 | 24,266 | null | """T-503 — full T-dependent torsion-orbit augmentation for Easy e1.
For N=323=17*19, the unit multipliers satisfying u^(2^T)=1 form kernels of
size 4, 8, and 16 at T=1,2,3. Multiplying x by any row-valid u preserves the
repeated-squaring target. During training this arm samples the appropriate
kernel uniformly and w... | {
"id": "485c1a2a-2905-4f93-be3f-8c4fa957e2a3",
"created_at": "2026-08-12 20:04:01.587882+00:00",
"db_md5": "aeb45f9e03beeef047bc4bf14d8bec22",
"submitter": "Subhrajit Makur",
"github_login": "Avyukth",
"run_id": "ffcc0873-b1fe-474b-ba04-9c88c4b0d335",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 5.107313930988312,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.35000000335276127
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.76716136932373,
"example_count": 100,
... |
486c86af-e71e-4dd9-8eb8-75d499a718aa | easy | jordanrubin | 2026-08-03 20:55:56.568342+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": "486c86af-e71e-4dd9-8eb8-75d499a718aa",
"created_at": "2026-08-03 20:55:56.568342+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "a515693b-9cf5-4d2c-81c8-ebd6ca1e83f1",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.442143583858343,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.20229166156301895
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.43593342157468,
"example_count": 300,
... |
486febeb-6937-4d85-aa0d-527f1c99858a | easy | sapient-sapiens | 2026-08-15 20:54:20.690585+00:00 | succeeded | 469ae48173121f8a328b5c1f44d0bff0e2f098d243dcd227975cee12399b6829 | 17,772 | null | """T-only adaptive chunk routing over a learned shared operator.
The controller sees only the categorical tokens inside the T prompt field and
is trained solely by final-answer loss. It never parses T numerically and T is
not used as an intermediate supervision target. Inputs use learned additive
token, semantic-gro... | {
"id": "486febeb-6937-4d85-aa0d-527f1c99858a",
"created_at": "2026-08-15 20:54:20.690585+00:00",
"db_md5": "38cda1d694d318221e86f82dd23bbc39",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "8b4dd3d8-cc73-462d-b0e3-736ca909657a",
"tier": "easy",
"dataset_id": "e4",
"status": "su... | {
"score": {
"mean_loss": 2.0651984426573913,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0034375
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.06712847386775,
"example_count": 1200,
... |
48700067-9391-4f84-9365-aeabb83a702e | easy | benjaminW2025 | 2026-08-18 00:22:24.703356+00:00 | succeeded | 2338d96efce40bad07eb6769ac2d78ae5b2770c4be287ab2ec2b6e121b661bd7 | 11,189 | null | """Basic single-pass Transformer with PyTorch AdamW."""
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 = 256
NUM_HEADS = 4
... | {
"id": "48700067-9391-4f84-9365-aeabb83a702e",
"created_at": "2026-08-18 00:22:24.703356+00:00",
"db_md5": "444137f56ceb89263359b6884e84fce9",
"submitter": "benjawesome",
"github_login": "benjaminW2025",
"run_id": "a2131f5e-0981-4654-acb3-28956a1d6783",
"tier": "easy",
"dataset_id": "e5",
"status": "... | {
"score": {
"mean_loss": 2.2482531374880317,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2411023479940644,
"example_count": 600,... |
48785645-44e0-44c4-a0ab-4a3679aa5593 | easy | Yalyenea | 2026-08-11 20:38:22.731827+00:00 | succeeded | 6d78943e5f57f580ac7c177a353da4fb6a5993d04baab0720705b6f27c87e5b7 | 12,919 | null | """R4 full-bandwidth candidate with RoPE and a 2000-step cosine horizon."""
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,
... | {
"id": "48785645-44e0-44c4-a0ab-4a3679aa5593",
"created_at": "2026-08-11 20:38:22.731827+00:00",
"db_md5": "e1ab4883f30066856bcfaaad405fd185",
"submitter": "yfff",
"github_login": "Yalyenea",
"run_id": "6a22d5ac-f64a-4d48-bbcb-199ec619cc89",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",
... | {
"score": {
"mean_loss": 8.594528198242188,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011249999981373549
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.51823616027832,
"example_count": 800,
... |
487d4fd5-9afb-49e9-a7b8-a124d416cd2d | easy | int3 | 2026-08-18 12:33:57.695582+00:00 | succeeded | 4f80168e3c089d7e12cb79b1e0f5c9323080a7f2389a4b9224826a226532cf86 | 12,185 | null | """Automata bank with bilinear pair-table readout and deep state supervision.
Architecture (all predictive computation in learned weights):
- The prompt's N / X / T decimal fields are parsed to integers (discrete index
math, the same class as T-conditioned loop counts / adaptive computation).
- State: one distributi... | {
"id": "487d4fd5-9afb-49e9-a7b8-a124d416cd2d",
"created_at": "2026-08-18 12:33:57.695582+00:00",
"db_md5": "5dcf7664ce49ce9468c223c1ee009ae3",
"submitter": "Jez Ng",
"github_login": "int3",
"run_id": "4e4ebe20-b428-4756-850b-aaaac9471a33",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.1019519524836916,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04500000002483527
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.109004343690478,
"example_count": 1200... |
487eb726-4ed4-4baf-b253-7a59f0095191 | easy | erdavis0 | 2026-08-27 02:23:39.487890+00:00 | succeeded | dea5a50d1a7e1542b606c505bb2e064069826b4fcc486a84827b8cbefc9b781f | 31,293 | null | """Typed-prefix reversible transition B0: ordinary evaluator token CE.
Public delimiter syntax routes opaque N, X, and T fields into three strict
functional roles. X initializes only an evolving auxiliary tape, N initializes
only an invariant environment lane, and T reaches only a learned monotone
prefix controller. ... | {
"id": "487eb726-4ed4-4baf-b253-7a59f0095191",
"created_at": "2026-08-27 02:23:39.487890+00:00",
"db_md5": "9a82ec9520ce6d08ad83d5dbf700b0ab",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "f62aefb1-40ab-4a52-88aa-9813ff18eea3",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",... | {
"score": {
"mean_loss": 5.310205698013306,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06500000134110451
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.450310707092285,
"example_count": 100,
... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.