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 |
|---|---|---|---|---|---|---|---|---|---|---|
2059aa5f-0907-412a-b939-8bdf4e841f72 | easy | Bananafly | 2026-08-12 09:32:46.924121+00:00 | succeeded | ec4ea8144c23c88eeaf8ed008629328243d3545a8b9678cc088f7aa471ae81af | 11,652 | null | """Easy-3 T-aligned soft residue model with a 6e-3 learning rate."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_M... | {
"id": "2059aa5f-0907-412a-b939-8bdf4e841f72",
"created_at": "2026-08-12 09:32:46.924121+00:00",
"db_md5": "59974190070b9d0e97510d824f5d266e",
"submitter": "Andre Kreidemann",
"github_login": "Bananafly",
"run_id": "4ea30f8f-ca64-4be5-81bd-9b52a68a767b",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.215923547744751,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2290353775024414,
"example_count": 100,
... |
20640b7d-0544-43d3-8806-9778b6614fa2 | easy | Maxwells-Demons | 2026-08-13 14:56:56.481355+00:00 | succeeded | 83e7bdc272d805616280b5d877e464aaad5a698ddde6d3efaf572498675ee081 | 120,067 | null | """One shared digit operator, its two free neighbours of x, and two learned maps.
The backbone is unchanged from v47/v96: the state IS the digit one-hot, one
application of the operator is
(state digits, N digits) -> next state digit logits
and the straight-through code of those logits is the whole state handed to... | {
"id": "20640b7d-0544-43d3-8806-9778b6614fa2",
"created_at": "2026-08-13 14:56:56.481355+00:00",
"db_md5": "ccf7080430ccb5455ebe511d5f88fefd",
"submitter": "Maxwells_Ayakashi",
"github_login": "Maxwells-Demons",
"run_id": "e6e63710-015c-4a37-8a59-aa62d07de598",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 13.183456714669626,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.5087500002980232
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 12.408450062499453,
"example_count": 600,... |
206a23f7-8cce-40d6-9915-5643c9b2d456 | easy | viridale | 2026-08-04 23:12:38.399864+00:00 | succeeded | 7b905ffd5ec9945fb9b75255e3eba273ea11e721b6647779f23933421f8adedf | 8,692 | null | """Clean autonomous recurrent Transformer with pure rotary positions.
hypothesis: v174's learned absolute position vectors support row-layout memorization;
applying generic rotary relative geometry to every attention call while removing
absolute vectors may preserve order but make the tied computation shift-st... | {
"id": "206a23f7-8cce-40d6-9915-5643c9b2d456",
"created_at": "2026-08-04 23:12:38.399864+00:00",
"db_md5": "848c289fe7646d8524859752ebf69818",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "dd0e88d7-e799-409d-baec-ac3df718796d",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.283447583224682,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018125000000000002
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.286914264548963,
"example_count": 800,... |
206ceb37-8d22-4280-ab8a-275d2cda84c6 | easy | erdavis0 | 2026-08-26 04:19:06.933899+00:00 | succeeded | b063dd87e5c8ed86565b7dc9bb9038c969fb662780579ead5b62238da52e334a | 30,346 | null | """Latent-role continuous REV control C0: ordinary evaluator token CE.
Public delimiter syntax only routes opaque token categories into distinct,
right-aligned N, X, and T tapes. Learned local cells compile separate
interaction, carry, and modulus-conditioned workspaces. N is isolated from
interaction and carry and ... | {
"id": "206ceb37-8d22-4280-ab8a-275d2cda84c6",
"created_at": "2026-08-26 04:19:06.933899+00:00",
"db_md5": "b092954054ec5242242d7829dc8330c1",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "39fa0de2-9d3c-404d-b6c1-84fa2070652e",
"tier": "easy",
"dataset_id": "e8",
"status": "succeeded",... | {
"score": {
"mean_loss": 4.774971961975098,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06729055382311344
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.179647922515869,
"example_count": 85,
... |
206d27b2-0689-4ffe-95a0-5c78897df13c | easy | Amr-Hegazy1 | 2026-08-03 22:08:24.901189+00:00 | succeeded | 23d3eac6e8359ca0592a721183b6c2cbbc90fd81f51b7076ae656b35a8600120 | 17,919 | null | """RecurrentGPT: prelude / gated shared recurrent core / coda.
Port of the recurrentGPT architecture to the One Layer Deeper contract.
The model is a depth-recurrent transformer:
* a fixed *prelude* stack embeds the prompt and produces ``head_output``;
* a *shared core* stack is applied ``r`` times. Every step ... | {
"id": "206d27b2-0689-4ffe-95a0-5c78897df13c",
"created_at": "2026-08-03 22:08:24.901189+00:00",
"db_md5": "225a7eaced136a47a69a5ea3a332cbca",
"submitter": "Amr Hegazy",
"github_login": "Amr-Hegazy1",
"run_id": "dbff3a5c-b68f-4ef4-852d-29b6721a4ac0",
"tier": "easy",
"dataset_id": "e3",
"status": "suc... | {
"score": {
"mean_loss": 6.743612324268586,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000009313225
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.676800127234036,
"example_count": 800,... |
206e10a5-01fc-41b4-9189-2be8d5ed9afb | easy | jordanrubin | 2026-08-05 18:22:10.129470+00:00 | failed | f4ac2ec8c82398633cef45114d3c64a0d7472cc2033a11ff994c128413d030b9 | 14,403 | 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": "206e10a5-01fc-41b4-9189-2be8d5ed9afb",
"created_at": "2026-08-05 18:22:10.129470+00:00",
"db_md5": "2225171dc05e63c7384b3e4291912986",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "5c76db49-5ec5-4aaa-84c1-e038fb386c5b",
"tier": "easy",
"dataset_id": "e2",
"status": "f... | null
|
206f19eb-1fd7-4fda-8435-979ae4c741c4 | easy | sapient-sapiens | 2026-08-16 19:53:04.721404+00:00 | succeeded | 8cdfe1e547c11be8b2427b0ba999f984ee1145f6f8673599044d984a0e395683 | 21,874 | null | """E6 SAM-family variant: frozen v55 with a one-variable SAM change.
Architecture, representation, HybridMuon split, rho=0.02, batch, wall-clock
schedule, and loss are unchanged from act64_e6_w1024_dh4_L6_v55. T is
diagnostic-only.
"""
from __future__ import annotations
import math
import time
import torch
import t... | {
"id": "206f19eb-1fd7-4fda-8435-979ae4c741c4",
"created_at": "2026-08-16 19:53:04.721404+00:00",
"db_md5": "a3c477e88181772c69603a2230c0e2f2",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "0f5f86e3-46ce-4e7b-86cf-cc098bdb3970",
"tier": "easy",
"dataset_id": "e6",
"status": "su... | {
"score": {
"mean_loss": 7.734316058354835,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.10518018221734343
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.708580493927002,
"example_count": 60,
... |
20735f59-812c-4461-b218-bd1672ed2143 | easy | shirvani-jr | 2026-08-22 13:18:02.514370+00:00 | succeeded | 1398bd08847c5eaab4959d387d99f9b6d5d587478909c52a6d433aa99a043c89 | 23,140 | null | """Looped exact-arithmetic digit machine.
The recurrent state is the decimal digit string of the current residue. One
loop application performs one modular squaring, so depth extrapolates in T by
construction. Every functional primitive — digit-by-digit product, full
adder, full subtractor, digit comparator, and com... | {
"id": "20735f59-812c-4461-b218-bd1672ed2143",
"created_at": "2026-08-22 13:18:02.514370+00:00",
"db_md5": "a1310d423bb5360fa56d444443de24fa",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "fbb9a294-0414-4bb4-9730-6a865123dfc6",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded"... | {
"score": {
"mean_loss": 0.04981634808150272,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.04983068424252413,
"example_count": 1200,
"... |
2076ca5d-cb60-48c1-9c27-a42d24b9f15a | easy | ehonig | 2026-08-14 01:06:10.304957+00:00 | succeeded | 6cc7f0ea1126c8f8a210569bdbe0d322f6256813c1a3e5cecb857c62b7684268 | 54,754 | 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": "2076ca5d-cb60-48c1-9c27-a42d24b9f15a",
"created_at": "2026-08-14 01:06:10.304957+00:00",
"db_md5": "5e22f84b86ad97e9c17d31e7481f6320",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "eeee08e5-9a54-4902-9341-cf88af4b680a",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 5.826894760131836,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.17000000923871994
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.4493021965026855,
"example_count": 100,... |
20b04701-39e8-42a4-91f6-353f9e2ea6da | easy | DDanlov | 2026-08-09 00:42:34.980725+00:00 | succeeded | 4e181a546b272b03db7d768867b588745819c557661877e5058817e28e645838 | 15,978 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with eigenvalue ma... | {
"id": "20b04701-39e8-42a4-91f6-353f9e2ea6da",
"created_at": "2026-08-09 00:42:34.980725+00:00",
"db_md5": "c486af7547a7d3524e0adac485073cb8",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "529dd032-a68f-43d1-9c64-1c0121da0193",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 41.54358922151759,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333333333335
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 46.67870185407288,
"example_count": 600... |
20b68470-6775-436b-83f3-76761379a65e | easy | khushidahi | 2026-08-03 20:47:30.956716+00:00 | succeeded | d4b36f63c940ed51e8c489ee3bf58a461fd3c2c8541f6a36ce5024050ec0f2ce | 15,251 | 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": "20b68470-6775-436b-83f3-76761379a65e",
"created_at": "2026-08-03 20:47:30.956716+00:00",
"db_md5": "fd36de4f4eeef128f3f54dea79b60e96",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "111b78ea-039e-419a-a02c-6650853ea4e5",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.2059743404388428,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008750000037252903
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2002992630004883,
"example_count": 60... |
20c1c67f-94fb-4878-8736-2ab3fb70d381 | easy | yashkant | 2026-08-14 17:39:55.236863+00:00 | succeeded | f20a4720dcf9ee66573b53375f41618a31658dd28a460c18e66a83063d3cb26a | 42,047 | null | """Round1818 parent-preserving temporal Anderson-Dykstra mixing.
The complete Round1792 AdamW5/EMA.995 W8/D2 carrier remains byte-identical at
a matched seed. Initialization of the added temporal mixer is isolated from
the inherited CPU RNG stream, and one exact-zero scalar gate keeps the scored
path equal to the par... | {
"id": "20c1c67f-94fb-4878-8736-2ab3fb70d381",
"created_at": "2026-08-14 17:39:55.236863+00:00",
"db_md5": "6172f737ec835d007e565a29eeea9512",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "059392bd-60dd-4697-8bbf-bfab48749d6f",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 2.8092466592788696,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07375000417232513
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.5021278858184814,
"example_count": 300... |
20c5c44d-54de-4456-a2a3-27d27ac678b0 | easy | ehonig | 2026-08-21 04:57:33.499357+00:00 | succeeded | a8101c7bf20f5fb52423225be6359acd59e36c1d9a8a08942c855acf40ae70ac | 31,882 | 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": "20c5c44d-54de-4456-a2a3-27d27ac678b0",
"created_at": "2026-08-21 04:57:33.499357+00:00",
"db_md5": "75d8fa2efbdc026e18b4eec8692e50b8",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "e60b4f57-1c2e-4cb2-ab78-b0003e8489b7",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.199367402809429,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006250000080714623
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.205705165863037,
"example_count": 600,... |
20c698d1-5347-4b91-9e04-3bcdcfa839a4 | easy | sapient-sapiens | 2026-08-23 17:29:02.283706+00:00 | succeeded | 4654ea70a10ec2d8e00d460c911bcef9ee74e7c4fd2814ef146e56930667c1a7 | 17,851 | null | """Scaled six-digit reverse-decoder tensor train with adaptive refinement.
Legitimate learned model: inputs are categorical token/group/left-relative
embeddings, field markers are used only for structural alignment, and every
readout is supervised against the final answer. No field is decoded into a
number and no int... | {
"id": "20c698d1-5347-4b91-9e04-3bcdcfa839a4",
"created_at": "2026-08-23 17:29:02.283706+00:00",
"db_md5": "408caaea0f49488bd165afeeb0e31ad8",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "1a64de39-aa6a-4564-bf23-9281f3ce920e",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.530931895646101,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01250000002483527
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5503166927350476,
"example_count": 600,... |
20d7086a-868e-4608-ba5c-025f3a577696 | easy | pculaf | 2026-08-25 07:29:03.956362+00:00 | succeeded | 8dff2262b7df53634018f9ca9a77dea10af0ef71d8cfa9a0fa9477ce965774ba | 7,753 | null | """Small single-pass post-RMSNorm Transformer for One Layer Deeper."""
from __future__ import annotations
from collections.abc import Iterable
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
... | {
"id": "20d7086a-868e-4608-ba5c-025f3a577696",
"created_at": "2026-08-25 07:29:03.956362+00:00",
"db_md5": "72caae5a8c1ea6bfd3a7a4ea46fac47d",
"submitter": "Pavle Culafic",
"github_login": "pculaf",
"run_id": "c3c074dc-3e52-490c-a685-fa137f187f43",
"tier": "easy",
"dataset_id": "e6",
"status": "succe... | {
"score": {
"mean_loss": 1.9264346957206726,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03851351607590914
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8965206146240234,
"example_count": 60,... |
20e6b3d3-02a9-495b-b632-c7808e5cb890 | easy | yashkant | 2026-08-09 02:49:56.692003+00:00 | succeeded | 3a333715b69c5a689189d9c634295053e6d180d123a9c36fcf53d1761c802a6f | 22,836 | null | """Round 211 common TRM-style x/y/z shared-transition template."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
from torch import... | {
"id": "20e6b3d3-02a9-495b-b632-c7808e5cb890",
"created_at": "2026-08-09 02:49:56.692003+00:00",
"db_md5": "edd41928889c4a77132c6427945a565b",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "ed417ae0-f9b4-45c2-bca4-eb14c9e824b3",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.1164598959388083,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01937500004656613
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1149879454763987,
"example_count": 800... |
20e9978a-7ac8-41f9-ba11-644b1f810ce0 | easy | oupadhyay | 2026-08-28 02:04:58.427951+00:00 | succeeded | 9b967d2e12d017317b3604fd0bd3303de9161ee725e8b3b300094d7c65603d7b | 3,842 | null | """Factorized digit-place encoding with a persistent globally mixed polynomial cell."""
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
W,D,R=16,48,128
class C:
def __init__(s,vocab_s... | {
"id": "20e9978a-7ac8-41f9-ba11-644b1f810ce0",
"created_at": "2026-08-28 02:04:58.427951+00:00",
"db_md5": "884d93435fe383d7e9e4d5f74c569f05",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "0e6b3d60-2475-498e-a547-e6ee960b2df4",
"tier": "easy",
"dataset_id": "e6",
"status": "s... | {
"score": {
"mean_loss": 1.9540529251098633,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.23963965475559235
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.624762773513794,
"example_count": 60,
... |
20ea0185-6126-49be-918b-bc1f095c8c4b | easy | yashkant | 2026-08-09 02:58:50.980983+00:00 | succeeded | d5fde88fe21de21e443530ce520067a660084202845f6e63cc3e645dcbef5720 | 22,836 | null | """Round 211 common TRM-style x/y/z shared-transition template."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
from torch import... | {
"id": "20ea0185-6126-49be-918b-bc1f095c8c4b",
"created_at": "2026-08-09 02:58:50.980983+00:00",
"db_md5": "c941879f160765469a08cd6840b8b992",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "d66f8713-b66c-4a0f-ad41-c46f726bfc2c",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 4.373364090919495,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05375000275671482
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.366385459899902,
"example_count": 300,
... |
20ecba00-69aa-42d1-8458-29fa4319d806 | easy | oupadhyay | 2026-08-07 11:34:16.702471+00:00 | succeeded | 6bd3231e657807ae6c7caf9f6138e2ef6359518ef842a193bfacb0b948bf55eb | 8,723 | null | """Compressed axial C64 control with shared row/column decimal-place encoding."""
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, PLACE_TABLE_SIZE, BOUNDARY_PLACE, ST... | {
"id": "20ecba00-69aa-42d1-8458-29fa4319d806",
"created_at": "2026-08-07 11:34:16.702471+00:00",
"db_md5": "6a103cabb023e68755f3e150071253d8",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "79f9a3da-e97b-45f2-81de-d2942b00a9ae",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.2412861047527604,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333457509677
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2411325769039547,
"example_count": 6... |
20f51a3a-44c3-4f59-abef-63b5843bec52 | easy | liam-gb | 2026-08-16 23:03:38.035782+00:00 | succeeded | 82632945ab7e0c87d203dfc4e2f107a9e7f49bf180d7e486a194d56b896646c1 | 14,535 | null | """charblock: a from-scratch body whose per-prime step has a known closed-form
solution, trained in the regime where that solution is documented to be found
from random initialisation.
Five ingredients, declared as one ARCHITECTURE+REGIME bundle:
1. Per-prime two-layer blocks with a QUADRATIC activation. For each ban... | {
"id": "20f51a3a-44c3-4f59-abef-63b5843bec52",
"created_at": "2026-08-16 23:03:38.035782+00:00",
"db_md5": "a79f5e7cd40035de9d3d3adb1e10b7b5",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "9e0fe181-4a80-4f68-82f1-1477078b75fe",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.3706767559051514,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3933333344757557
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5836727619171143,
"example_count": 100,... |
20f82d97-ea31-4492-8fe1-d1a54df5d710 | easy | liam-gb | 2026-08-16 19:47:48.555678+00:00 | succeeded | 08861a83832278cc0578fac15abf7aa0b72524ff644357684630ce2e20c6c3db | 15,720 | null | """rns_pm_gather_ct_widet_l32nore: norelrt with LOOPS = 32.
SINGLE CHANGE vs rns_pm_gather_ct_widet_norelrt: LOOPS 8 -> 32. A counted
loop capped at 8 cannot certify past T=8 by construction while the ladder
runs to 64; this is the corrected-recipe body that can. The loop trim keeps
the extra depth free at train time ... | {
"id": "20f82d97-ea31-4492-8fe1-d1a54df5d710",
"created_at": "2026-08-16 19:47:48.555678+00:00",
"db_md5": "c5770d075888ed9d73d2bd866ce45fca",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "3c715dda-1ceb-402c-b8f5-ffdaae48ae6c",
"tier": "easy",
"dataset_id": "e8",
"status": "succeeded"... | {
"score": {
"mean_loss": 4.569880366325378,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3846702282316983
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.8675055503845215,
"example_count": 85,
... |
20f83421-cdd9-4ea9-8255-0ffb40d27301 | easy | sapient-sapiens | 2026-08-17 18:00:29.520564+00:00 | succeeded | 853099ed2d8a675acdec11de42d42228cc6c913667f973274419166c4252761b | 30,225 | null | """Field-local ACT candidate: T-free operator, countable halt, answer-slot readout.
Easy-screen cap is MAX_CHUNKS=8. Raise to 64 before a certification packet or Hard.
SAM is kept. T is never used to compute the answer. Token sequences of N/X/T are
matched in the loss only to route existing training labels onto recurr... | {
"id": "20f83421-cdd9-4ea9-8255-0ffb40d27301",
"created_at": "2026-08-17 18:00:29.520564+00:00",
"db_md5": "1caea6aafb15190f270ce890a5fbc456",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "1550298c-9ca2-433f-adcf-d958704a1837",
"tier": "easy",
"dataset_id": "e9",
"status": "su... | {
"score": {
"mean_loss": 1.9188502818270954,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013333333432674408
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.92389976978302,
"example_count": 90,
... |
20fcb3f0-7fb6-4b20-8b9d-03539df0f0d3 | easy | Maxwells-Demons | 2026-08-13 08:53:07.500787+00:00 | succeeded | 6abbebeebcc3863c286eb320bc9e9af440ef4451af836d64d9c3174be89853de | 109,857 | null | """One shared digit operator, its two free neighbours of x, and two learned maps.
The backbone is unchanged from v47/v96: the state IS the digit one-hot, one
application of the operator is
(state digits, N digits) -> next state digit logits
and the straight-through code of those logits is the whole state handed to... | {
"id": "20fcb3f0-7fb6-4b20-8b9d-03539df0f0d3",
"created_at": "2026-08-13 08:53:07.500787+00:00",
"db_md5": "dab8b95432da3f12736d74330e25643d",
"submitter": "Maxwells_Ayakashi",
"github_login": "Maxwells-Demons",
"run_id": "4f0a492c-c03c-4abe-acdd-ab0d650aab25",
"tier": "easy",
"dataset_id": "e4",
"st... | {
"score": {
"mean_loss": 0.21780250834965645,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.9878125011920929
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.37677678769509115,
"example_count": 12... |
20fd79de-9fad-4121-94ce-b92c09efc7cf | easy | yashkant | 2026-08-17 00:16:11.509443+00:00 | succeeded | 9b2d230bb19e53828319f90a7434c87266aedfff50f82fda48c974958dce3dca | 41,475 | null | """Round2192 late progressive effective batch on the best official source.
The exact model, initialization, loss formula, two-pass SAM plus
Compositional-Muon, LR schedule, and EMA are retained. The evaluator
supplies 96 physical rows. A prefix carries loss before a fixed number
of completed updates; afterward all 96 ... | {
"id": "20fd79de-9fad-4121-94ce-b92c09efc7cf",
"created_at": "2026-08-17 00:16:11.509443+00:00",
"db_md5": "5efba82dda7cebec1d2a13635caacad7",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "2bca13ff-85bd-4cc9-9491-010c98b129e4",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 2.2177754640579224,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.015000000596046448
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.22735595703125,
"example_count": 300,... |
20fd9061-12bd-4c83-a80e-7eac6a0af7a0 | easy | DDanlov | 2026-08-17 11:47:17.799058+00:00 | succeeded | fe6d9746e702113ac667df0bba4de447d35cc99149966d95046cbafa0a3c4561 | 30,329 | null | """
Scalable Recurrent Deliberation Transformer (ScalableDEPHNet)
============================================================
Architecture:
- Pure No-Norm: Single unit RMS normalization strictly isolated to the input token embedding layer.
Zero LayerNorm / RMSNorm inside recurrent deliberation blocks.
- Pure No-Skip... | {
"id": "20fd9061-12bd-4c83-a80e-7eac6a0af7a0",
"created_at": "2026-08-17 11:47:17.799058+00:00",
"db_md5": "3b89ab38649aa714dcea00f059af04e5",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "0053d119-ffe6-4036-bcac-5cda18bd406d",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.1811311239019,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333307504654
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1940901279449463,
"example_count": 100,
... |
210067a4-200e-4358-a985-f16d9953e29d | easy | erdavis0 | 2026-08-26 03:15:43.701119+00:00 | succeeded | c67927e1f1a3eb13d9778a95b41512101bbd0baa70d6fd25c9f9b89607e3826f | 30,360 | null | """Latent-role continuous REV C2: exact length-normalized sequence mixture.
Public delimiter syntax only routes opaque token categories into distinct,
right-aligned N, X, and T tapes. Learned local cells compile separate
interaction, carry, and modulus-conditioned workspaces. N is isolated from
interaction and carry... | {
"id": "210067a4-200e-4358-a985-f16d9953e29d",
"created_at": "2026-08-26 03:15:43.701119+00:00",
"db_md5": "b3917270504f5239445a48889c48e9ae",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "8c489108-317c-4a5f-a5f0-058d8ac6fbaa",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",... | {
"score": {
"mean_loss": 3.5487264621713432,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.535203855668483,
"example_count": 600... |
210739cb-a737-47de-b955-bdcc0d03ca5b | easy | khushidahi | 2026-08-16 01:18:41.395653+00:00 | succeeded | d0fc2cf7817c69df3ad27f741643b9c8e6eda12092b2cd56c33a4870bf193926 | 24,237 | null | """R83: hard-gated diagonal recurrent tape.
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.
The transition re... | {
"id": "210739cb-a737-47de-b955-bdcc0d03ca5b",
"created_at": "2026-08-16 01:18:41.395653+00:00",
"db_md5": "b2f11eff26a74627029433b9421193ef",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "803e07d2-3623-4c66-9831-fa21fe9d8cc5",
"tier": "easy",
"dataset_id": "e6",
"status": "succ... | {
"score": {
"mean_loss": 0.6768182814121246,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.8283784091472626
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.4357982277870178,
"example_count": 60,
... |
210bc0df-72d9-449e-b4af-17fcc92719eb | easy | erdavis0 | 2026-08-29 01:21:52.570237+00:00 | succeeded | c7f78ef5748d23ee6c570a69dda0f9e2f1d45cb5b22c279e1ef01c84faeb2f27 | 17,575 | null | """Sealed N-conditioned circulant operator over an X-initialized latent tape.
Public delimiters only route opaque categories into independent right-aligned
N, X, and T tapes. N alone compiles a small bank of learned cyclic place
kernels and a continuous program vector; X alone initializes the recurrent
state; T is co... | {
"id": "210bc0df-72d9-449e-b4af-17fcc92719eb",
"created_at": "2026-08-29 01:21:52.570237+00:00",
"db_md5": "c929d07e2f2fc815762284c89335e51b",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "19b875bf-0bec-4ab5-8a98-6623b7815086",
"tier": "easy",
"dataset_id": "e7",
"status": "succeeded",... | {
"score": {
"mean_loss": 7.125635385513306,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09888592455536127
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.475359916687012,
"example_count": 85,
... |
21186864-5281-41b1-a433-66eda1cf842b | easy | jordanrubin | 2026-08-03 21:05:51.353469+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": "21186864-5281-41b1-a433-66eda1cf842b",
"created_at": "2026-08-03 21:05:51.353469+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "ee5e7200-895e-4055-965a-6735ae6c2912",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.2799896779821673,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.19958332863946757
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.011237894893385,
"example_count": 300,... |
211c24fd-f38f-4334-a8ad-3a1ae82a6cfd | easy | ehonig | 2026-08-21 01:16:38.103261+00:00 | succeeded | 66f84915a4e39dce066e348521e26897daeb379b8c72116cfbb8ed3a25a44c7d | 31,878 | 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": "211c24fd-f38f-4334-a8ad-3a1ae82a6cfd",
"created_at": "2026-08-21 01:16:38.103261+00:00",
"db_md5": "a47092414289ee3f0cbb9fe119e076e4",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "92bd8126-85ef-44b4-a385-7f6824a58877",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 2.3159477710723877,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06333333067595959
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.580030679702759,
"example_count": 100,... |
211f1e21-240a-40c8-a45d-d4418aea03d3 | easy | repst | 2026-08-26 21:03:27.189073+00:00 | failed | 45b19e898899e0a770f5c6616339fe7828fddaad24d6fb791bcac6e7edd2a53e | 167,339 | null | # R751: hierarchical sparse-polynomial recurrent program induction.
#
# R751 preserves hosted-proven R749's parser, complete-answer objective,
# structured posterior particles, learned LocalCell tape machine, recurrent T
# loop, and learned serializer. It replaces exhaustive STEP/interface crossing
# with an LM-like tw... | {
"id": "211f1e21-240a-40c8-a45d-d4418aea03d3",
"created_at": "2026-08-26 21:03:27.189073+00:00",
"db_md5": "94108ad4e5770d4b0870e848de02f6dc",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "5e6d1307-06a9-42f0-b9da-21f34e9ae170",
"tier": "easy",
"dataset_id": "e1",
"status": "failed"... | null
|
2120c60b-5521-4253-8211-f070a6ea6ac7 | easy | oupadhyay | 2026-08-29 18:10:40.364869+00:00 | succeeded | 2f117d6dcce73a53253aeb7b599eac0444feea9ca195c0d449d632a5ebe36b59 | 3,988 | null | """Conditional optimizer screen: lr3."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,assert_model_state
W,D,H=4,64,32
class C:
def __init__(s,vocab_size,max_seq_len):s.vocab_size,s.... | {
"id": "2120c60b-5521-4253-8211-f070a6ea6ac7",
"created_at": "2026-08-29 18:10:40.364869+00:00",
"db_md5": "5875b6ee33ec06641c91ae76c8ce6b0c",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "5c3f2079-a9ed-4a51-979a-fbb382000161",
"tier": "easy",
"dataset_id": "e10",
"status": "... | {
"score": {
"mean_loss": 1.985803484916687,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.019575757905840874
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9218688011169434,
"example_count": 125... |
21281f7f-778a-4e38-874f-c8d978922c39 | easy | k-penchev | 2026-08-22 21:18:19.856157+00:00 | succeeded | b0c4e93a6d4bdfddd4c92312880fdff15edde6db11f0e0454b72393a20eefecb | 4,581 | null | """Two-loop tied Transformer with digit places and immutable input recall."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL... | {
"id": "21281f7f-778a-4e38-874f-c8d978922c39",
"created_at": "2026-08-22 21:18:19.856157+00:00",
"db_md5": "be9f41abaf0f6535d5af24c2a497dfa3",
"submitter": "Kaloyan Penchev",
"github_login": "k-penchev",
"run_id": "e60f16f9-89f7-43e8-880b-8116b226723d",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 2.509803593158722,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04333333298563957
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.057401180267334,
"example_count": 100,
... |
212adc10-b371-4d45-9cd7-abac56b7dc5c | easy | DDanlov | 2026-08-11 22:06:00.992850+00:00 | succeeded | 35a068f213f2acd8d92e935f1ce45e0f2db1d2de62090292d02b04fdf638d095 | 20,813 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
def init_cutting_the_skip(module):
"""SVD-based orthogonal initialization to ensure initia... | {
"id": "212adc10-b371-4d45-9cd7-abac56b7dc5c",
"created_at": "2026-08-11 22:06:00.992850+00:00",
"db_md5": "d1c53c08329d28c9db42c287e502cb5d",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "64747b12-ca67-463c-8be4-43eafbc1079e",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 49.10432861413358,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 49.58456633871446,
"example_count": 600,
"exact... |
212dd6fd-5b9b-45e9-bd7c-3d9955b9f5ce | easy | sapient-sapiens | 2026-08-20 17:30:28.489309+00:00 | succeeded | 31ee43115ba37c3df1a6895dd2abd06fee95edf3f6730c954b5fab84d276d8f1 | 44,771 | null | """Persistent pair-indexed relational state around a categorical operator.
The recurrent answer state remains a bank of right-aligned categorical digit
slots. A separate continuous S-by-S grid retains every ordered digit-pair
cell across applications. Learned bilinear bindings and categorical modulus
context update ... | {
"id": "212dd6fd-5b9b-45e9-bd7c-3d9955b9f5ce",
"created_at": "2026-08-20 17:30:28.489309+00:00",
"db_md5": "b17bbb875c9d65c219993ce40aedcf37",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "0a85eb15-d8fe-46d2-bef0-d734a689aaa0",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.4070016175921767,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010000000024835268
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3846343900056164,
"example_count": 60... |
212e4c23-bfe4-4042-aab9-fd87435014e4 | easy | CodeReclaimers | 2026-08-18 22:28:46.329709+00:00 | succeeded | d12107e96c7eb0c5b87f90836a677a9f097d59b50f7b9d9f61e7b2216153ea29 | 13,459 | null | """T=1-first field-aware functional-root search.
Four cells share one reversible Feistel mechanism and differ only in how the
visible N field is interpreted: ordinary decimal, reversed decimal, inverse
Gray, or bit-reversed. X and Y remain in the visible 20-bit state coordinate.
Each cell also has a small, zero-initi... | {
"id": "212e4c23-bfe4-4042-aab9-fd87435014e4",
"created_at": "2026-08-18 22:28:46.329709+00:00",
"db_md5": "ee2be1eafc03e9d2321fa079c29427d7",
"submitter": "CodeReclaimers",
"github_login": "CodeReclaimers",
"run_id": "012a1129-78d0-4b4f-b18c-4ce06365f264",
"tier": "easy",
"dataset_id": "e2",
"status... | {
"score": {
"mean_loss": 21.428699493408203,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333972841501
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 21.463415145874023,
"example_count": 30... |
2131c0dc-88ec-4089-8233-134b82441ced | easy | DDanlov | 2026-08-13 19:44:46.235525+00:00 | succeeded | 939f58d818c3f953053dfcbc068030b31142e50f323a87b6d7f84ab5df0f2dd5 | 26,904 | null | from __future__ import annotations
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
try:
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
except ... | {
"id": "2131c0dc-88ec-4089-8233-134b82441ced",
"created_at": "2026-08-13 19:44:46.235525+00:00",
"db_md5": "e3de2d9e8ccead933e2b8372c4ddba6b",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "d734ad47-ff55-4a66-8dbd-bc92457129d5",
"tier": "easy",
"dataset_id": "e2",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.3230648968406205,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.019791666766007743
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3172093609068867,
"example_count": 30... |
213f349a-5195-4d5c-a973-3c6ef456fdac | easy | sapient-sapiens | 2026-08-14 17:41:19.380494+00:00 | succeeded | 28ba469aae3e42ca679270105f219b6bd47433fa80f83dc154b3c3e55e44c7c3 | 14,299 | null | """Digit-space recurrent operator for repeated modular squaring.
Nothing here converts a token into a decimal number. Digits stay as embeddings,
so what a digit is worth, and every carry and reduction, is learned. That also
sidesteps a hard numeric wall: at Hard's modulus scale `x^2` runs past the 53-bit
exact-integer... | {
"id": "213f349a-5195-4d5c-a973-3c6ef456fdac",
"created_at": "2026-08-14 17:41:19.380494+00:00",
"db_md5": "1efadda9e01ae84ec2857432a666c542",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "6b3f2b22-d6cd-49b5-a9ff-5aa58b10d410",
"tier": "easy",
"dataset_id": "e4",
"status": "su... | {
"score": {
"mean_loss": 2.070760418426566,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008645833333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.0754522683315155,
"example_count": 120... |
213fa0ef-bf5a-4567-8e05-a69b9db396da | easy | KaustubhKumar05 | 2026-08-28 17:26:58.721493+00:00 | failed | 4069ab75919b6843dc21e38aa802299da7360e1c94018f18c01c64428a5f8397 | 15,137 | null | """w1_value -- the §124 value-space core wired into the harness.
THE MECHANISM (FINDINGS §121-§124). One modular squaring is exactly a scalar map on the
normalised value u = y/N:
u' = frac(N * u^2) since (y^2 mod N)/N = frac(y^2/N) = frac(N*(y/N)^2)
A POPULATION of randomly-initialised triples (a,b,c) with ... | {
"id": "213fa0ef-bf5a-4567-8e05-a69b9db396da",
"created_at": "2026-08-28 17:26:58.721493+00:00",
"db_md5": "4460cc801d215f1140bec2ea17315b67",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "e6f0e080-c581-4c58-a9fe-6d52f577cba5",
"tier": "easy",
"dataset_id": "e5",
"status": "failed... | null
|
21412ef6-b67b-472e-b26d-0c0ac0fa4ba6 | easy | jordanrubin | 2026-08-02 17:52:30.913480+00:00 | succeeded | 695cb20b4638147c2f4355dfbb227ff1dcf6f892b0188f80bdac2511d008ad7f | 20,828 | null | """Looped bidirectional selective-SSM (Mamba-style), pure PyTorch.
The statefulness bet: input-dependent SSM dynamics along the token axis,
with the block weight-tied and applied NUM_LOOPS times in depth with input
injection — recurrence in depth is where the task's T-iteration lives.
The selective scan is computed i... | {
"id": "21412ef6-b67b-472e-b26d-0c0ac0fa4ba6",
"created_at": "2026-08-02 17:52:30.913480+00:00",
"db_md5": "aabbea9f4f3eb2e2cd3c1011cfd4c1e7",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "3f3669c4-3252-4c4c-95db-a47139fb3877",
"tier": "easy",
"dataset_id": "e1",
"status": "s... | {
"score": {
"mean_loss": 4.860153578476649,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03999999910593033
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.7866146457729055,
"example_count": 100,... |
21443919-8947-4363-8048-004fa6215b13 | easy | jordanrubin | 2026-08-05 21:00:31.793509+00:00 | succeeded | b274baac55d74bcf98433731b295da92f1217792b897a463d3fcab46d583d486 | 24,859 | null | """Adaptive exploration template for the exact-T looped Transformer.
The T=1 computation is the proven value-Fourier baseline cell: a D=256,
three-block bidirectional Transformer with abacus digit positions and learned
Fourier features over the parsed integer x. For T>1 the *same* three-block
cell is reused exactly T... | {
"id": "21443919-8947-4363-8048-004fa6215b13",
"created_at": "2026-08-05 21:00:31.793509+00:00",
"db_md5": "81660c623ad99444cb3c48d3da4f2f92",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "d3384d32-b5b0-4f9e-91b9-8292976580e7",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 7.777645756480153,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.18374999744507173
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 11.48886879423631,
"example_count": 300,
... |
214ac33a-1f56-45ff-963a-b433b9c57469 | easy | mnida | 2026-08-23 18:30:01.344155+00:00 | failed | 22c9d3c643f90947b262cf31d6406bcde8e9e06e9cf05b8e321a783ca40a5467 | 6,221 | 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": "214ac33a-1f56-45ff-963a-b433b9c57469",
"created_at": "2026-08-23 18:30:01.344155+00:00",
"db_md5": "c5570677005004c307cb30cc3e4f5ea1",
"submitter": "mnida",
"github_login": "mnida",
"run_id": "3edac525-ab0e-4440-b1f4-b024eca37a14",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
"st... | null
|
2151e8f7-9035-4806-baad-373ed44fded1 | easy | DDanlov | 2026-08-18 10:34:13.220401+00:00 | succeeded | 9e2e886a7f72218c6f6788250c79c4edcba3908cfb09a02021dbdad2941008ba | 23,584 | null | """
Official Submission for One Layer Deeper Challenge
Architecture: 4-Layer Recurrent Core Pure No-Norm No-Skip DKS Reasoning Model:
- 4 Distinct Pure No-Norm No-Skip Transformer Layers in the Recurrent Core: L1 -> L2 -> L3 -> L4
- Unrolled T times for serial modular squaring
- dim=512, 16 heads, d_ff=1024, RohanShamp... | {
"id": "2151e8f7-9035-4806-baad-373ed44fded1",
"created_at": "2026-08-18 10:34:13.220401+00:00",
"db_md5": "d6174d2a9211082098f18de894eca987",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "ace5d758-4d57-4f7b-93d0-1c01e2de70a4",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.037996196458864,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06500000054637592
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.005561535038165,
"example_count": 100,
... |
215d6a19-1784-4edd-b091-db7fe110c02a | easy | velocizapkar | 2026-08-16 07:22:51.159087+00:00 | succeeded | 3d594d881d9e5c1ec1c636e06f00aac9c08e07aae33c4d291961825e58ceb79b | 9,694 | null | """Standard residual Transformer with one block tied four times.
This is the ordinary L4 standard-Transformer control with one architectural
intervention: one physical pre-RMS residual Transformer block is applied four
times with shared parameters. Training and evaluation both execute all four
applications. There is... | {
"id": "215d6a19-1784-4edd-b091-db7fe110c02a",
"created_at": "2026-08-16 07:22:51.159087+00:00",
"db_md5": "cc8bb72a9785373930bb9d57fce2aca0",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "9c9bbec4-d04f-4e62-bfd9-742c3e1247e9",
"tier": "easy",
"dataset_id": "e6",
"statu... | {
"score": {
"mean_loss": 9.380731582641602,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05180180352181196
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.045494079589844,
"example_count": 60,
... |
2173823e-1ba4-425d-964d-9152dad077ef | easy | khushidahi | 2026-08-16 13:03:22.694561+00:00 | succeeded | 2e371d5f156e0ebd361c0044fe5a517f1e9f316577b08dddd74fa033401d17ac | 24,535 | null | """R107: high-dropout decimal transition control.
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.
The transit... | {
"id": "2173823e-1ba4-425d-964d-9152dad077ef",
"created_at": "2026-08-16 13:03:22.694561+00:00",
"db_md5": "aba8d44fb16e918c71d6cd7a3c242190",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "7a977937-7874-4b04-9c89-3f5296d4edd8",
"tier": "easy",
"dataset_id": "e10",
"status": "suc... | {
"score": {
"mean_loss": 1.1542063057422638,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.7266060709953308
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.7690520882606506,
"example_count": 125,... |
2179cde1-b4e0-4182-94d8-a42204f1080c | easy | himalalps | 2026-08-07 05:08:28.200425+00:00 | succeeded | 11e74513bebef3436a5311354fd021c686767b76c227f06a10f094da16142808 | 69,331 | null | """Algorithmically constrained latent recurrence for One Layer Deeper.
One masked cross-attention pass creates two independent representations:
``state_0`` from X only and an immutable context from N only. T is excluded
from both representations and is used solely to choose how many times the
learned recurrent MLP tr... | {
"id": "2179cde1-b4e0-4182-94d8-a42204f1080c",
"created_at": "2026-08-07 05:08:28.200425+00:00",
"db_md5": "ef6e5cff942f5020b8c4249ccaea1b2f",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "150093f9-e017-44b5-8fdc-56b26b152c04",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 3.7295593719694526,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333289871613
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.414841413497925,
"example_count": 600... |
217c342c-f44c-48b2-bd12-e04c1f2c7dca | easy | KaustubhKumar05 | 2026-08-28 17:28:45.400377+00:00 | failed | f3c1c96fa099332861b80d17954e137ea0a749bd90afbd76193f9f904612f65d | 15,832 | null | """w1_value -- the §124 value-space core wired into the harness.
THE MECHANISM (FINDINGS §121-§124). One modular squaring is exactly a scalar map on the
normalised value u = y/N:
u' = frac(N * u^2) since (y^2 mod N)/N = frac(y^2/N) = frac(N*(y/N)^2)
A POPULATION of randomly-initialised triples (a,b,c) with ... | {
"id": "217c342c-f44c-48b2-bd12-e04c1f2c7dca",
"created_at": "2026-08-28 17:28:45.400377+00:00",
"db_md5": "61a9f53066b6961b4adf36933bd5576e",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "8cb40610-df0f-4e45-9264-9ec119fc46aa",
"tier": "easy",
"dataset_id": "e5",
"status": "failed... | null
|
2181e08f-dc22-4831-b4f7-640971cda55f | easy | DDanlov | 2026-08-10 09:11:22.929100+00:00 | succeeded | 1d95a12c86ada719242d9e69a7b1a337ba6885bc25c27319efb512b93cbbd29d | 15,611 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RMSNorm(nn.Module):
def __init__(self, dim: int, eps: float = 1e-6):
super()... | {
"id": "2181e08f-dc22-4831-b4f7-640971cda55f",
"created_at": "2026-08-10 09:11:22.929100+00:00",
"db_md5": "009b5f2005ad2ad440ee24955adcd16a",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "29db9b68-5208-4ad2-a164-4150740db1f7",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 34.02553638020656,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333333333335
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 43.359004974365234,
"example_count": 10... |
219f48f3-336b-4928-9809-2cac9c265cc0 | easy | LauraGomezjurado | 2026-08-30 21:45:54.224380+00:00 | succeeded | cd67b521e635e576bcf20ae237fcb173c9531bd16e5ee0c9b709d96c027946e7 | 9,045 | null | """Looped depth-recurrent Transformer for repeated modular squaring.
Design notes
------------
The evaluator reads answers off the *last* ``target_len`` prompt positions, so
answers are right-aligned with the least-significant digit at the final valid
position. Digit significance is therefore a fixed offset from the e... | {
"id": "219f48f3-336b-4928-9809-2cac9c265cc0",
"created_at": "2026-08-30 21:45:54.224380+00:00",
"db_md5": "30f792a2fe0639423503695e4977766e",
"submitter": "LauraGomezjurado",
"github_login": "LauraGomezjurado",
"run_id": "9f63b39a-c233-4c8f-9733-fa19870d41ae",
"tier": "easy",
"dataset_id": "e7",
"st... | {
"score": {
"mean_loss": 11.711393810453869,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.029010695997964252
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 12.36397647857666,
"example_count": 85,... |
21a1ba5b-fd39-415f-978d-9bc64d8c4f47 | easy | isaac0804 | 2026-08-04 17:28:33.214379+00:00 | succeeded | a844da2bc4a2d90c1bcf4dadaa9998fc67f6e057443651dede6848ee5d7c69a6 | 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": "21a1ba5b-fd39-415f-978d-9bc64d8c4f47",
"created_at": "2026-08-04 17:28:33.214379+00:00",
"db_md5": "6847937dc71381e1fa0eb292a1fba5b8",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "6deebda5-701f-47ff-b63b-28a4d7cedbbb",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 1.9224886894226074,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0566666666418314
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8938798904418945,
"example_count": 100,... |
21aa83f9-87f9-4273-a80c-fdf716ad2081 | easy | AmanPriyanshu | 2026-08-03 23:24:54.546856+00:00 | succeeded | a4d82b3edf1d84985963ccf27f54e3a08dad1f4a06a8f6d37291306f6845012e | 9,054 | null | """One Layer Deeper — attempt 2.
Changes vs attempt 1 (which memorized x and never used the loop):
* SHORTCUT BLOCKED. No fat transformer over the whole prompt. Each field is
read on its own: N-digits -> modulus context c; x-digits -> initial residue
state h0. The ONLY compute that yields the answer is the w... | {
"id": "21aa83f9-87f9-4273-a80c-fdf716ad2081",
"created_at": "2026-08-03 23:24:54.546856+00:00",
"db_md5": "e0614f817eb90ccaac11628249f38e6c",
"submitter": "Aman Priyanshu",
"github_login": "AmanPriyanshu",
"run_id": "4bd25f1e-0dfe-4eb2-b530-ecf4a47dd7fc",
"tier": "easy",
"dataset_id": "e1",
"status"... | {
"score": {
"mean_loss": 1.994743525981903,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0566666666418314
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9303830862045288,
"example_count": 100,
... |
21ad0a63-fbc7-4f54-8d6c-b694df8d6016 | easy | anirudh-chakravarthy | 2026-08-12 04:16:29.080908+00:00 | succeeded | defd48c07ed24a42afa52d8d672b2caa211de4f5dda8bec42baf1a813d07ebec | 15,401 | null | """A learned state transition applied once per requested time step."""
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,
asser... | {
"id": "21ad0a63-fbc7-4f54-8d6c-b694df8d6016",
"created_at": "2026-08-12 04:16:29.080908+00:00",
"db_md5": "97b7b03f7bfa1d4515ed1a005cac05e8",
"submitter": "Anirudh S Chakravarthy",
"github_login": "anirudh-chakravarthy",
"run_id": "3311a531-6ffa-4b20-b072-f05a795f98ba",
"tier": "easy",
"dataset_id": "... | {
"score": {
"mean_loss": 3.7449017495683856,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011250000037252903
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.5175401704728335,
"example_count": 60... |
21b4d349-894e-4de6-8c86-0d401b3a6d18 | easy | velocizapkar | 2026-08-09 15:53:19.580938+00:00 | succeeded | 32d61e3e9da2aa3a99fbfa1ffb056d15567b0cd8352b122b8a053802e45c2e2a | 143,748 | null | """Fast packed residual-free Attention-Residual transformer with Muon.
The model is a generic learned sequence model. It contains no parser,
arithmetic primitive, task constant, generated data, label feedback, or answer
construction. Every macro-state is produced by a serial pack of attention and
MLP transformations... | {
"id": "21b4d349-894e-4de6-8c86-0d401b3a6d18",
"created_at": "2026-08-09 15:53:19.580938+00:00",
"db_md5": "5b4187f454aa3ebe1d7fd02a837c991c",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "64f76d07-bd0c-42c8-a800-05ac4d268075",
"tier": "easy",
"dataset_id": "e1",
"statu... | {
"score": {
"mean_loss": 3.4823487997055054,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999921768904
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.789088726043701,
"example_count": 100,... |
21ba0a50-e3d2-4b59-9d4e-3a07f0666360 | easy | MarcoF1 | 2026-08-13 06:38:20.823915+00:00 | succeeded | 1f64159b085297de9734b8b0a3b5b77cc5c1ff041ab69a479665cca9f7402c45 | 3,892 | null | """Untied 4-block transformer control (fixed feedforward depth, no recurrence)."""
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_... | {
"id": "21ba0a50-e3d2-4b59-9d4e-3a07f0666360",
"created_at": "2026-08-13 06:38:20.823915+00:00",
"db_md5": "74b8c3eb349ccafa247ae3d5f0f2facc",
"submitter": "Marco Fleming",
"github_login": "MarcoF1",
"run_id": "de3aee06-de91-43c2-8e44-9c9f9998b8be",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 4.815616638583139,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.019999999776482584
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.80443811416626,
"example_count": 100,
... |
21c128c9-f040-4be5-a1b0-bf3e60f8c7bf | easy | blake-camp-surge | 2026-08-05 19:05:54.906863+00:00 | succeeded | 33ce5d3185cdfe9949dcb271fba5496c276a71a83b0bdaf4c58c3821be0af3c3 | 10,518 | null | """Vector-state neural cellular automaton over attention edges."""
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,
)
TOKEN... | {
"id": "21c128c9-f040-4be5-a1b0-bf3e60f8c7bf",
"created_at": "2026-08-05 19:05:54.906863+00:00",
"db_md5": "e02ce0cff1f74d0a954a1c4509708d2a",
"submitter": "Blake Camp",
"github_login": "blake-camp-surge",
"run_id": "c51d3b36-5b67-40aa-8ee4-c80b7fc14a62",
"tier": "easy",
"dataset_id": "e3",
"status":... | {
"score": {
"mean_loss": 2.129553635222386,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000000000001
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1273640209153064,
"example_count": 800... |
21c1f295-e839-4302-b607-fcdfc7bb031f | easy | CodeReclaimers | 2026-08-16 23:23:01.261753+00:00 | succeeded | f530469f2ed397bcd446c8f231af2ea9e23dd9d11b0e8e5b1e5fd2e5191de930 | 7,843 | null | """Tied learned Feistel transition for hidden-permutation recovery.
The decimal boundary is decoded into a 20-bit state. A single learned,
N-conditioned round function is reused once per literal T while the Feistel
wiring supplies an invertible state transition. The round values and key
schedule remain learned; no c... | {
"id": "21c1f295-e839-4302-b607-fcdfc7bb031f",
"created_at": "2026-08-16 23:23:01.261753+00:00",
"db_md5": "3433eced7c89d2f67c221bb7753d823d",
"submitter": "CodeReclaimers",
"github_login": "CodeReclaimers",
"run_id": "65c79016-5ed9-43aa-bb5a-1d65a71f3bf9",
"tier": "easy",
"dataset_id": "e1",
"status... | {
"score": {
"mean_loss": 20.787620544433594,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.014999999664723873
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 20.328357696533203,
"example_count": 10... |
21ca1621-6253-4e85-af45-99e1260f9b11 | easy | 0Chris5R | 2026-08-07 09:53:38.898687+00:00 | succeeded | 512b497ab96a63251c8fbb30f00d691fb4443533f3f3521773737b454fd9f960 | 14,519 | null | """Learned convolutional-GRU over an aligned field-by-digit workspace.
The public field markers and decimal significance define only the layout of
the neural workspace. Every transition and output is learned; there is no
arithmetic routine, recurrence assumption, lookup table, or generated target.
"""
from __future__... | {
"id": "21ca1621-6253-4e85-af45-99e1260f9b11",
"created_at": "2026-08-07 09:53:38.898687+00:00",
"db_md5": "a9a596b5467befb031d3e7e4f48c8a6d",
"submitter": "Chris ",
"github_login": "0Chris5R",
"run_id": "d468ef8d-750c-40e6-b1ca-e0baefffdf26",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 7.108837771263948,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013750000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.333759810357885,
"example_count": 600,... |
21cc2195-6f3b-4d1a-b78d-6e51a29cd099 | easy | DDanlov | 2026-08-07 09:17:15.022789+00:00 | succeeded | b92edf30a80bcf5946014793c044ecf8f929f0368000c6e308cde84bedf6c38f | 10,896 | null | """
Highly Scaled Official Competition Submission:
1. Architecture: Scaled PLER K=30 Unrolled Recurrence + Modular Fourier Phase Embeddings + Causal Decoder Head (D_MODEL=768, NUM_HEADS=48, NUM_STEPS=30).
2. Optimizer: RohanShampoo & BlockShampoo with Pseudo-Inverse Eigenvalue Trick (S > 1e-15) + Sutskever 2011 Nestero... | {
"id": "21cc2195-6f3b-4d1a-b78d-6e51a29cd099",
"created_at": "2026-08-07 09:17:15.022789+00:00",
"db_md5": "38d70b909ebd4d3491f7cafcff041218",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "844315ec-c4be-48a0-a168-54094d8a0aec",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 13.54390287399292,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 14.25390911102295,
"example_count": 100,
"exact... |
21d0f1fc-1156-4776-872a-4a32a2c22de1 | easy | sapient-sapiens | 2026-08-30 22:38:28.970340+00:00 | succeeded | 01e2ebb2355af11bcf7e4e29943b8829755b18f8ac4f255c0df1942da33bd786 | 19,443 | null | """Eligible E5 direct-Horner model with detached same-final suffix credit.
The ordinary eight-square branch retains full BPTT and weight 0.9. A T-aware
selector chooses a detached predecessor from X/A1--A7, and one fresh tied
square receives weight 0.1 against the same final answer labels. No
intermediate value, depth... | {
"id": "21d0f1fc-1156-4776-872a-4a32a2c22de1",
"created_at": "2026-08-30 22:38:28.970340+00:00",
"db_md5": "2e15853276285a2c82bdc677cc5fec7c",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "81c8d75a-a020-4dfd-aa73-52b97d2e5e88",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.2177478812626816,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000012417634
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.221694741013873,
"example_count": 600... |
21debf90-6a6e-45c0-94d5-1f04a36a0f2b | easy | AFeinfield | 2026-08-28 04:11:01.659215+00:00 | succeeded | ebc2e492aa5c558230e4172d0a761b926fb7db1491c70dacf6698415f71a5dd0 | 9,266 | null | """189: DIGIT-TUPLE GROKMAP — the h1-scale fusion.
F maps per-place digit distributions to per-place digit logits through
one dense quadratic trunk, conditioned on N via per-place digit
embeddings. The output IS the composable state (per-place CE lands on
it directly — no head to shortcut through, the 181 lesson), com... | {
"id": "21debf90-6a6e-45c0-94d5-1f04a36a0f2b",
"created_at": "2026-08-28 04:11:01.659215+00:00",
"db_md5": "e6c0ba375db62da06f001e62befe1ea9",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "68643191-1650-4190-9edf-49b08d4bbb27",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 5.66194442062047,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006250000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.9359762080581735,
"example_count": 600,... |
21e267ef-64b1-459f-8ae8-175d5a9b0349 | easy | yashkant | 2026-08-13 00:42:57.221532+00:00 | succeeded | c258f69b5c6297ef14dfef19f6380a3e891cfc4f5cb2882d145e9de77fc5b058 | 31,392 | null | """Round1702 one-projection-per-T Dykstra recursion on the D416 carrier.
Each parsed public-T cycle applies one shared learned smooth projection and
alternates the A/B set code and correction memory across cycles. The immutable
anchor is reinjected on every call; all active cycle outputs are averaged, and
terminal CE... | {
"id": "21e267ef-64b1-459f-8ae8-175d5a9b0349",
"created_at": "2026-08-13 00:42:57.221532+00:00",
"db_md5": "73b4ae94b7b4b114c03adec2acb6db0d",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "628d9808-169b-4b8a-a4c0-9b7112c4d486",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 4.342406272888184,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.2756249997764826
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.885671138763428,
"example_count": 300,
... |
21e68712-5d80-4f85-92d9-dd6b2eea3baf | easy | oupadhyay | 2026-08-12 07:03:44.891999+00:00 | succeeded | 2b46b514557cfa66ccbb90ec5e4143aed9d85d03074d80b413f51e551733bdb1 | 8,480 | null | """Standalone RMS/axial gate-calibrated C44 for the Easy E5 tier."""
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
PAD, N_MARK, X_MARK, T_MARK, DIGIT... | {
"id": "21e68712-5d80-4f85-92d9-dd6b2eea3baf",
"created_at": "2026-08-12 07:03:44.891999+00:00",
"db_md5": "8417a8d871eacb667debddd3806ac54c",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "cf059146-062c-4dd3-a134-f7e8fca529f4",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.283354864810748,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00458333338300387
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.194888452777948,
"example_count": 600,
... |
21ebb641-2b05-4613-9728-a8ac921de7c3 | easy | jordanrubin | 2026-08-03 20:42:33.662339+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": "21ebb641-2b05-4613-9728-a8ac921de7c3",
"created_at": "2026-08-03 20:42:33.662339+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "e7dedba1-b82c-4496-a6bf-10d82c6fe531",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.39995121625902,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.2020833301357925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.33300342017073,
"example_count": 300,
... |
21ecafbd-5e08-4cea-9ef8-460179854513 | easy | yashkant | 2026-08-16 00:44:35.107650+00:00 | succeeded | 2584011f05bbc145175fe808a5f2c287405d2780b0931910b585d1d19e4990a0 | 34,958 | null | """Round1776 anisotropy-only centered diagonal-metric Dykstra.
The proven Round1766 W8/D2 recurrent carrier is retained. Before each named
set projection, a shared network maps the immutable input anchor and that set's
code to a positive diagonal metric. Its bounded log-diagonal is mean-centered
so the metric has un... | {
"id": "21ecafbd-5e08-4cea-9ef8-460179854513",
"created_at": "2026-08-16 00:44:35.107650+00:00",
"db_md5": "3c1f0b2a1ae90d8ea343a18948370c09",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "3f9d133f-1204-4a97-bcc3-497ffffe540b",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.1230965982384253,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.015625000158324836
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.118781420680019,
"example_count": 800... |
21f1ced5-6094-4510-8abb-73eda7218b12 | easy | gauravmishra | 2026-08-10 03:57:20.220339+00:00 | succeeded | c4ec4077778d73f2fed53ba22226a5bc64955358a7f9e3ff4d08eb331a411778 | 37,802 | null | """Information-gain candidate ig60_e42_e23_e20_late_deep_supervision_e5 (selected_e23_convlstm_e20)."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
Token... | {
"id": "21f1ced5-6094-4510-8abb-73eda7218b12",
"created_at": "2026-08-10 03:57:20.220339+00:00",
"db_md5": "efc731a1ba9935c7f93db6dad8f91915",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "6feb19ca-ff54-4d84-91c1-ea61c427ef24",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.2829304193194497,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000024835269
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2923719284245787,
"example_count": 60... |
21f4df11-43ce-4037-beec-fee7f2442c5d | easy | TZZheng | 2026-08-06 17:31:00.923381+00:00 | succeeded | e5617b9321f35a17380da2b7a02c73a222134533d04697461ce4211bb191c03b | 7,295 | 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": "21f4df11-43ce-4037-beec-fee7f2442c5d",
"created_at": "2026-08-06 17:31:00.923381+00:00",
"db_md5": "ad23af3eca4cb6eda12274c0c16da9c5",
"submitter": "TZZheng",
"github_login": "TZZheng",
"run_id": "614ec0d9-ca43-4401-830c-7dc5f089b2dd",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.963404655456543,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666828095913
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.862797975540161,
"example_count": 100,... |
21f6ebc9-9710-4bc3-9b1e-1fd63167a90e | easy | DDanlov | 2026-08-20 22:44:40.703546+00:00 | succeeded | 0458dfb13ab4fa21fe031a02fd364a797dd0b61babf6b1712ed3433574fce79b | 21,253 | null | from __future__ import annotations
import math
import time
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except Imp... | {
"id": "21f6ebc9-9710-4bc3-9b1e-1fd63167a90e",
"created_at": "2026-08-20 22:44:40.703546+00:00",
"db_md5": "49b5dd3da95b2dfbf04c288b6da6c3e1",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "37f2be8a-bf37-4298-851b-beadb722cedd",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.695224105009083,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002083333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.695526641847841,
"example_count": 600,... |
21fd369b-a7ff-4ac2-87c7-0077c0c8e7ce | easy | ehonig | 2026-08-19 21:31:27.444705+00:00 | succeeded | 7dc2043cecd13cab5fb3e1008fe569ad6ac785e7d8f1a50678606a994ebdceac | 19,541 | 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": "21fd369b-a7ff-4ac2-87c7-0077c0c8e7ce",
"created_at": "2026-08-19 21:31:27.444705+00:00",
"db_md5": "c448d4fc9b9956de69876379d279e004",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "c4f6d270-9675-446a-8e60-c85edcfd0712",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 2.222408890724182,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011041667079553008
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2221014499664307,
"example_count": 300... |
21fe6c26-aa93-4164-a887-027b94e37798 | easy | karanganesan | 2026-08-08 04:54:11.695532+00:00 | succeeded | 8d9eecdc8607c91b190336e24b49597738decc69ce7d4e24ee5930ae97cfcad7 | 27,127 | 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": "21fe6c26-aa93-4164-a887-027b94e37798",
"created_at": "2026-08-08 04:54:11.695532+00:00",
"db_md5": "3e8439ba8ab561733e7dc883fab4ecb2",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "304746c3-be4c-4f41-b12f-5ecfcd497180",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 7.0292041301727295,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002916666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.586562633514404,
"example_count": 600... |
22099502-eea9-4b91-8995-ed139ff18fe9 | easy | shreyash-chonkie | 2026-08-26 12:34:38.014981+00:00 | succeeded | a34f8cb48ae57557b50c7e97ccab1d282173b8282c8657a5981af384a27efba4 | 16,102 | null | """Hard source/opcode compiler with a separately owned shared executor."""
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,
a... | {
"id": "22099502-eea9-4b91-8995-ed139ff18fe9",
"created_at": "2026-08-26 12:34:38.014981+00:00",
"db_md5": "2f7c2cccd0cd518c883e9b8c55c4e2a7",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "6f0be4ce-ce2c-4d6a-be6b-8471a7fae97e",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 4.942991018295288,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05833333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.973415851593018,
"example_count": 100,
... |
220a3aa0-a03d-4ebf-b406-9a1e779848cf | easy | ehonig | 2026-08-20 09:27:01.092582+00:00 | succeeded | a7154c44f5851446647af8a4af8ea058fb31949be86993d1b443619ec5f5d2d6 | 23,852 | 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": "220a3aa0-a03d-4ebf-b406-9a1e779848cf",
"created_at": "2026-08-20 09:27:01.092582+00:00",
"db_md5": "0b9155ecb51838be79f2399af3d031ac",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "a15616f6-24dc-42b9-b3f0-27d6c98aac56",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 2.161450743675232,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1725947856903076,
"example_count": 100,
... |
22115603-f937-4ad0-b528-5edda538e173 | easy | DDanlov | 2026-08-09 00:29:55.480668+00:00 | succeeded | 28ba57acef822e56594ad6b76f6a1a8ee532fba76d43b02a794cf04f0bdb5da9 | 15,978 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with eigenvalue ma... | {
"id": "22115603-f937-4ad0-b528-5edda538e173",
"created_at": "2026-08-09 00:29:55.480668+00:00",
"db_md5": "83ac8c461fec9fd0f4f1ac556b0de5bc",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "601c139a-1f27-47c2-94d0-581b34a2d8da",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 35.08291795071422,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007083333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 41.26314616096394,
"example_count": 600,... |
221a47ec-4b7b-440c-a5df-219b1c3eaa6a | easy | velocizapkar | 2026-08-16 07:34:36.590202+00:00 | succeeded | 684e4ce71f1429a11cd2d4c102654ffe396fc6c3dc2bcda9be300d8b26fec6f4 | 9,485 | null | """Matched standard bidirectional Transformer control.
This is deliberately an ordinary fixed-depth Transformer: learned token and
absolute-position embeddings, full softmax self-attention, pre-RMSNorm,
SwiGLU, residual connections, a final RMSNorm, and an untied vocabulary head.
It has no recurrence, convolution, scr... | {
"id": "221a47ec-4b7b-440c-a5df-219b1c3eaa6a",
"created_at": "2026-08-16 07:34:36.590202+00:00",
"db_md5": "007f76a50ca36a338a51d681a1cf8009",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "36f24d85-e139-4ca7-9bbb-c6c856493643",
"tier": "easy",
"dataset_id": "e7",
"statu... | {
"score": {
"mean_loss": 7.0725319385528564,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.042468806728720665
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.878723621368408,
"example_count": 85,... |
22233939-dbc4-4ea1-817c-9628e01862b6 | easy | yashkant | 2026-08-10 20:11:33.899941+00:00 | succeeded | 102d896087066490f6ebb4f18324d4d7ca2ac60bc688db06d565ff2e52b46913 | 31,137 | null | """Round1282 shared continued-fraction denominator-recursion pair.
The complete Round1141 D416 MemGLU carrier remains byte-identical. Its pooled
representation seeds a positive terminal denominator. Every shared call adds
one tied continued-fraction level, ``P(x) + softplus(beta)/(state + epsilon)``.
The matched contr... | {
"id": "22233939-dbc4-4ea1-817c-9628e01862b6",
"created_at": "2026-08-10 20:11:33.899941+00:00",
"db_md5": "df57b8b95b0a939244b6cb428b1315f0",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "cdff8a0e-d55b-4059-91c7-ab7fab9fe180",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 2.1471344180920147,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666679084302
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1518826762656995,
"example_count": 60... |
22287f22-7107-4afb-bd3b-7932060297a0 | easy | viridale | 2026-08-24 23:30:18.649563+00:00 | succeeded | 2f89b4066e50c15045471df2c2a43a4196a22f90a3adc1875e0dbfd8d36bffd2 | 18,055 | null | """Shared-coordinate binding quotient-ideal Transformer.
hypothesis: v869 subtracts independently projected, separately normalized tensors;
placing every law/certificate route in one digit-algebra basis and subtracting
before one shared projection makes modular proof geometry real and learnable
axis: arch
targ... | {
"id": "22287f22-7107-4afb-bd3b-7932060297a0",
"created_at": "2026-08-24 23:30:18.649563+00:00",
"db_md5": "703dcb6c26df13e9ce66232511a448d0",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "7d4cfd18-6cf0-4f58-b0f4-c018e8d626f7",
"tier": "easy",
"dataset_id": "e6",
"status": "succe... | {
"score": {
"mean_loss": 2.2558958530426025,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.20450451225042343
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.852726697921753,
"example_count": 60,
... |
222b6722-2e4e-4e8d-9081-aab66ee07c7d | easy | chad-atexpedient | 2026-08-14 20:41:46.380947+00:00 | succeeded | 526acf6733e7979a4b0a7fe8439db15cf5f8e86fae9839b620dd70f0ff30f565 | 17,094 | null | """X49: X18 with batch_size=1024 for more consistency pairs."""
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,
)
WI... | {
"id": "222b6722-2e4e-4e8d-9081-aab66ee07c7d",
"created_at": "2026-08-14 20:41:46.380947+00:00",
"db_md5": "63606e192e2c447fe05cecead76d4ede",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "71bad07e-3794-448e-92cd-c9d5b9f9b5cb",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.2023112811229435,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2216922306693725,
"example_count": 60... |
22304cbb-c1ad-44f9-ba14-951aafddf674 | easy | isaac0804 | 2026-08-05 08:15:15.319482+00:00 | succeeded | d629acb71440ad7d10ec0bd2f4b70f33b583ab141aaaa5bf0c03a04d6bcd0655 | 11,974 | null | """Weight-tied looped Transformer (v6 core: damped gate, 2 sublayers/iteration)
plus two cheap, legal additions from the standard looped/recurrent-transformer
toolkit, both gated behind module-level flags for fast local ablation:
INPUT INJECTION (``USE_INPUT_INJECTION``): at every loop iteration, re-inject
a lear... | {
"id": "22304cbb-c1ad-44f9-ba14-951aafddf674",
"created_at": "2026-08-05 08:15:15.319482+00:00",
"db_md5": "cdaa12ec368e952ab41a81edd8282593",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "8042e774-03d3-4409-90c4-028fb3ac8b5f",
"tier": "easy",
"dataset_id": "e4",
"status": "succe... | {
"score": {
"mean_loss": 2.1000036914604903,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0071875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.114354984167505,
"example_count": 1200,
... |
2230bc54-93be-45fe-8ed4-708d584314d7 | easy | shreyash-chonkie | 2026-08-31 23:31:23.223916+00:00 | succeeded | 6eff5fcee60bfa79c0bb876258e3055f87a207f3de9c9a94b77c3a0be38c64f3 | 15,659 | null | """Recurrent attention with improvement-weighted pool feedback."""
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_mo... | {
"id": "2230bc54-93be-45fe-8ed4-708d584314d7",
"created_at": "2026-08-31 23:31:23.223916+00:00",
"db_md5": "543829cb6abefa025e973d133e3406a6",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "7c7f9ab1-f1ab-4c4e-87f8-268bbf75c816",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 2.080645203590393,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999828636646
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1711931228637695,
"example_count": 100,... |
223113b5-f8bc-48ba-a892-90b406c90af9 | easy | yashkant | 2026-08-23 20:17:20.748125+00:00 | failed | 35db76f466990cd8e74833393b591825bd4ccfdd7634c8698c959b560bb138b1 | 11,550 | null | """One Layer Deeper submission: CRT residue-register machine with a learned squaring
transition and a Fourier "clock" bilinear reconstruction readout.
Idea (end-to-end learned; nothing about the task is hard-coded):
* State = per-modulus residue registers over a fixed bank of small primes. The input
integer x is... | {
"id": "223113b5-f8bc-48ba-a892-90b406c90af9",
"created_at": "2026-08-23 20:17:20.748125+00:00",
"db_md5": "54b80a60a8aca82369ec363db1993af0",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "e186b1c5-fbcf-42f4-b2b0-ce1e69fe3028",
"tier": "easy",
"dataset_id": "e1",
"status": "failed"... | null
|
22366102-14d8-409d-869e-0c64d81f8fc5 | easy | DDanlov | 2026-08-16 00:12:53.120042+00:00 | succeeded | 9dc7d0a8a1b0fb8e3f6a502893bbf46dd0534671810d611796e8a743b5340e8c | 16,544 | null | """
Production Port-Hamiltonian RoPE Dynamic Submission
Config ID: 5 | Name: Medium_S6_d768_Balanced
dim=768, stages=6, heads=12, dt=0.1, lr=0.007, wd=0.04, opt=shampoo
"""
from __future__ import annotations
import math
from typing import Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as... | {
"id": "22366102-14d8-409d-869e-0c64d81f8fc5",
"created_at": "2026-08-16 00:12:53.120042+00:00",
"db_md5": "b50e26665b921647738f124e0e6642da",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "0c9cb709-8fe7-4312-bab1-9b5e411c4713",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.205814624975107,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0079166666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.248492681392105,
"example_count": 600,
... |
223672d1-a297-4f1c-8b69-58aeeb93785d | easy | jordanrubin | 2026-08-03 21:00:51.543155+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": "223672d1-a297-4f1c-8b69-58aeeb93785d",
"created_at": "2026-08-03 21:00:51.543155+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "db6d4255-a872-47f0-9ace-66c12f3f35dd",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.39995121625902,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.2020833301357925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.33300342017073,
"example_count": 300,
... |
223aa461-988a-4162-b622-e2e146f8da0d | easy | jordanrubin | 2026-08-08 09:39:59.609344+00:00 | succeeded | 078cd8887c504bdcfcf580d442de465beb5b3a0ea5e1726a6c95fd2267731c91 | 16,615 | 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": "223aa461-988a-4162-b622-e2e146f8da0d",
"created_at": "2026-08-08 09:39:59.609344+00:00",
"db_md5": "1f66d0c9d8b2addfecdef9a9aa694dae",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "541dd77b-6bad-4758-8890-243696b4b7f4",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.6900046035899146,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.1774999942475309
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.707041072513168,
"example_count": 300,
... |
22426a70-6a97-4696-94af-bcc268ff6f8b | easy | jordanrubin | 2026-08-07 01:33:31.022678+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": "22426a70-6a97-4696-94af-bcc268ff6f8b",
"created_at": "2026-08-07 01:33:31.022678+00:00",
"db_md5": "497d08edf33fa52921753c6fa8942084",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "93836a85-866c-4847-a452-d9e3e8c9f990",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.26245204147677,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01083333301357925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.291995466557319,
"example_count": 600,
... |
224533ab-5fa6-4f21-a6fe-66452ace23ca | easy | anirudh-chakravarthy | 2026-08-12 01:46:27.360318+00:00 | succeeded | de0a9a0e4e1f219a6746bc76ba9ec301a1f80cda51014067f85774bc9405d3d3 | 14,668 | null | """A learned state transition applied once per requested time step."""
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,
asser... | {
"id": "224533ab-5fa6-4f21-a6fe-66452ace23ca",
"created_at": "2026-08-12 01:46:27.360318+00:00",
"db_md5": "bc331106422177d7bd330d58c1e6a57f",
"submitter": "Anirudh S Chakravarthy",
"github_login": "anirudh-chakravarthy",
"run_id": "e6c9a75e-f3ee-4aca-b6e4-01d21ccd9a69",
"tier": "easy",
"dataset_id": "... | {
"score": {
"mean_loss": 2.208517074637334,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333345750967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.222476280323593,
"example_count": 600,... |
2247eb58-d713-49a3-806b-93f6d484809e | easy | erdavis0 | 2026-08-21 13:55:00.231640+00:00 | succeeded | 5bf0d5def60fcd3ccb8e375df6f6e943236f7db572ccb2b2c679fbf82745c1b0 | 6,171 | null | """Absolute-position output-slot encoder/decoder 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,
)
WIDTH = 64
HEADS = 4
EXP... | {
"id": "2247eb58-d713-49a3-806b-93f6d484809e",
"created_at": "2026-08-21 13:55:00.231640+00:00",
"db_md5": "e998fc08007161251583d6a17dcd3b43",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "18b3568c-27ea-4f81-b404-1b87055e5889",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",... | {
"score": {
"mean_loss": 7.4763500690460205,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07364865392446518
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.6702561378479,
"example_count": 60,
... |
224d6c26-26ca-45cd-b3ee-ce2d9823ec15 | easy | DDanlov | 2026-08-24 04:21:26.702155+00:00 | succeeded | 6dd93f7cafb5c69106ced329cdeaa3b7880aad32844e04201abeaa219435a11c | 40,841 | null | """
Unified Hybrid Masked Deliberation Model (44.0M Parameters) - Pure Skipless, No Cross-Attention
Architecture:
- Scale: dim=1536, num_heads=24, d_ff=6144 (Exactly 43,978,752 parameters)
- Hybrid Prefix-Bidirectional + Causal Target Attention Mask
- Token Layout: [Prompt (L_p), 4 Pause Registers, Target Tokens (L_tgt... | {
"id": "224d6c26-26ca-45cd-b3ee-ce2d9823ec15",
"created_at": "2026-08-24 04:21:26.702155+00:00",
"db_md5": "89ded7efaa19552b2c5de229d32afa70",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "e22fa47a-9dac-47a5-b29a-356f2f11c131",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.3213099946936833,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004166666691501935
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.359858189997652,
"example_count": 600... |
2250da67-8c78-4f7f-8d0c-b7129ea3de32 | easy | DDanlov | 2026-08-25 23:34:35.605059+00:00 | succeeded | fb1a6a1896cdf525dac84b86831ef3c4e13ac35ce3a3c5696db5c14a7b52f7b7 | 38,939 | null | """
Architecture 1: Two-Stage Pure Skipless Deliberation Model with Cross-Attention-First Decoder
Static Input Re-Injection ($z_0$ conditioning) + Unconstrained SwiGLU + SUO Initialization.
Design Invariants:
1. Deliberation Core: Multi-Head Self-Attention (Mimetic Q-K) + Unconstrained SwiGLU FFN (No Givens, No Rotati... | {
"id": "2250da67-8c78-4f7f-8d0c-b7129ea3de32",
"created_at": "2026-08-25 23:34:35.605059+00:00",
"db_md5": "105b90c1a6f5b351cb6fdf327e16d50c",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "8b27ee53-fd2c-489e-bf3d-22c42ffea595",
"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... |
2254f5cc-4482-4977-aa27-ef20ee363892 | easy | Alan-Qin | 2026-08-18 17:46:51.924760+00:00 | succeeded | 36fac39034cdc8e14c2033695c54fc10eeac36df55a1e4dbff401db88dc06916 | 17,899 | null | """One-hot CRT loop + sibling-consistency supervision (v24).
Architecture: per-band one-hot residues (parameter-free encoding of x),
a learned per-band transition matrix W_m applied T times (STE one-hot), and a
shared digit decoder read at every loop step.
Training loss (token_training_loss): besides the final-answer... | {
"id": "2254f5cc-4482-4977-aa27-ef20ee363892",
"created_at": "2026-08-18 17:46:51.924760+00:00",
"db_md5": "4d752e5f846a8f7f3f8aef9a12f5216d",
"submitter": "Zeyu Qin",
"github_login": "Alan-Qin",
"run_id": "15e08217-5e79-4edb-a567-f45ec6490981",
"tier": "easy",
"dataset_id": "e3",
"status": "succeede... | {
"score": {
"mean_loss": 3.37575548333339,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09687500045634806
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.9506914385938825,
"example_count": 800,
... |
2257fb8a-941e-4aae-9be7-fb6299839641 | easy | DESU-CLUB | 2026-08-24 11:12:21.397106+00:00 | succeeded | 5b4cb40e009c5f829aa07af7f80020a02bbbc5732624226aaf678cd8ed26671b | 10,167 | null | """Huginn-style looped Transformer whose iteration count is read from the T token.
prelude -> [core applied T times, input re-injected each step] -> coda
Forked from experiments/_baselines/adamw_transformer.py; the Block, RMSNorm and
optimizer are unchanged so that the only varied axis is the depth mechanism.
"""
fr... | {
"id": "2257fb8a-941e-4aae-9be7-fb6299839641",
"created_at": "2026-08-24 11:12:21.397106+00:00",
"db_md5": "da4b3352acf106f091459ec563e69cd7",
"submitter": "Low Keng Hoong, Warren",
"github_login": "DESU-CLUB",
"run_id": "632b7efb-0aa7-475e-89f6-ecb60886cb33",
"tier": "easy",
"dataset_id": "e6",
"sta... | {
"score": {
"mean_loss": 8.483613007689176,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07364865316933877
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.829289436340332,
"example_count": 60,
... |
225a4f06-50eb-4d2f-b828-2bf6d07e0e70 | easy | KaustubhKumar05 | 2026-08-25 16:39:14.377330+00:00 | succeeded | 66c0d17ac95330bfa44c7d3f372988009e3ed4e7cbcca72adeca514afd3de49f | 14,717 | null | """Neural-GPU-style conv GRU, v6: v2s + relaxed weight sharing.
Four cell copies cycled by iteration index (Kaiser & Sutskever's "crucial"
trick for multiplication). A pull-together penalty on copy divergence is
returned as auxiliary and added to the loss, annealing the copies toward one
shared cell while keeping earl... | {
"id": "225a4f06-50eb-4d2f-b828-2bf6d07e0e70",
"created_at": "2026-08-25 16:39:14.377330+00:00",
"db_md5": "cebdbe1f9f283b2f115dc4641a0cdd29",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "1d7b12f1-815d-4d8d-b021-28693e1ed8cf",
"tier": "easy",
"dataset_id": "e6",
"status": "succee... | {
"score": {
"mean_loss": 2.9735372066497803,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02342342445626855
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8710529804229736,
"example_count": 60,... |
2266e2f8-0c82-453d-b6b3-e5f822079872 | easy | sapient-sapiens | 2026-08-17 02:38:52.227576+00:00 | succeeded | 96d137d90e52e00d4bfe7084329085f05efee0981345708cdf647be0670dae02 | 15,120 | null | """Fixed-4 recurrence candidate: v70 source without adaptive halt.
Parent: act64_e6_w1024_dh4_L6_wd1_v70 (width 1024, d_h=4, L=6, WD=1.0).
Primary variable: adaptive ACT over 1--64 (typically hard-halt 2) versus a
fixed four applications of the shared core. The halt controller is removed.
T remains an ordinary prompt ... | {
"id": "2266e2f8-0c82-453d-b6b3-e5f822079872",
"created_at": "2026-08-17 02:38:52.227576+00:00",
"db_md5": "a360270b1bd62e5c960d2cc497021af4",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "4f5970b4-24c2-4089-9331-ef2478999e16",
"tier": "easy",
"dataset_id": "e9",
"status": "su... | {
"score": {
"mean_loss": 7.300091194566011,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.046666668504476545
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.891551494598389,
"example_count": 90,
... |
2267d28d-8a7c-49ce-9763-648f13034e62 | easy | kushasareen | 2026-08-12 02:24:32.130445+00:00 | succeeded | c43e4b042de597d8d7701c23d1f865189e890ec82136dac2f2a5f5481eafe79c | 6,884 | null | """v2 weight decay bracket: wd=0.3.
T-conditioned recurrent Transformer with the measured stability fixes:
- 2 untied pre-blocks (pre-norm), 1 untied post-block, tied loop block
- prompt re-injection every loop iteration (gradient + information path)
- UT-style double post-norm inside the loop block (state bounded at ... | {
"id": "2267d28d-8a7c-49ce-9763-648f13034e62",
"created_at": "2026-08-12 02:24:32.130445+00:00",
"db_md5": "26baffc7dcaed00bd8150e5ab2ec501f",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "c81ebbd7-9062-44a4-b7cd-af1a80961ae1",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 8.051304463293668,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005416666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.30338464831023,
"example_count": 600,
... |
226afd3b-2735-4bfe-abfb-1bc2e2c30093 | easy | pculaf | 2026-08-25 14:19:59.091090+00:00 | succeeded | b3eb3fb8a08cc858183f73e7a647b0710c03cb193362f6b9cce0ea5e9c67ced0 | 8,288 | null | """Four-loop post-RMSNorm Transformer for One Layer Deeper."""
from __future__ import annotations
from collections.abc import Iterable
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_m... | {
"id": "226afd3b-2735-4bfe-abfb-1bc2e2c30093",
"created_at": "2026-08-25 14:19:59.091090+00:00",
"db_md5": "06d3398262efcdfa4bf712724e235f4d",
"submitter": "Pavle Culafic",
"github_login": "pculaf",
"run_id": "cb059e97-1050-4662-9d4a-10e9ed820f7a",
"tier": "easy",
"dataset_id": "e9",
"status": "succe... | {
"score": {
"mean_loss": 1.9658262133598328,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666828095913
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9859439134597778,
"example_count": 90... |
226c36b2-27fc-44ff-bd49-7f512382440a | easy | harmya | 2026-08-07 09:44:57.499645+00:00 | succeeded | 8c07f6413812ce87dac1f1bfa1c0331cac82c150a7fcd063510f86a573369689 | 36,771 | null | """Character-lane model with a combinatorial assignment teacher.
Learns repeated modular squaring end-to-end from token sequences. Digit
tokens are embedded as angles on the unit circle, mapped to complex lanes
by a learned linear map, and advanced by a parameter-free complex-squaring
recurrence. A learned scalar dept... | {
"id": "226c36b2-27fc-44ff-bd49-7f512382440a",
"created_at": "2026-08-07 09:44:57.499645+00:00",
"db_md5": "3c939e7fe5480abe1a7f39385b5b002b",
"submitter": "Harmya Bhatt",
"github_login": "harmya",
"run_id": "f77d1e52-eeb0-4c16-baaa-62dcb0920c9b",
"tier": "easy",
"dataset_id": "e1",
"status": "succee... | {
"score": {
"mean_loss": 2.4886664152145386,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08166666701436043
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.134493589401245,
"example_count": 100,... |
22750d16-8d95-4b32-9b22-cbe64caa32b1 | easy | velocizapkar | 2026-08-16 06:53:29.366502+00:00 | succeeded | 684e4ce71f1429a11cd2d4c102654ffe396fc6c3dc2bcda9be300d8b26fec6f4 | 9,485 | null | """Matched standard bidirectional Transformer control.
This is deliberately an ordinary fixed-depth Transformer: learned token and
absolute-position embeddings, full softmax self-attention, pre-RMSNorm,
SwiGLU, residual connections, a final RMSNorm, and an untied vocabulary head.
It has no recurrence, convolution, scr... | {
"id": "22750d16-8d95-4b32-9b22-cbe64caa32b1",
"created_at": "2026-08-16 06:53:29.366502+00:00",
"db_md5": "007f76a50ca36a338a51d681a1cf8009",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "e8a1a180-e184-4845-88dd-e09e25c10315",
"tier": "easy",
"dataset_id": "e6",
"statu... | {
"score": {
"mean_loss": 5.981003522872925,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.061936939135193825
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.898272514343262,
"example_count": 60,
... |
227e0fdd-d61a-43d6-99d0-6c6e9a9f4a45 | easy | sankalp1999 | 2026-08-04 11:37:49.427390+00:00 | succeeded | abf1a029ea21ced22e692437f0f942aef66b4eab7b30ff925f35590ac5655a74 | 4,869 | null | """Parametrized flattened-grid residual MLP for E1 sweeps."""
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,
assert_model_state,
)
... | {
"id": "227e0fdd-d61a-43d6-99d0-6c6e9a9f4a45",
"created_at": "2026-08-04 11:37:49.427390+00:00",
"db_md5": "2a1b653a83dd176bc4b8423a5034ad90",
"submitter": "Sankalp ",
"github_login": "sankalp1999",
"run_id": "ef8896b7-4e38-42f5-9b0b-0913af92cf28",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 2.0109644532203674,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04666666593402624
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.915906548500061,
"example_count": 100,... |
228053cf-5480-46c0-8a23-6205c68f910f | easy | yunjiangster | 2026-08-27 00:03:37.581768+00:00 | succeeded | 06bcf5264eb2767d31461b124df405a1811306257c2a92a7abf0733ec1a9dc5b | 18,378 | null | """Position-shared recurrent grid probe for One Layer Deeper.
This diagnostic deliberately trains only T=1 rows on E5. It tests whether a
small shared local machine can learn a transferable one-step map across varying
moduli. It contains no full-tape dense lookup path and no task solver.
"""
from __future__ import a... | {
"id": "228053cf-5480-46c0-8a23-6205c68f910f",
"created_at": "2026-08-27 00:03:37.581768+00:00",
"db_md5": "47211de9a78da3fd9ba42a2e35d8159b",
"submitter": "Yunjiang Jiang",
"github_login": "yunjiangster",
"run_id": "6c0bc496-097b-486b-8fc0-06acee324a36",
"tier": "easy",
"dataset_id": "e5",
"status":... | {
"score": {
"mean_loss": 2.390426637273811,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0029166667039195694
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.395210524845551,
"example_count": 600... |
228416e1-44e1-45df-9e40-ba5c40b2fae4 | easy | sapient-sapiens | 2026-08-23 17:47:02.104335+00:00 | succeeded | f3f6c084895788109ece7b900d12fa2d816701c160795bba6fd837a573e3c66a | 16,078 | null | """Scaled six-digit two-phase scan with adaptive recurrent reduction.
The model uses categorical learned embeddings and marker-derived alignment only.
No N, X, or T field is decoded into a number. All recurrent sweep readouts are
supervised exclusively against the same final answer.
"""
from __future__ import annota... | {
"id": "228416e1-44e1-45df-9e40-ba5c40b2fae4",
"created_at": "2026-08-23 17:47:02.104335+00:00",
"db_md5": "c808e15f43bcbff83e1b5d4630e063f0",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "d2edb513-5b50-45ea-8929-cdb047b5c031",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.152576730286457,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000012417634
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.153511720788853,
"example_count": 600,... |
2287d139-67b5-41d1-8582-cc7b21b3c996 | easy | SunstarKK | 2026-08-04 21:40:40.981054+00:00 | succeeded | 10492a113e961455e0a3d32cede7636e97cf8b7c6926ad5abec902316934080d | 15,413 | null | """Compute-aware recurrent-candidate model for One Layer Deeper."""
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_m... | {
"id": "2287d139-67b5-41d1-8582-cc7b21b3c996",
"created_at": "2026-08-04 21:40:40.981054+00:00",
"db_md5": "e36577fb1e5bb1a9fc5742dad02a3f6f",
"submitter": "SunstarK",
"github_login": "SunstarKK",
"run_id": "3080c098-3e0a-4672-8d05-3dcb23e2b697",
"tier": "easy",
"dataset_id": "e1",
"status": "succeed... | {
"score": {
"mean_loss": 3.4433810710906982,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03333333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.6315701007843018,
"example_count": 100... |
228dfe6b-138f-4edf-913e-401fe54ae363 | easy | rudygt | 2026-08-04 19:33:11.094089+00:00 | succeeded | 8c91cf07ee48421de4950025b933d38d318079229e96d5e6c84ea1f16bda05cb | 20,499 | null | """OL-2 — quantized tied-step recurrence in the contract's digit I/O ("contract
wrapper", LOG.md 2026-08-04 T-series recipe).
Pipeline: digit-token prompt -> 1-block bidirectional encoder -> attention-pooled
h0 -> tied step block (2x RMSNorm-MLP residual + zero-init complex-square path),
unrolled T times (T parsed fro... | {
"id": "228dfe6b-138f-4edf-913e-401fe54ae363",
"created_at": "2026-08-04 19:33:11.094089+00:00",
"db_md5": "811e04c6d5c685787bf7861f65b25de0",
"submitter": "Rudy Alvarez",
"github_login": "rudygt",
"run_id": "1d1393d9-ec91-4e73-a1a0-476aa1b5c4ec",
"tier": "easy",
"dataset_id": "e1",
"status": "succee... | {
"score": {
"mean_loss": 9.476113319396973,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06333333440124989
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 12.569642066955566,
"example_count": 100,... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.