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 |
|---|---|---|---|---|---|---|---|---|---|---|
1e239227-e5c4-437c-bf54-3bb5cc10795b | easy | oupadhyay | 2026-08-08 09:47:52.813436+00:00 | succeeded | 939ee0aa97f9b51b9e27e7880e00b2226a7fb9cf643866d186e01ac2efb02018 | 7,426 | null | """Generic universal recurrent Transformer candidate."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
D = 128
HEADS = 4
FFN = 384
SCRATCH = 32... | {
"id": "1e239227-e5c4-437c-bf54-3bb5cc10795b",
"created_at": "2026-08-08 09:47:52.813436+00:00",
"db_md5": "2eb85e17dfaf843a20765bf460ac3cbf",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "1f5e4fd2-383d-4648-b04f-672f2c47e90d",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 3.4353739566564023,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666896392902
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.588104724884033,
"example_count": 600... |
1e247687-ea5b-4f86-bdb0-f58184ca9f22 | easy | sapient-sapiens | 2026-08-31 22:15:36.336558+00:00 | succeeded | 22a881b9f8358420d4ee44cc1d36bc6128ccccf06979b79c79c1cc4b971f336e | 16,043 | null | """One-block nested E5 model: four learned Horner calls per square, eight squares."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_mod... | {
"id": "1e247687-ea5b-4f86-bdb0-f58184ca9f22",
"created_at": "2026-08-31 22:15:36.336558+00:00",
"db_md5": "f3b8498ad0d2fb103a413f576400c1f9",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "2c8371a1-3f07-4198-aca7-04343ce071df",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.158227125596243,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0054166666790843015
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.16006171970624,
"example_count": 600,... |
1e252176-866f-4361-a1f0-9f1dc1c13ad9 | easy | Koenneker | 2026-08-05 09:10:49.014716+00:00 | succeeded | 7bf30e19d90e7f6cadb4ad67d0b2cd51634a4fd2d29a507be6f840a0dd78f8c1 | 3,628 | null | """Basic single-pass Transformer with PyTorch AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 128
NUM_HEADS = 4
... | {
"id": "1e252176-866f-4361-a1f0-9f1dc1c13ad9",
"created_at": "2026-08-05 09:10:49.014716+00:00",
"db_md5": "eaac69fe345f82973fcf392c504a40ff",
"submitter": "Ferdinand Könneker",
"github_login": "Koenneker",
"run_id": "dbac71e1-e6ad-43a0-aa7e-5772c7a35670",
"tier": "easy",
"dataset_id": "e4",
"status"... | {
"score": {
"mean_loss": 2.5564008813929275,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0032291666666666666
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.733548319512003,
"example_count": 12... |
1e2ae27b-c3c5-44d0-ac19-c10d9587891d | easy | Ranoobaba | 2026-08-20 02:26:06.911646+00:00 | succeeded | 526adfba9c8750f6937251f7e1f9ca0f54714a6c5d38ddde115949f5b79ec985 | 6,640 | null | """Looped transformer, iteration count conditioned on T read from the prompt.
Core idea: learn one latent "squaring step" operator and apply it T times.
Depth generalization comes from the loop structure itself; the model never
needs more parameters for deeper problems, only more iterations.
"""
from __future__ impor... | {
"id": "1e2ae27b-c3c5-44d0-ac19-c10d9587891d",
"created_at": "2026-08-20 02:26:06.911646+00:00",
"db_md5": "60db6aade93ea21fd0088b18751ca1b2",
"submitter": "Syed Rayyan Ali",
"github_login": "Ranoobaba",
"run_id": "7719e1ec-6557-4d44-94d2-b93bc338a41c",
"tier": "easy",
"dataset_id": "e3",
"status": "... | {
"score": {
"mean_loss": 6.029870278341619,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.092912283974833,
"example_count": 800,
"e... |
1e2b157f-2a1f-4fe3-835a-e161482eb32d | easy | Alan-Qin | 2026-08-18 17:19:49.590694+00:00 | succeeded | 0bc5d50fb167839ac0b3414e658dfef674352a761e2d3917c11b42c088a3bbec | 16,654 | 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": "1e2b157f-2a1f-4fe3-835a-e161482eb32d",
"created_at": "2026-08-18 17:19:49.590694+00:00",
"db_md5": "cf3600bf05b023d27d45e55095750f32",
"submitter": "Zeyu Qin",
"github_login": "Alan-Qin",
"run_id": "168b5bf6-0555-4ae3-bac3-7f98c30dd8f2",
"tier": "easy",
"dataset_id": "e10",
"status": "succeed... | {
"score": {
"mean_loss": 0.02558335615321994,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.9898989796638489
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.013305402360856533,
"example_count": 1... |
1e2bacc3-6730-4a9f-a7a5-19930cb2a628 | easy | SunstarKK | 2026-08-18 12:35:39.988801+00:00 | succeeded | ce344aee19105b4948f37f574c603c224916183493ab1f742328aef4cbba7755 | 13,789 | null | """Learned multiscale digit pyramid for One Layer Deeper."""
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,
)
PAD ... | {
"id": "1e2bacc3-6730-4a9f-a7a5-19930cb2a628",
"created_at": "2026-08-18 12:35:39.988801+00:00",
"db_md5": "d53da454efb55f1ac47befb27f242bd6",
"submitter": "SunstarK",
"github_login": "SunstarKK",
"run_id": "8873c27a-0b66-464d-83da-20ce3578883d",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 4.51008512074057,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.020000000037252906
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.387566472382823,
"example_count": 600,
... |
1e3239e2-ab02-42bf-96ff-55a7737f9de6 | easy | alirezashirvani-jr | 2026-08-31 00:31:07.572870+00:00 | succeeded | d9be1270c66f4cec678728ecb8d3888d74765a4fd36126c152a8eb335ac95f65 | 19,728 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PERIOD_MINIMUM = 3
PERIOD_MAXIMUM = 100
EX... | {
"id": "1e3239e2-ab02-42bf-96ff-55a7737f9de6",
"created_at": "2026-08-31 00:31:07.572870+00:00",
"db_md5": "d28e9bd3af3854ff3baff14eaf466501",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "7a6fa8a1-6d65-46cf-aae9-e7dbbeb850bb",
"tier": "easy",
"dataset_id": "e4",
... | {
"score": {
"mean_loss": 2.271340177227133,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04635416653007269
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6414999426094288,
"example_count": 1200... |
1e390917-e36c-4075-9b35-f1dc40284318 | easy | DDanlov | 2026-08-19 21:36:06.608720+00:00 | succeeded | 48a83d0a4e040d27d70b1e98d73a32993304e663adf7b5cd7e67fbc7b19d3bc2 | 17,787 | null | """
Dynamic Submission for One Layer Deeper API Benchmark
Config: dim=1536, num_heads=32, d_ff_mult=4, fixed_trec=156
"""
from __future__ import annotations
import math
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:
... | {
"id": "1e390917-e36c-4075-9b35-f1dc40284318",
"created_at": "2026-08-19 21:36:06.608720+00:00",
"db_md5": "c1dfc62d6a01fca33c7e50684eae9f95",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "ec39ffe5-2ff7-4a45-9a2d-8442e278af60",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.4236361472394297,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333345750968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5081524518840514,
"example_count": 60... |
1e43b74f-b3cf-423d-8196-3b68e5c867ad | easy | lzy54 | 2026-08-29 09:46:11.886053+00:00 | succeeded | 68bd0b087634f84c979703698cb1a9a0c7622a21df5d3ccc1ef0bb9753f7a2c1 | 6,234 | null | """R3: compact transfer of the official looped-tf arithmetic recipe.
The recurrent state starts at zero. Each of four loops re-injects the encoded
problem, then applies the same three-block d=256 core. The original repository
uses d=256, 8 heads, three shared blocks, batch 64, and AdamW at 1e-4 for its
addition sett... | {
"id": "1e43b74f-b3cf-423d-8196-3b68e5c867ad",
"created_at": "2026-08-29 09:46:11.886053+00:00",
"db_md5": "cf1a9a41a43c95cccb3a117c9c9ae4de",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "d2b75132-6058-4306-8d5a-77058666caf2",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.1634542683147497,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.169600239783659,
"example_count": 600,
... |
1e4eac0e-fb09-463d-85d7-7f098e455f76 | easy | mnida | 2026-08-24 19:53:44.797577+00:00 | succeeded | f75c72b7e7648ef50b1915b5b147d1c21cf9ad13d608230f65b8de5587bd72f7 | 15,686 | null | """Known-step projective homogeneous CP recurrence with PyTorch AdamW."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
... | {
"id": "1e4eac0e-fb09-463d-85d7-7f098e455f76",
"created_at": "2026-08-24 19:53:44.797577+00:00",
"db_md5": "2a20ef30768835b907a08c35baa7ec15",
"submitter": "mnida",
"github_login": "mnida",
"run_id": "31147d29-844d-4e74-8450-cf21d608b328",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 4.628069480730671,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004583333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.076403557986956,
"example_count": 600,... |
1e54361c-c4e5-421d-aad4-6fc46fe517f5 | easy | DDanlov | 2026-08-14 07:45:01.713994+00:00 | succeeded | 52c8f780bd80643100bd4df9fb6975672fa53422c49f0c797c4d45f15cba1465 | 21,975 | null | import math
from typing import Optional
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark import Submission, assert_model_state, OptimizerBundle
except ImportError:
try:
from client import Submission, assert_model_state, OptimizerBu... | {
"id": "1e54361c-c4e5-421d-aad4-6fc46fe517f5",
"created_at": "2026-08-14 07:45:01.713994+00:00",
"db_md5": "02f5b2add60f865a1a867fb13b8d4762",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "a7748ea5-62f3-45ac-8069-8a24f838ff57",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.833085265187732,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8330883342321678,
"example_count": 600,
... |
1e560e8a-95b2-480c-af8b-5460a8146b4a | easy | chad-atexpedient | 2026-08-09 12:57:42.785243+00:00 | succeeded | 26c3233e941228813c8a030f6a4455ac6d2236a319de4d8ca7959d908af86717 | 17,666 | null | """X11: full-budget direct T1 transition identification.
The mutable register is initialized from x once. Every subsequent application
of the tied cell receives only the current LSD-first digit state and immutable
N digits. Requested T controls only the number of applications. The state
logits are also the answer ... | {
"id": "1e560e8a-95b2-480c-af8b-5460a8146b4a",
"created_at": "2026-08-09 12:57:42.785243+00:00",
"db_md5": "af999aa7ec366e8e9fb1e50aeff2ebf7",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "fd3ab74a-e916-423d-bb6b-0aac86a5cdc7",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 4.730082841204093,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333426465591
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.119244575500488,
"example_count": 600,... |
1e5681f5-2a7e-4c93-9cc6-427711a36a8f | easy | Alan-Qin | 2026-08-16 15:56:49.784182+00:00 | succeeded | 8a0cad7297fcacc5e8f0ddb693bb03a4f7c1062f89c78c710f709489fda68eb1 | 9,008 | null | """Prompt-conditioned fixed-slot recurrent transformer (v22).
This candidate is a task-agnostic information-flow probe. It deliberately
does not parse N or X digit widths, does not construct a decimal state, and
does not hard-code a modular arithmetic operation. The whole input prompt is
encoded once into immutable ... | {
"id": "1e5681f5-2a7e-4c93-9cc6-427711a36a8f",
"created_at": "2026-08-16 15:56:49.784182+00:00",
"db_md5": "11bca38f76d4f8be3dfad146ff1edffa",
"submitter": "Zeyu Qin",
"github_login": "Alan-Qin",
"run_id": "1205323b-400f-40fa-b01d-a70b7b0b6e29",
"tier": "easy",
"dataset_id": "e5",
"status": "succeede... | {
"score": {
"mean_loss": 5.498215366494284,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.389159305213278,
"example_count": 600,... |
1e5764b7-701f-48f3-a0dd-aa25fa3b3b7f | easy | yashkant | 2026-08-14 16:01:27.199178+00:00 | succeeded | e57cec3808f343eb197d7f49dbe04f00dad79836504bdedda5200c7630cc41b2 | 33,286 | null | """Round1786 evolutionary crossover of two controlled survivors.
The Round1780 W8/D2 carrier keeps its winning contrastive, exactly
antisymmetric learned set codes. The candidate additionally removes each
initial and recurrent Dykstra correction residual's coordinate mean before
the unchanged tanh bound, combining Ro... | {
"id": "1e5764b7-701f-48f3-a0dd-aa25fa3b3b7f",
"created_at": "2026-08-14 16:01:27.199178+00:00",
"db_md5": "e6fab685e84a73c72c50180ab67b5830",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "f83400e5-ff34-43a4-81a3-d3e4fc5f7a4b",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 2.314720869064331,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03354166820645332
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5257906913757324,
"example_count": 300,... |
1e66305a-9f1c-4306-bb7e-ac20baf6c163 | easy | ipsa-loquitur | 2026-08-27 01:42:03.706484+00:00 | succeeded | 83f68730e4e43f5eb3c07898f599ef006fc8abeb0f38d2cd386151ecce2b499b | 15,929 | null | """Standalone NetF-V1 submission candidate for One Layer Deeper.
The learned rule is shared across bit positions, modular-square iterations,
moduli, and tape widths. During training, exact double-and-add states provide
dense one-step supervision. During evaluation, the model receives no state
corrections and rolls out... | {
"id": "1e66305a-9f1c-4306-bb7e-ac20baf6c163",
"created_at": "2026-08-27 01:42:03.706484+00:00",
"db_md5": "dab86dfb4159e53a6251a25be1bbabfa",
"submitter": "ipsa-loquitur",
"github_login": "ipsa-loquitur",
"run_id": "86fa0e8e-e09a-4697-bd72-6ab0ecaf3bf8",
"tier": "easy",
"dataset_id": "e3",
"status":... | {
"score": {
"mean_loss": 19.139833602800763,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09250000037252903
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 20.36516645188386,
"example_count": 800,... |
1e67fb9d-7292-45f1-bd88-b4258110efe9 | easy | lpbb | 2026-08-19 11:51:46.470627+00:00 | succeeded | e5c36210ef7a2cc02ed2a8318eaf4016354209f5411142438e7dc0177d85717c | 9,299 | null | """S0 + RoPE + abacus embeddings + dropout, with a looped block: instead of
`num_layers` independent layers, a block of `num_layers` unique layers is run
`num_repeats` times with shared weights (e.g. 12 layers x 4 repeats = depth 48
but only 12 layers' worth of parameters). Weight tying across repeats forces
the block ... | {
"id": "1e67fb9d-7292-45f1-bd88-b4258110efe9",
"created_at": "2026-08-19 11:51:46.470627+00:00",
"db_md5": "a0755f04bc673fc370bf0581fd11372c",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "627ec601-488e-4ab2-b788-cd4d12ea9847",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 4.99759578704834,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03999999910593033
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.508913516998291,
"example_count": 100,
... |
1e6e0619-36d6-454a-8ea9-255f3bbbf275 | easy | chad-atexpedient | 2026-08-20 15:43:26.625115+00:00 | succeeded | 85f4d241b753d69a1e8be75cacbe5601bb33dc0c757669c5ae3b07395adea284 | 24,438 | null | """X116: X105 with same-row deeper-rollout self-distillation.
Continuation-chain matches are sparse within a random batch. X116 uses X105's
already-computed four training rollouts as a detached teacher: when a deeper state
predicts the supplied endpoint with lower token CE than the requested-T state, the
requested log... | {
"id": "1e6e0619-36d6-454a-8ea9-255f3bbbf275",
"created_at": "2026-08-20 15:43:26.625115+00:00",
"db_md5": "bd99df5a29303f8194f31a4a2199badb",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "90810478-ff87-433c-934f-ba29c4f80c6c",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.1896546781913395,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.012500000012417634
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.206651098525043,
"example_count": 600... |
1e6fe252-4bed-4fc4-a045-8fa514f7e70e | easy | erdavis0 | 2026-08-24 12:52:42.295210+00:00 | succeeded | dfd006ad8396f639cb082005e79403fa55d871cd15b25e6f39a8f2ebe7b70687 | 30,151 | null | """Reversible augmented-state transducer over aligned learned digit tapes.
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": "1e6fe252-4bed-4fc4-a045-8fa514f7e70e",
"created_at": "2026-08-24 12:52:42.295210+00:00",
"db_md5": "70e27e2d6467b2d09f657fb1a6551554",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "eb429fd1-093c-4b1c-834d-5cb962f7aa80",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",... | {
"score": {
"mean_loss": 6.60203400126609,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01500000001862645
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.63398708111607,
"example_count": 800,
... |
1e72f473-580e-4b24-afa2-db224f3c7484 | easy | alirezashirvani-jr | 2026-08-31 23:28:35.274873+00:00 | succeeded | 2d99e858015257ddae96afe14d5febc7ae86c874a66033164ad9b8eb731ed7ec | 20,604 | null | from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
HP = {
"width": 0,
"maximum_outer_steps": 64,... | {
"id": "1e72f473-580e-4b24-afa2-db224f3c7484",
"created_at": "2026-08-31 23:28:35.274873+00:00",
"db_md5": "ae1d2d5fbb720c3499f6b5524595516e",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "7f0e29fa-f560-4e84-b727-37a05c9472e3",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 3.103555520115358,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00125
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.127441119452763,
"example_count": 600,
"e... |
1e76cf26-b74b-4984-9ea1-15ee0b10c15e | easy | repst | 2026-08-21 22:42:58.518189+00:00 | succeeded | 328901da9db1503b10ab51d1c273be1d34d56a6ff546b5b482d99db755aed1b3 | 30,214 | null | """R355: evaluator-trained full mixture over guarded integer programs.
This organizer-review candidate deliberately authors parsing, integer ALU
semantics, guarded execution, and decimal serialization. It contains no
target program, analytic posterior, MAP path, or label-derived candidate loss.
"""
from dataclasses i... | {
"id": "1e76cf26-b74b-4984-9ea1-15ee0b10c15e",
"created_at": "2026-08-21 22:42:58.518189+00:00",
"db_md5": "fad4762f7d933f57e97740e49b67f9df",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "99faee5b-8a73-4cb1-97b8-7c01fa522c39",
"tier": "easy",
"dataset_id": "e1",
"status": "succeed... | {
"score": {
"mean_loss": 8.11278514127046e-05,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.156296123038984e-05,
"example_count": 100,
... |
1e7b922f-42ee-4957-9fba-3c6c7005905b | easy | chad-atexpedient | 2026-08-05 04:24:14.123109+00:00 | succeeded | 3bcaebaa97a7c9426b8a015f544a235d3dae387ad3a49fc71e90913d42d7401c | 11,222 | null | """R029MH: soft behavior routing with hard retain/write choice."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
LANES = 6
CHANNELS... | {
"id": "1e7b922f-42ee-4957-9fba-3c6c7005905b",
"created_at": "2026-08-05 04:24:14.123109+00:00",
"db_md5": "3159cb859b18c465f753a8a31251f320",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "155e8727-e882-4e79-a2d7-1a225f98240b",
"tier": "easy",
"dataset_id": "e3",
"st... | {
"score": {
"mean_loss": 2.2429195602377714,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.015000000027939676
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.246369010919676,
"example_count": 800... |
1e7d0daf-aae9-41c2-b47a-04dfe675dbd5 | easy | jordanrubin | 2026-08-03 21:09:32.731895+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": "1e7d0daf-aae9-41c2-b47a-04dfe675dbd5",
"created_at": "2026-08-03 21:09:32.731895+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "037ca469-26c6-41b6-9574-0c3655a08206",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.4173998250150417,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.23708332842215896
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.611532835456414,
"example_count": 300,... |
1e869b57-de89-49fe-92e2-bb31cc889548 | easy | benjaminW2025 | 2026-08-18 00:26:59.831896+00:00 | succeeded | f5ebb7f418cf7f5ba52b5087a527e1607bb9561564f065b3a41efb6c1963e384 | 11,190 | null | """Basic single-pass Transformer with PyTorch AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 256
NUM_HEADS = 4
... | {
"id": "1e869b57-de89-49fe-92e2-bb31cc889548",
"created_at": "2026-08-18 00:26:59.831896+00:00",
"db_md5": "6f8c2d1b7d7730bd88a073abdd7262c4",
"submitter": "benjawesome",
"github_login": "benjaminW2025",
"run_id": "1e60c4ef-829f-4225-85e2-0d2362b4cd06",
"tier": "easy",
"dataset_id": "e5",
"status": "... | {
"score": {
"mean_loss": 2.27015723947604,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333345750967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2755140584680533,
"example_count": 600,... |
1e87d7d3-2c25-4935-8828-c0c95eadd85b | easy | ehonig | 2026-08-19 16:32:54.728602+00:00 | succeeded | 6d42e01507be0730537633b3d3a4fb8bfc38930ae4dc8d3062ffa95a185e6235 | 16,556 | 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": "1e87d7d3-2c25-4935-8828-c0c95eadd85b",
"created_at": "2026-08-19 16:32:54.728602+00:00",
"db_md5": "bfe57912fa1a5dff86799f66aee7a84e",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "628d1918-0cdd-4a86-bb57-9a32adb0312f",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 1.914002239704132,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06500000134110451
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9287046194076538,
"example_count": 100,... |
1e94bc83-5913-4171-9d64-b84a2a490a1b | easy | lpbb | 2026-08-19 11:31:20.918621+00:00 | succeeded | 4973f33f7c047c5485eb8e2df804b82f988f51691799511b7d541cb1e69d9664 | 8,863 | null | """S0 + RoPE + abacus embeddings + dropout, with a looped block: instead of
`num_layers` independent layers, a block of `num_layers` unique layers is run
`num_repeats` times with shared weights (e.g. 12 layers x 4 repeats = depth 48
but only 12 layers' worth of parameters). Weight tying across repeats forces
the block ... | {
"id": "1e94bc83-5913-4171-9d64-b84a2a490a1b",
"created_at": "2026-08-19 11:31:20.918621+00:00",
"db_md5": "0a66dc570b8d60726921f64a8306008b",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "417d0584-5e22-40d6-8732-17cb3cdf955e",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 3.054724097251892,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05166666489094496
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.5938363075256348,
"example_count": 100,... |
1e95f1b2-48fe-4734-ac59-b52a32c607b9 | easy | jordanrubin | 2026-08-06 20:00:24.498364+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": "1e95f1b2-48fe-4734-ac59-b52a32c607b9",
"created_at": "2026-08-06 20:00:24.498364+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "1a79002d-c91d-4ee3-9b2d-e8ee4b94a43d",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.4050381808964554,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.21749999669070047
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.413389370970333,
"example_count": 300,... |
1e9b2df1-1fe5-4b1a-a218-c7a25dd99307 | easy | jordanrubin | 2026-08-12 07:06:47.895900+00:00 | succeeded | 60a2a54debdbc8b340d199e9bf0aa3ebed2491bdc562ee97b5c6f1b19eda4130 | 24,919 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "1e9b2df1-1fe5-4b1a-a218-c7a25dd99307",
"created_at": "2026-08-12 07:06:47.895900+00:00",
"db_md5": "c1ea7b63327dc7beb4d7cf3e3c9676a0",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "573a0831-7d7a-4f80-9a08-c221549325cb",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.928895026227316,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.022916666232049465
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.22128128257033,
"example_count": 600,
... |
1eb60615-f5ad-4f8e-9832-69718956a7fa | easy | DDanlov | 2026-08-27 21:36:48.393044+00:00 | failed | 9727ed09ce226e523111926e56eaba3db88c81e6cc456eac5947115e8d73c13e | 38,195 | null | import math
import time
from dataclasses import dataclass
from typing import Optional, Tuple, Dict, Any, List, Union
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from onelayerdeeper.api import Submission, OptimizerBundle, ModelSpec, OptimizerSpec
except ... | {
"id": "1eb60615-f5ad-4f8e-9832-69718956a7fa",
"created_at": "2026-08-27 21:36:48.393044+00:00",
"db_md5": "b04a2f7c3045d941beaab007b2e35920",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "3f939dcc-28a8-42d1-970c-bcd02c5aa892",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
1eb99f5b-0b7b-415c-92f2-f7a17f3c4d1f | easy | anirudh-chakravarthy | 2026-08-12 17:40:24.296850+00:00 | succeeded | 71fc6eed964dfb470c8e027ca9c1e1a40d3b26b5112617b7fcbaf46ebabfe4d5 | 16,746 | null | """The locked model with dense whole-number auxiliary supervision."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert... | {
"id": "1eb99f5b-0b7b-415c-92f2-f7a17f3c4d1f",
"created_at": "2026-08-12 17:40:24.296850+00:00",
"db_md5": "ebc631dca08e6e88faac25fcbf7526ff",
"submitter": "Anirudh S Chakravarthy",
"github_login": "anirudh-chakravarthy",
"run_id": "4396e829-a8dc-4541-99ae-c72f22bdd7c5",
"tier": "easy",
"dataset_id": "... | {
"score": {
"mean_loss": 4.937106300151559,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.203933102133028,
"example_count": 600,... |
1ebba2a3-bd24-48db-969d-e8166cb90075 | easy | DDanlov | 2026-08-12 22:38:52.598060+00:00 | failed | da74ebe078117f7172f81293f5b4a218595df7006a955243fa9374a2bdc913fb | 27,361 | 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": "1ebba2a3-bd24-48db-969d-e8166cb90075",
"created_at": "2026-08-12 22:38:52.598060+00:00",
"db_md5": "65003311306aad6cda7d6f03df2d3421",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "ee960ecf-bfb5-43ff-aed1-933ab3c5014f",
"tier": "easy",
"dataset_id": "e2",
"status": "failed",
... | null
|
1ebe5348-c06d-43b6-ae4c-445d0307a52e | easy | oupadhyay | 2026-08-12 07:40:34.220627+00:00 | succeeded | 08df3377d8f5e6d8caf7d5b1e56fd668bcbfd54f445f92bb3c6812ed2eebf486 | 8,293 | null | """Standalone sgu_opt_warm16_easy_v1 optimization of RMS/axial crossing C21 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
PA... | {
"id": "1ebe5348-c06d-43b6-ae4c-445d0307a52e",
"created_at": "2026-08-12 07:40:34.220627+00:00",
"db_md5": "dc654677d226b96a12b22609163b42ec",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "19219a3e-e331-4070-aec3-0299bd56de03",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.20329349527285,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0083333333581686
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1543262496657434,
"example_count": 600,
... |
1ec512d3-dd6a-43ad-8c99-7ed44103b6ea | easy | oupadhyay | 2026-08-07 14:20:51.109947+00:00 | succeeded | ac7f7725b9186f44c243371519311a695ddf45fd67301e8fb7d27f5966c42c44 | 6,609 | null | """Frozen E1 bidirectional decimal scan, packaged for the full remote budget."""
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 = 64
class Config:
def __init__(self, sp... | {
"id": "1ec512d3-dd6a-43ad-8c99-7ed44103b6ea",
"created_at": "2026-08-07 14:20:51.109947+00:00",
"db_md5": "3eb710f0334fc0f5503607a3704b3e13",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "aa435d4c-1fbd-4c71-bccd-5a204c66aa4f",
"tier": "easy",
"dataset_id": "e1",
"status": "s... | {
"score": {
"mean_loss": 2.647630453109741,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06666667014360428
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.782134532928467,
"example_count": 100,
... |
1ec92ce2-34e0-4372-a8c3-8ce03f017316 | easy | chad-atexpedient | 2026-08-27 02:54:42.771299+00:00 | failed | 36da7ecec06b45707a56da4dd3a425b65192198343f6af163f99391b7579f408 | 18,909 | null | """X132: direct paired-successor supervision.
For supplied same-N chains where A.output equals B.input, decode one shadow
successor from A's terminal state and train it directly on B's supplied output.
This replaces weak latent-only alignment with the exact evaluator-provided
behavioral target. No arithmetic or genera... | {
"id": "1ec92ce2-34e0-4372-a8c3-8ce03f017316",
"created_at": "2026-08-27 02:54:42.771299+00:00",
"db_md5": "0be16a4533b0c510b8e23c9fc1daca00",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "5c1db71c-d124-45be-be39-63493887aeac",
"tier": "easy",
"dataset_id": "e5",
"st... | null
|
1ecb753c-de9a-4176-9e41-a89cb51e7901 | easy | himalalps | 2026-08-07 15:56:10.195752+00:00 | succeeded | 3dc356fe9fbdf67c4ecb5c7464240207f677be3dee94881dc88a5c26e9ffa6e7 | 69,745 | 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": "1ecb753c-de9a-4176-9e41-a89cb51e7901",
"created_at": "2026-08-07 15:56:10.195752+00:00",
"db_md5": "450316a83c94f0b619e73af1a05c0758",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "f67fb6ad-8dd0-4228-80bc-d56df15df826",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 5.680757711196804,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007500000040357312
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.792383670806885,
"example_count": 600,... |
1ece3511-65da-4097-9ca2-1c09b4d877da | easy | DDanlov | 2026-08-13 22:42:07.811349+00:00 | failed | bc4a7bc45b10a3c13b98738aff70c8c4dc8398d26bef4179e6f6bd5aa3533f0c | 22,144 | null | import torch
import torch.nn as nn
import torch.nn.functional as F
import math
from dataclasses import dataclass
from typing import Optional
from torch.optim import Optimizer
try:
from client import Submission, assert_model_state, OptimizerBundle
except ImportError:
Submission = None
assert_model_state = N... | {
"id": "1ece3511-65da-4097-9ca2-1c09b4d877da",
"created_at": "2026-08-13 22:42:07.811349+00:00",
"db_md5": "0b6da4ec56c8a0c3fa0f0516e6b97e39",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "54ab19ea-914e-400f-8e9e-b6d59f825415",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
1ed4f1eb-6304-43ec-b525-d785bf9ee2f2 | easy | KaustubhKumar05 | 2026-08-27 03:30:47.990967+00:00 | succeeded | 69d5c5212a352379b9174f01d5b63f7d5492a872fedb0517e4ff61c30ff75120 | 13,196 | null | """abacus_2block — candidate B3: B with TWO distinct blocks instead of one tied one.\n\nEverything else identical to B, so a difference is attributable to block count.\nSee N_BLOCKS below for why (§47).\n\nOriginal B docstring follows.\n\nabacus_modsquare — candidate B: digit-ALIGNED arithmetic transformer.
FINDINGS §... | {
"id": "1ed4f1eb-6304-43ec-b525-d785bf9ee2f2",
"created_at": "2026-08-27 03:30:47.990967+00:00",
"db_md5": "f53b65f18d98a51f819bda1a43d65036",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "996a346a-0c29-49ec-87e8-5112f505213b",
"tier": "easy",
"dataset_id": "e3",
"status": "succee... | {
"score": {
"mean_loss": 8.116636140568383,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013125000000000001
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.224926535301867,
"example_count": 800,... |
1ee349d3-e411-4daf-8b5f-782fa3a71f9d | easy | nikolageorgiev2000 | 2026-08-04 12:05:27.129711+00:00 | succeeded | 58d1d9176532ad60de5373ce659de4c7d50350d01b0bb0299c30d110817ce6f1 | 10,915 | null | """T-composed categorical-digit reasoner with fully learned pair/reduction maps."""
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,
TokenLossBat... | {
"id": "1ee349d3-e411-4daf-8b5f-782fa3a71f9d",
"created_at": "2026-08-04 12:05:27.129711+00:00",
"db_md5": "a151624833dbed6e15f3f0aa771a5bc6",
"submitter": "Nikola Georgiev",
"github_login": "nikolageorgiev2000",
"run_id": "d3f9f6b5-e5bb-4913-9e72-6ab464c6b7e8",
"tier": "easy",
"dataset_id": "e4",
"s... | {
"score": {
"mean_loss": 7.269361975239983,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006770833345750968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.225989362168324,
"example_count": 1200... |
1ee47975-687d-4d44-8fb1-caaa7e2df631 | easy | lpbb | 2026-08-19 13:00:52.112642+00:00 | succeeded | d3398ae290c00525d0f4c68556c2c47f4029ce4f8e7c95f2dea583bf556f6df8 | 9,874 | null | """S0 + RoPE + abacus embeddings + dropout, with a looped block: instead of
`num_layers` independent layers, a block of `num_layers` unique layers is run
`num_repeats` times with shared weights (e.g. 12 layers x 4 repeats = depth 48
but only 12 layers' worth of parameters). Weight tying across repeats forces
the block ... | {
"id": "1ee47975-687d-4d44-8fb1-caaa7e2df631",
"created_at": "2026-08-19 13:00:52.112642+00:00",
"db_md5": "b8f803e2045f1f935925d201f8e4d9ed",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "83b1de53-ef03-4162-9965-6ed6f1221893",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 2.7986984252929688,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04999999888241291
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.3214313983917236,
"example_count": 100... |
1eedbf16-615a-4f9e-99c5-c09159600bf0 | easy | Yalyenea | 2026-08-12 17:44:42.135936+00:00 | failed | 8ca95040258a23c92a8400aedea4d47bede508fa86798136d762072200308e95 | 11,411 | null | """Small generic recurrent Transformer for hidden function composition."""
from __future__ import annotations
from dataclasses import dataclass
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerS... | {
"id": "1eedbf16-615a-4f9e-99c5-c09159600bf0",
"created_at": "2026-08-12 17:44:42.135936+00:00",
"db_md5": "590fb80490b3f7895ea4deb950bc035b",
"submitter": "yfff",
"github_login": "Yalyenea",
"run_id": "158bc5c3-13df-407d-bfed-b70b6db97ba0",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
"... | null
|
1efad6c7-bfda-4748-9f74-1f9051eee9e5 | easy | jordanrubin | 2026-08-05 18:56:37.988501+00:00 | succeeded | fe9b40d55e805d01fd2a26b26a22f3c208612e4f202f842974410049f26f64e8 | 11,503 | null | """Continuous-state 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 times.
T controls o... | {
"id": "1efad6c7-bfda-4748-9f74-1f9051eee9e5",
"created_at": "2026-08-05 18:56:37.988501+00:00",
"db_md5": "69d1907ec985af05eaefe042069e3800",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "5d151bc9-199b-467a-b62f-9f17641712a1",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.652084774489203,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666599921882
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.413952409874964,
"example_count": 300,... |
1efe313b-e762-4e81-afb2-a5e41f75483a | easy | Ranoobaba | 2026-08-19 13:19:03.030135+00:00 | succeeded | b71314738962d767d7c6eeb1c9d67deabd95a17ff4fcfd2b69ef853953d51498 | 6,601 | null | """Looped transformer, iteration count conditioned on T read from the prompt.
Core idea: learn one latent "squaring step" operator and apply it T times.
Depth generalization comes from the loop structure itself; the model never
needs more parameters for deeper problems, only more iterations.
"""
from __future__ impor... | {
"id": "1efe313b-e762-4e81-afb2-a5e41f75483a",
"created_at": "2026-08-19 13:19:03.030135+00:00",
"db_md5": "462bd1aee09ff4498696aeaa01594429",
"submitter": "Syed Rayyan Ali",
"github_login": "Ranoobaba",
"run_id": "aa7345e4-87d9-4fb7-ada7-d2b6e7430f9b",
"tier": "easy",
"dataset_id": "e5",
"status": "... | {
"score": {
"mean_loss": 2.5389763519165793,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6019787819128934,
"example_count": 60... |
1f0c3618-badc-4f98-8e23-861059c6c83d | easy | DDanlov | 2026-08-16 00:11:44.222114+00:00 | succeeded | 94b8408cfca8d38d2b504daf4b357649b5e5d5f4eafde8d612c8897288a44a36 | 16,554 | null | """
Production Port-Hamiltonian RoPE Dynamic Submission
Config ID: 4 | Name: HighWidth_S4_d1024_Classic
dim=1024, stages=4, heads=16, dt=0.12, lr=0.006, wd=0.05, opt=shampoo
"""
from __future__ import annotations
import math
from typing import Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.function... | {
"id": "1f0c3618-badc-4f98-8e23-861059c6c83d",
"created_at": "2026-08-16 00:11:44.222114+00:00",
"db_md5": "e796d8504420eb3375e552b82b068b96",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "46367956-39d6-4afa-a04f-b635b4a5cd1f",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.244940283686941,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666691501935
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.314469194198403,
"example_count": 600,... |
1f1205f0-d082-4e92-9439-db93052506af | easy | khushidahi | 2026-08-21 03:52:27.360499+00:00 | succeeded | 965c76f9ef75eb75e6c508fa239905eb3f687fe4a4c378e00d63b311d13ae6f0 | 16,754 | null | """R335: fixed-width residue objective for the Table-5 Abacus loop.
The research claim is deliberately narrow: on fixed-modulus Easy E6, can the
published Abacus representation plus a maximally recurrent Transformer learn
the one-step modular-squaring transition on held-out operands?
McLeish et al.'s arithmetic model... | {
"id": "1f1205f0-d082-4e92-9439-db93052506af",
"created_at": "2026-08-21 03:52:27.360499+00:00",
"db_md5": "b01a00c444fa4419f48095a07a0bcde7",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "1da3038c-15ee-4545-97fe-872a33d551e3",
"tier": "easy",
"dataset_id": "e6",
"status": "succ... | {
"score": {
"mean_loss": 5.7072690153775145,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.051801803043565235
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.461006164550781,
"example_count": 60,... |
1f1329c7-ac7b-43da-a67d-ae1ae83c4e3a | easy | jordanrubin | 2026-08-12 04:46:42.791672+00:00 | succeeded | 5e67d1a282bda9fab14fa7d6e3200be9c6aac5fca5eb9b7f7a229db9bcf84720 | 22,494 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "1f1329c7-ac7b-43da-a67d-ae1ae83c4e3a",
"created_at": "2026-08-12 04:46:42.791672+00:00",
"db_md5": "38f0396c0fc03a18e69ddda1edcd8120",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "c59c8366-8957-4950-a601-191efb24534e",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.554158915752302,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00666666644004484
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.173258880462347,
"example_count": 600,
... |
1f17cfc8-151b-446c-88d3-fe7e40559af6 | easy | erdavis0 | 2026-08-24 12:39:32.405929+00:00 | succeeded | eaea7b22df53c8a50c29a15d16a2cf3a12bf39b0866119d4f9a98b513887d2fe | 30,668 | null | """Reversible augmented-state transducer over aligned learned digit tapes.
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": "1f17cfc8-151b-446c-88d3-fe7e40559af6",
"created_at": "2026-08-24 12:39:32.405929+00:00",
"db_md5": "5ebf532547b56dac63fc61f7e3a1c9a9",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "7341c7b9-f876-43df-9cf5-e66a731abd0a",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.3287116271023285,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3455394950148234,
"example_count": 60... |
1f22938c-2ef2-40b4-ba9f-40ad10890077 | easy | chad-atexpedient | 2026-08-11 04:07:04.381145+00:00 | succeeded | b2b435e0a79b915f188f2cc4964da03221012e6393aae402587e571787f493e9 | 11,137 | null | """X13A: R028F forced semantic router at width 96."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
LANES = 6
CHA... | {
"id": "1f22938c-2ef2-40b4-ba9f-40ad10890077",
"created_at": "2026-08-11 04:07:04.381145+00:00",
"db_md5": "f69bd3b651f119c8a596cd741598d5a1",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "e804fbe3-53dd-4978-a182-becf5915adc9",
"tier": "easy",
"dataset_id": "e3",
"st... | {
"score": {
"mean_loss": 2.328092065415833,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0031250000093132257
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.329637477013761,
"example_count": 800... |
1f2affd1-bfc3-43e3-9931-30bc184755aa | easy | yashkant | 2026-08-15 14:29:17.512063+00:00 | succeeded | 9b1e5133ff06ac386bbd97fdabf43cabfad8aca861640827949a6dbe9e49d072 | 31,765 | null | """Round1702 three-projections-per-T Dykstra recursion on the D416 carrier.
Each parsed public-T cycle applies an A-B-A sequence through one shared learned
smooth projection, reinjecting the immutable anchor on every call and updating
the corresponding Dykstra correction memory after each projection. All active
cycle... | {
"id": "1f2affd1-bfc3-43e3-9931-30bc184755aa",
"created_at": "2026-08-15 14:29:17.512063+00:00",
"db_md5": "520fd45ba52d67a06d79ff8be9c27218",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "9c633e57-9fb4-49d2-85e3-970f62c07486",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.1270420431124353,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018750000204890968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.126501095711207,
"example_count": 800... |
1f2c9980-c23e-4704-bd32-58e8f5732fb6 | easy | himalalps | 2026-08-06 16:04:26.611071+00:00 | succeeded | b7a5e3bcd88ad16c5d6fd3b68640b02130129febbcd9e9cdd323ce429e5343d9 | 64,126 | 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": "1f2c9980-c23e-4704-bd32-58e8f5732fb6",
"created_at": "2026-08-06 16:04:26.611071+00:00",
"db_md5": "58eb14749dfd628aab943f7e40aea185",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "2494175d-b5ba-4937-b4c5-a8c2d3af5879",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 2.270895631698606,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008750000229726235
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.256821632385254,
"example_count": 600,... |
1f2cfd62-6949-41a9-8ebb-ebae285c1ebd | easy | Ranoobaba | 2026-08-16 09:08:51.958973+00:00 | succeeded | 0c67fec1cc3134b4a53d530d8e3885ca1ff7d9700d63253dbded73844f0062a7 | 5,723 | null | """Looped transformer, iteration count conditioned on T read from the prompt.
Core idea: learn one latent "squaring step" operator and apply it T times.
Depth generalization comes from the loop structure itself; the model never
needs more parameters for deeper problems, only more iterations.
"""
from __future__ impor... | {
"id": "1f2cfd62-6949-41a9-8ebb-ebae285c1ebd",
"created_at": "2026-08-16 09:08:51.958973+00:00",
"db_md5": "aa1920b1d8b46fe158f6bf85741b3f4a",
"submitter": "Syed Rayyan Ali",
"github_login": "Ranoobaba",
"run_id": "8cc0760d-cf77-487f-98ee-8ee234813808",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 6.004384244172056,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02833333299805721
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.31915283203125,
"example_count": 100,
... |
1f34879a-7db3-4d78-83eb-1d391dd33227 | easy | yashkant | 2026-08-31 19:27:16.224255+00:00 | failed | a6539a7e76206318ee2873249fe9e043ae66aa58637275f3187bdfbb2a5d456f | 107,958 | null | """One Layer Deeper submission -- a learned register machine.
Architecture (all outputs produced by learned parameters; no task arithmetic on the answering path):
a bank of R=6 digit registers (W=16 positions each) carried as DISCRETE tokens between micro-steps: five
data registers and one COUNTDOWN register holding t... | {
"id": "1f34879a-7db3-4d78-83eb-1d391dd33227",
"created_at": "2026-08-31 19:27:16.224255+00:00",
"db_md5": "eb03058609823bd88ccd78afd714ff0c",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "46e66921-d73d-44ee-8f8a-f69e362bb4e4",
"tier": "easy",
"dataset_id": "e1",
"status": "failed"... | null
|
1f3d0f45-f8a1-4ceb-9028-1aea4f817d57 | easy | chad-atexpedient | 2026-08-09 13:30:55.480613+00:00 | succeeded | c736e1cdd359fc029015e19a6ff6e6acc65fa540d44e4b74ce29eac475810864 | 10,397 | null | """X12: persistent phase-typed cellular microprogram."""
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
WIDTH = 64
WORK_LANES = 4
PHASES = 4
... | {
"id": "1f3d0f45-f8a1-4ceb-9028-1aea4f817d57",
"created_at": "2026-08-09 13:30:55.480613+00:00",
"db_md5": "5a02a4f06708f347b702fddc28e6aabe",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "d0712beb-f091-4633-81f3-47647bec9381",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 9.20097601762462,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00625
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.64676886930594,
"example_count": 600,
"exa... |
1f405afd-4170-4f85-b6b0-21e92b2afdde | easy | yashkant | 2026-08-10 20:22:56.500791+00:00 | succeeded | 8e3ad45dde3bc081bffc8263034b1e05b611d49e69651cefad2f7bc53a779744 | 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": "1f405afd-4170-4f85-b6b0-21e92b2afdde",
"created_at": "2026-08-10 20:22:56.500791+00:00",
"db_md5": "60c097b303e1d8abc2a017917010a304",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "3ca0aaef-c89c-4c7d-a297-c9547d684f62",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.1282284625165575,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013750000158324838
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1262681827256293,
"example_count": 80... |
1f40b6a5-7521-4235-a5ce-2ffd5bda182b | easy | viridale | 2026-08-13 11:20:33.397107+00:00 | succeeded | b94ddeef5de06ed7869fa8dad2e4fdaea0ea993b591b20f6a7be818a8651bb25 | 23,114 | null | """Bit-canonical recurrent neural multiplier/reducer.
hypothesis: exact bit atoms plus a native learned bilinear product-column route and
learned N-conditioned associative carry/reduction scan make a small recurrence
edit constructively learnable instead of asking an unconstrained chart to invent
arithmeti... | {
"id": "1f40b6a5-7521-4235-a5ce-2ffd5bda182b",
"created_at": "2026-08-13 11:20:33.397107+00:00",
"db_md5": "e9abb543b51185dfc87e28d0ab959a3f",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "48f1ea78-e53e-4011-bb09-829b7721b702",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 17.278656005859375,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 17.56584930419922,
"example_count": 100,
"exac... |
1f4339a2-6527-4d4d-9dcf-ba629b79feec | easy | anirudh-chakravarthy | 2026-08-12 00:35:33.843363+00:00 | succeeded | 5911ab3527210f2ac038faedefa13d52dcdfca52e7f4648fc18fb43c2697e34a | 14,118 | 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": "1f4339a2-6527-4d4d-9dcf-ba629b79feec",
"created_at": "2026-08-12 00:35:33.843363+00:00",
"db_md5": "1bb78b8b2d522dcc3b2f7fe561241e2f",
"submitter": "Anirudh S Chakravarthy",
"github_login": "anirudh-chakravarthy",
"run_id": "455000cc-faf3-4bfc-9fd0-6c7ac6d0ec23",
"tier": "easy",
"dataset_id": "... | {
"score": {
"mean_loss": 2.3611556477391717,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008750000024835268
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.493849252906081,
"example_count": 600... |
1f494e4d-beb5-4441-bccf-db44e69a4a2b | easy | ehonig | 2026-08-15 01:04:46.248606+00:00 | failed | 782475e484c85dcc6616309477ed0d16be3d5bb6b9c91532b72c7f4efef513ba | 89,099 | 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": "1f494e4d-beb5-4441-bccf-db44e69a4a2b",
"created_at": "2026-08-15 01:04:46.248606+00:00",
"db_md5": "4be7d37e97097e9db9bfc9b998d7c0a1",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "de4c297f-6014-43d3-83d9-4323040a90b2",
"tier": "easy",
"dataset_id": "e1",
"status": "fail... | null
|
1f4ced58-430d-47be-8940-2404639888c5 | easy | khushidahi | 2026-08-07 03:21:16.714477+00:00 | succeeded | c19a42a866f0e3271768206dd35e3e4c074769021001ebcb559f617c7719753e | 16,917 | 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": "1f4ced58-430d-47be-8940-2404639888c5",
"created_at": "2026-08-07 03:21:16.714477+00:00",
"db_md5": "2f55a24e1a13f68932c6d545bc1c7c4e",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "44738fc3-a6b9-437d-8af4-04660859b516",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.1807671785354614,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666567325592
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1789722442626953,
"example_count": 60... |
1f517477-a4ed-442c-aa75-de2432696c8f | easy | erdavis0 | 2026-08-21 14:25:20.684881+00:00 | succeeded | e37358a21622100ab57ead40071bc75817c1cd70ff5e1eca862fdf0ea3de1faf | 4,764 | null | """Single fused forward GRU augmented by an invariant global prompt summary."""
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... | {
"id": "1f517477-a4ed-442c-aa75-de2432696c8f",
"created_at": "2026-08-21 14:25:20.684881+00:00",
"db_md5": "2ddc044b08096c26eb27e0bacf9f9213",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "30fcde1f-c740-4926-8cdb-393b7a0c4474",
"tier": "easy",
"dataset_id": "e7",
"status": "succeeded",... | {
"score": {
"mean_loss": 9.589510440826416,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.028609626460820436
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.160004615783691,
"example_count": 85,... |
1f57f6b1-85d1-488e-9eb6-f69f83fbbdc8 | easy | khushidahi | 2026-08-23 19:49:24.751211+00:00 | failed | dc30f96367eeccb5fe562f1ff3bdff32abfe024a824930c0ad6f787d71aa41f0 | 11,734 | null | """ART-DT recall candidate.
A learned recurrent sequence model with right-to-left Abacus positions and
global-sequence rotary attention. This variant changes only optimizer LR and
warmup schedule from its architectural parent. This branch ties the two
within-visit layer calls to one TransformerLayer object. One imm... | {
"id": "1f57f6b1-85d1-488e-9eb6-f69f83fbbdc8",
"created_at": "2026-08-23 19:49:24.751211+00:00",
"db_md5": "25f8e724d7ec7f02d7c56a4ef3401964",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "783e1ac4-99fa-45e9-8bae-b270defe76c5",
"tier": "easy",
"dataset_id": "e9",
"status": "fail... | null
|
1f5ee962-12c4-46ee-9291-74106e366154 | easy | viridale | 2026-08-24 23:40:33.075903+00:00 | succeeded | 89e09d0fee51e7075b92a27003013efd728481329a75bcf95572a2b5a9d8bbc2 | 17,925 | null | """Barycentric-witness quotient-ideal Transformer.
hypothesis: v870's ephemeral quotient proposal needlessly crosses a hard categorical
boundary; keeping that proof witness on the digit simplex gives endpoint gradients
a short continuous path while retaining the binding shared-coordinate ideal
axis: arch
targe... | {
"id": "1f5ee962-12c4-46ee-9291-74106e366154",
"created_at": "2026-08-24 23:40:33.075903+00:00",
"db_md5": "4d9184460e5d2839c005834bc71a3f78",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "7244d09a-c271-457e-8f56-7e26740e8462",
"tier": "easy",
"dataset_id": "e6",
"status": "succe... | {
"score": {
"mean_loss": 2.189384341239929,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07522523030638695
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9571483135223389,
"example_count": 60,
... |
1f652d8d-cc54-47df-96d7-4e150c419c35 | easy | shirvani-jr | 2026-08-30 23:37:36.975555+00:00 | succeeded | 5e04560fff6d76c6350ed134dce34bbeb5ddf03992d210d9cdc6dda6ff09212c | 17,544 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PERIOD_MINIMUM = 11
PERIOD_MAXIMUM = 100
E... | {
"id": "1f652d8d-cc54-47df-96d7-4e150c419c35",
"created_at": "2026-08-30 23:37:36.975555+00:00",
"db_md5": "cfec548a5587d3a2942b9c10220452bc",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "67c76e14-62e2-4368-925e-f02ebba7f917",
"tier": "easy",
"dataset_id": "e8",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.1292214393615723,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.030704100616276264
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.199883460998535,
"example_count": 85,... |
1f6646b3-abcf-4e61-8394-de71afc8594c | easy | shan23chen | 2026-08-08 02:46:41.895175+00:00 | succeeded | 78ca8e5d6066f76f50aec1313b2c8f2efbec8fe58ed743244822b66dc3237607 | 24,961 | null | """Boundary-reset recurrent register Transformer.
One learned transition is applied once per requested outer step. The
transition receives an immutable decimal N register and the current learned
boundary state, but never receives T or a separate copy of the original X.
Within each outer application, scratch persists ... | {
"id": "1f6646b3-abcf-4e61-8394-de71afc8594c",
"created_at": "2026-08-08 02:46:41.895175+00:00",
"db_md5": "57c6213eda41c31c213c0ac18948a5dd",
"submitter": "Shan Chen",
"github_login": "shan23chen",
"run_id": "f2ab1e14-5d5d-4bb7-a91d-96e0a105fac1",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 2.8001887798309326,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009166667005047202
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.7178754806518555,
"example_count": 60... |
1f6cb305-c775-4a5d-8a07-96729105378c | easy | khushidahi | 2026-08-06 05:25:07.339021+00:00 | succeeded | 2ab0219829bcbb3502a8296bc650fa7df34bfda71fa8f92f50f0d72379fa8c26 | 14,803 | 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": "1f6cb305-c775-4a5d-8a07-96729105378c",
"created_at": "2026-08-06 05:25:07.339021+00:00",
"db_md5": "fcf5ecd559c5ae67823a062bdf7a1ad9",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "ba2e9875-9423-4864-a258-f8fa6a373725",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 2.2310925722122192,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01229166705161333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2324938774108887,
"example_count": 300... |
1f6cf356-42dd-45c5-9346-a1966ef2bb35 | easy | alirezashirvani-jr | 2026-08-29 05:16:55.085586+00:00 | succeeded | 3d3211f7ed6eb6a8737f22c37599eff66c9da5c52e3bd9b47c7ffc449e7d66a8 | 44,251 | null | """Adaptive learned discrete recurrent fabric for One Layer Deeper.
A gradient-trained, operation-free recurrent computational machine. The
public prompt grammar only routes literal digit tokens into canonical,
tail-aligned categorical registers. A reusable learned transition circuit
is applied serially, a learned c... | {
"id": "1f6cf356-42dd-45c5-9346-a1966ef2bb35",
"created_at": "2026-08-29 05:16:55.085586+00:00",
"db_md5": "c594db05f26962ba8128579425869daf",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "9c51aae0-74ee-4d92-9ad0-39beae051883",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 2.7549807847821497,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013333333414047957
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.748147487640381,
"example_count": 600... |
1f775a4a-2872-476a-8107-fe2ef0cdde17 | easy | KaustubhKumar05 | 2026-08-28 06:23:44.683676+00:00 | succeeded | 5a815f515e555cb3c5f1aa15561ce421e7d119abfe07b394df334b7f72635e19 | 12,882 | null | """cell_budget -- allocate compute by DIFFICULTY, not by T.
THE OBSERVATION (S86, pure arithmetic, one of the two findings that survived today).
The image of y -> y^2 mod N collapses to a fixed set S_N by T <= 6 on all 29 moduli
tested, and squaring is a genuine bijection on S_N with |S_N| between 19 and 1351. So
rung... | {
"id": "1f775a4a-2872-476a-8107-fe2ef0cdde17",
"created_at": "2026-08-28 06:23:44.683676+00:00",
"db_md5": "13a8dc016b8f6e082280d0cd5647c6af",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "800ffd2a-619c-4669-aba8-fef3888205f4",
"tier": "easy",
"dataset_id": "e4",
"status": "succee... | {
"score": {
"mean_loss": 2.4751167472149147,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003958333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8711018695365627,
"example_count": 12... |
1f7f77d1-01aa-441a-99aa-d61a2ad83793 | easy | liam-gb | 2026-08-13 13:03:11.611354+00:00 | succeeded | 97ea56c561d2643022dfd035d72ac60f7995748f53262bd174cc07aca02061a9 | 17,316 | null | """rns_pm_gather_ct_widet_fast: norelrt with the dispatch waste removed.
Every change here is mathematically identical to norelrt. Nothing about the
model changes; only how many kernels it takes to say it.
Verified on real data: the phase read-in is **bit-identical** (theta and the
integer time-step match exactly). T... | {
"id": "1f7f77d1-01aa-441a-99aa-d61a2ad83793",
"created_at": "2026-08-13 13:03:11.611354+00:00",
"db_md5": "e42bafbc7d4380ef79add2b98bf1fb0c",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "23dd01a1-b71e-49ef-b510-967ec421a0a5",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 7.508300322602998,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.057500000006208814
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.233137130737305,
"example_count": 600,... |
1f849efd-e67b-4172-a9a9-e1863c2541db | easy | gauravmishra | 2026-08-10 03:51:07.326442+00:00 | succeeded | 534d7373bfa3a4a6d4564326eea7ec437841a494e0f5bbdfa5adc5026a4f9f6f | 12,545 | null | """Stage-1 information-gain K4 representation candidate."""
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,
)
POSITI... | {
"id": "1f849efd-e67b-4172-a9a9-e1863c2541db",
"created_at": "2026-08-10 03:51:07.326442+00:00",
"db_md5": "2721d0863048f4daafd2d2de127e3c1b",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "0f20b75a-3a9d-4440-ad92-bc97f712c9c5",
"tier": "easy",
"dataset_id": "e4",
"status": ... | {
"score": {
"mean_loss": 2.1985497249065,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006041666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2535600156171083,
"example_count": 1200,... |
1f88a7e5-ff6e-4e60-bd9d-7f4f740ce2be | easy | briheet | 2026-08-05 11:17:02.354652+00:00 | succeeded | 18d3879e64a76b06ae56f915a918f59a774549b1df97ccf6467ff64f94ce40b2 | 4,177 | 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": "1f88a7e5-ff6e-4e60-bd9d-7f4f740ce2be",
"created_at": "2026-08-05 11:17:02.354652+00:00",
"db_md5": "22bc28b7e1075477281c58b8421795ed",
"submitter": "Aizenn",
"github_login": "briheet",
"run_id": "ea5a3a98-c7a0-47e5-8be8-f8aa52e31177",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",... | {
"score": {
"mean_loss": 3.5998438596725464,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03333333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.9840917587280273,
"example_count": 100... |
1f929b58-db59-4c2e-9ecb-4a0537861d93 | easy | DESU-CLUB | 2026-08-24 10:04:22.962050+00:00 | succeeded | 17d1214f057cdc95b3719ddb6e48ed7536f4a35de1834c5314d0d0b23984e50e | 7,896 | 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": "1f929b58-db59-4c2e-9ecb-4a0537861d93",
"created_at": "2026-08-24 10:04:22.962050+00:00",
"db_md5": "deeff860ffa760174d8010a86ab315a3",
"submitter": "Low Keng Hoong, Warren",
"github_login": "DESU-CLUB",
"run_id": "0c54852b-851f-4547-a2ec-d62098c9309c",
"tier": "easy",
"dataset_id": "e6",
"sta... | {
"score": {
"mean_loss": 8.352590505390951,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09031531889293645
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.252146244049072,
"example_count": 60,
... |
1f97f453-7866-4ec0-9766-080856781a10 | easy | shreyash-chonkie | 2026-08-23 22:54:13.738182+00:00 | succeeded | 40de22d31f27866c48694b609d4f5e3e1c827019c114db117e2c9cd24053e964 | 8,233 | null | """Four-stage recurrent register Transformer for One Layer Deeper."""
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 = 3200
... | {
"id": "1f97f453-7866-4ec0-9766-080856781a10",
"created_at": "2026-08-23 22:54:13.738182+00:00",
"db_md5": "1b2b4783ef4373c9ad87b2452eed0d95",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "7587227b-18d0-43a6-ae78-100f35f29dcb",
"tier": "easy",
"dataset_id": "e6",
"status": "... | {
"score": {
"mean_loss": 2.1926727294921875,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02342342445626855
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1651763916015625,
"example_count": 60,... |
1f9837d2-4476-4027-a2d2-c0c4286d9422 | easy | johnowhitaker | 2026-08-03 21:32:50.622159+00:00 | succeeded | f4c39cfa3ad32db38b643378c612555e47260838654530bab22e5aca18708c3f | 11,742 | null | """Interleaved residue-state recurrent Transformer.
Four possible residue positions at the right edge of each prompt receive five
internal spacing slots apiece. A shared Transformer body is rerun six times.
After each pass, the state at the current prediction position is inserted into
the next spacing slot, creating ... | {
"id": "1f9837d2-4476-4027-a2d2-c0c4286d9422",
"created_at": "2026-08-03 21:32:50.622159+00:00",
"db_md5": "3b77a312cff7f863f665eb8019647756",
"submitter": "Jonathan Whitaker",
"github_login": "johnowhitaker",
"run_id": "64a9c8db-3104-417a-aa73-656a92167a0a",
"tier": "easy",
"dataset_id": "e1",
"stat... | {
"score": {
"mean_loss": 1.9070656299591064,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07999999821186066
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8852323293685913,
"example_count": 100... |
1f9ca580-eac7-4d27-b436-7bd07a612dc8 | easy | yashkant | 2026-08-19 01:31:33.117459+00:00 | succeeded | a0735e6814da5a0e030f01898248a107eb53b46eb33dfaf5d8794c12fbff5f5a | 43,555 | null | """Round2190 removes cross-step Dykstra correction carry.
The retained official carrier, primal recurrence, two learned projections,
antisymmetric set codes, parsed-T allocation, optimizer, and EMA are fixed.
The candidate reuses the learned initial A/B correction offsets on every
cycle instead of carrying hand-writte... | {
"id": "1f9ca580-eac7-4d27-b436-7bd07a612dc8",
"created_at": "2026-08-19 01:31:33.117459+00:00",
"db_md5": "a1e58993d313ec62f708dc7f75822fad",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "f5478246-72a2-42b8-9f64-916f00609401",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 3.736788034439087,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08083333447575569
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.742213249206543,
"example_count": 300,
... |
1f9f323d-1e4f-4e7d-bef9-c8bf5126ab79 | easy | DDanlov | 2026-08-31 21:25:07.488726+00:00 | succeeded | 0981212e7c32bbe813e123adfd4bacad4c23631e5eeadd8e39ce73e902ed95b5 | 32,408 | null | """
Unified Continuous Hopfield Equilibrium Propagation Deliberation Engine
=======================================================================
Architecture:
- Deliberation: Continuous Hopfield Multi-Layer Energy System (L=4, D=768, d_ffn=2304, H=16)
- Integration: In-Place Strang-Split Conformal Symplectic Velocit... | {
"id": "1f9f323d-1e4f-4e7d-bef9-c8bf5126ab79",
"created_at": "2026-08-31 21:25:07.488726+00:00",
"db_md5": "474bd6441c27824955cd2c8a0feae135",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "0eb29d2f-582f-4a70-833b-9c31b20b74a4",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.1123501918464744,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0025
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.1112988643100596,
"example_count": 800,
"... |
1f9fd8ed-6d88-42d5-b5bc-be9810984c52 | easy | khushidahi | 2026-08-06 03:20:09.580961+00:00 | succeeded | 98fa61132b1a1344a5d4261c059236ed182206f71630ddc6141eea0790063ddd | 14,370 | 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": "1f9fd8ed-6d88-42d5-b5bc-be9810984c52",
"created_at": "2026-08-06 03:20:09.580961+00:00",
"db_md5": "71c7be138366fd4f5a07237589d068fb",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "a8420cfa-36c0-45e6-a4a3-0f43e3ffc57d",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 2.2363829612731934,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011041667079553008
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.234870433807373,
"example_count": 300... |
1fa4e2b3-c461-4abb-af07-e1a3d9504900 | easy | karanganesan | 2026-08-05 10:46:16.230480+00:00 | succeeded | b942e61fcead25a4278bbf0197040699a289801b5341b3fdb7b460c71899f414 | 22,081 | 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": "1fa4e2b3-c461-4abb-af07-e1a3d9504900",
"created_at": "2026-08-05 10:46:16.230480+00:00",
"db_md5": "ef47b23429836827cec88a4bb5cc1969",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "f14ab51a-83e2-413e-ac7e-fad547ec7385",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 18.31827163696289,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02166666742414236
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 20.069591522216797,
"example_count": 100,... |
1fb124bf-aa16-45c6-a38b-e6370256a2bb | easy | yashkant | 2026-08-19 02:09:43.405650+00:00 | succeeded | 5bf8f121b988273162cf8750198b4fdcabe0fb154eb587dc36e58b26c3c9a185 | 39,755 | null | """Round2222 removes the reset model's dead correction-state updates.
The replicated reset winner never reuses either correction proposal: each
parsed-T cycle reinjects the same two learned offsets. This candidate is
therefore a primal-only recurrence with two shared bounded projections per
cycle. It preserves every... | {
"id": "1fb124bf-aa16-45c6-a38b-e6370256a2bb",
"created_at": "2026-08-19 02:09:43.405650+00:00",
"db_md5": "f8233dc122db1ac70fbbd92026effe96",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "4ed0c56d-5c41-437f-bbf5-22ecc0c705a6",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 2.145391420251706,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010000000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.149309851128959,
"example_count": 600,... |
1fb63c3a-04b3-40f6-ab6a-67d36b421219 | easy | isaac0804 | 2026-08-05 09:34:13.793219+00:00 | succeeded | 25b7bff2ebb05cb23fe8aae9e035a6b76177e23260184cf1f7340a7b67e20392 | 6,971 | null | """Vanilla (non-recurrent, non-weight-tied) deep Transformer, widened, to test
a specific hypothesis about the E3 collapse diagnosed in
submissions/baseline_adamw_deep: local CPU instrumentation showed the deep
(8-layer, D_MODEL=128) model's train loss on E3 (sampled N, fixed T=2) drops
fast for ~20-40 steps then g... | {
"id": "1fb63c3a-04b3-40f6-ab6a-67d36b421219",
"created_at": "2026-08-05 09:34:13.793219+00:00",
"db_md5": "92870940c6df7b3cf80b357308091a8c",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "93b7b50e-2085-4ca7-8ab1-e92ba0224dbf",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 9.05642453737281,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01562500004656613
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.933033489809851,
"example_count": 800,
... |
1fcfbf2b-fe5f-4dd1-abf6-e1c9fd5857c4 | easy | DDanlov | 2026-08-19 19:40:19.997325+00:00 | failed | 35a339eabbb1791c33c3add55e6294f26e72c832e915a33952d80ca441b080c5 | 17,088 | null | """
Dynamic Submission for One Layer Deeper API Benchmark
Config: dim=1024, num_heads=32, d_ff_mult=4, fixed_trec=224
"""
from __future__ import annotations
import math
from types import SimpleNamespace
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from to... | {
"id": "1fcfbf2b-fe5f-4dd1-abf6-e1c9fd5857c4",
"created_at": "2026-08-19 19:40:19.997325+00:00",
"db_md5": "f3f03b5154a39e5c0b405e87f8efcb76",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "b60fd356-7c26-4a42-8a8c-bbbe423b8497",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
1fdcb83d-afe9-41ff-ac9f-89e5242915dd | easy | khushidahi | 2026-08-07 02:16:29.903305+00:00 | succeeded | 0df7cfc854246bafc5a30de738991427b8405f02047a5ce3487c2e3bc9e0676a | 18,888 | 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": "1fdcb83d-afe9-41ff-ac9f-89e5242915dd",
"created_at": "2026-08-07 02:16:29.903305+00:00",
"db_md5": "4929b699f0a40cd94ae5888108374a24",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "7ae6d60e-6cdb-4a5f-82fd-903b77c70d91",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.1978633403778076,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666977107525
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1967785358428955,
"example_count": 60... |
1fe2852e-4e1a-45f8-9a1b-743499bb5cf7 | easy | kushasareen | 2026-08-27 03:52:57.988256+00:00 | succeeded | 1ac4d06cabd87a87f83daa4c0f3592668bef6e9ca57b7745f7ffd07a0e5d9235 | 10,895 | null | """loop2_T: digitcomp (4 blocks) where blocks 2..3 repeat T times; T token kept.
At T = 1 this is exactly the digitcomp model. Muon recipe, 3 updates per batch.
Each number's token is a learned function of its digits: every place has
its own embedding table, the place embeddings are summed and passed
through a small ... | {
"id": "1fe2852e-4e1a-45f8-9a1b-743499bb5cf7",
"created_at": "2026-08-27 03:52:57.988256+00:00",
"db_md5": "825703bce2a3f5a907566ff25735ee9b",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "4d16eb70-029b-4a3a-8bf2-b0599e10056c",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.408559024333954,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.4145833496004343
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.363450050354004,
"example_count": 300,
... |
1fe5ac71-acce-4d86-825d-f1a766ab9dbd | easy | pjjajal | 2026-08-07 03:25:21.676298+00:00 | succeeded | f635a625901210eb631f20b7e3760bf9b4b2922a7b4511014b08635e35e6e5be | 13,430 | null | """Looped Transformer with an outer-product residual state.
Each token carries several residual streams. A dynamic read contracts those
streams to one Transformer-width vector; a dynamic write forms an outer product
between per-stream write weights and the Transformer output, then adds that
matrix to the identity-car... | {
"id": "1fe5ac71-acce-4d86-825d-f1a766ab9dbd",
"created_at": "2026-08-07 03:25:21.676298+00:00",
"db_md5": "29e810e31b33bb4feae923476c0b2d20",
"submitter": "pjjajal",
"github_login": "pjjajal",
"run_id": "cd4eaa33-7535-4c45-8e61-259a6e5b1b26",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 1.9470096230506897,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0716666653752327
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9351799488067627,
"example_count": 100,... |
1fee0a04-17ea-489b-9cc9-42ce89a422ff | easy | richardcepka | 2026-08-13 23:09:20.980993+00:00 | succeeded | 60a90490b0dcc0cc1404f041a03e5b444e6ebfd5e2c7098f7dd3f285fd9f3f46 | 16,676 | null | """Depth-recurrent digit model for One Layer Deeper.
The prompt is encoded once, then injected into a shared transformer core on
every recurrent step. A small decimal register is the persistent answer state;
continuous scratch tokens carry everything that is not yet ready to become a
digit. The register stays soft d... | {
"id": "1fee0a04-17ea-489b-9cc9-42ce89a422ff",
"created_at": "2026-08-13 23:09:20.980993+00:00",
"db_md5": "60726ea1b89004c5736e533eeb624399",
"submitter": "Richard Cepka",
"github_login": "richardcepka",
"run_id": "aee711bf-75cd-401a-a03d-6a732cf1375a",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.62905552268431,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666908810537
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.735147714614868,
"example_count": 600,
... |
1ff253aa-adb5-4b14-9aed-dca1c69052d4 | easy | adwhit | 2026-08-14 15:17:17.946422+00:00 | succeeded | 48728092cb22ceae319da96ebbf3e047c303136c5970b4e6d2ca0a7ece442c4d | 3,235 | null | """Statistics-floor CONTROL, not a contender.
A model that is structurally incapable of arithmetic: its logits at position i
depend ONLY on i's offset from the RESULT marker — a learned frequency table of
"which token tends to appear at answer position k". It cannot see x, N, or T.
Whatever eval loss this reaches is t... | {
"id": "1ff253aa-adb5-4b14-9aed-dca1c69052d4",
"created_at": "2026-08-14 15:17:17.946422+00:00",
"db_md5": "376846a32e6935f7be5b6df6c093855a",
"submitter": "adwhit",
"github_login": "adwhit",
"run_id": "7527c7a0-fbe4-42ce-a533-5987f300d03d",
"tier": "easy",
"dataset_id": "e2",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.2644245624542236,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011041667079553008
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2771387100219727,
"example_count": 30... |
1ff548c3-cfc0-4dca-912d-a313a11bc758 | easy | DDanlov | 2026-08-29 13:21:45.180397+00:00 | succeeded | 7f16f63bd3f571df1fdf46c5afc1234a0577c2e741a144ab8fe60ed7641e4ae0 | 24,087 | null | """
trial_13_high_lr_fast_shampoo.py
T13: High Learning Rate (lr=0.005, update_freq=5) on Preconditioned HeavyBall (K=8).
"""
import math
import time
import contextlib
from dataclasses import dataclass
from typing import Optional, Tuple, Dict, Any, List, Union
import torch
import torch.nn as nn
import torch.nn.functio... | {
"id": "1ff548c3-cfc0-4dca-912d-a313a11bc758",
"created_at": "2026-08-29 13:21:45.180397+00:00",
"db_md5": "9fd6b60cdd3da7695ea1e918ef490da5",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "14d8abec-f4e8-4d8c-8858-55a62f0ab410",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.2960433551718604,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.012916666703919569
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3346485557042964,
"example_count": 60... |
1ffd79bc-fe84-44f5-aade-bcd3b21ee5ab | easy | yunjiangster | 2026-08-26 23:25:42.078710+00:00 | succeeded | dc80015e0cd465bded00bf6d486c6296afb45d44c75b25d5a2a003d512ba97ce | 22,736 | null | """Quantized Markov Digit Machine for the One Layer Deeper benchmark.
The model parses the public prompt format into fixed-width decimal tapes, applies
one learned, tied transition once per requested time step, quantizes the residue
back to categorical digits after every step, and right-aligns the resulting
state with... | {
"id": "1ffd79bc-fe84-44f5-aade-bcd3b21ee5ab",
"created_at": "2026-08-26 23:25:42.078710+00:00",
"db_md5": "c39f5f2a40ad111e6c4f7399d26d6828",
"submitter": "Yunjiang Jiang",
"github_login": "yunjiangster",
"run_id": "578c8f89-aaa6-47bb-9065-986c3f46f418",
"tier": "easy",
"dataset_id": "e5",
"status":... | {
"score": {
"mean_loss": 7.812041146106786,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666666666668
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.963008025301946,
"example_count": 600,... |
20025620-4c85-4ac4-b643-d49751b91145 | easy | 0Chris5R | 2026-08-02 16:33:49.813404+00:00 | succeeded | 33e58b936d14d63acff2bc246e9f8353504ffdbb56e968006203aa10833669bc | 24,606 | null | """Shared-digit Neural GPU with late full-objective EMA evaluation.
The model aligns decimal digits by significance, forms a learned low-rank
second-order feature map of the current state, and evolves a two-dimensional
workspace with the original convolutional-GRU equations. Each arithmetic
transition applies the sam... | {
"id": "20025620-4c85-4ac4-b643-d49751b91145",
"created_at": "2026-08-02 16:33:49.813404+00:00",
"db_md5": "911e744b595fc11bddb067f7ff1f6b7d",
"submitter": "Chris ",
"github_login": "0Chris5R",
"run_id": "fd7b47d1-4373-47c5-b47f-d011a857a5be",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.546759474782058,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.027499999999999997
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3327112967657935,
"example_count": 600... |
2005b50c-ec3c-4f91-91d3-068bbf68a29a | easy | nikolageorgiev2000 | 2026-08-04 11:26:24.748132+00:00 | failed | a0b87afe4f0c3b86c94ede89fca29d08d7822036a51f47196fa3c8938f86aa98 | 11,051 | null | """T-composed categorical-digit reasoner with fully learned pair/reduction maps."""
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,
TokenLossBat... | {
"id": "2005b50c-ec3c-4f91-91d3-068bbf68a29a",
"created_at": "2026-08-04 11:26:24.748132+00:00",
"db_md5": "beba8561d728b9acb5bb51752c3b9ea1",
"submitter": "Nikola Georgiev",
"github_login": "nikolageorgiev2000",
"run_id": "45642e09-b959-469c-ac8b-d7220589e478",
"tier": "easy",
"dataset_id": "e2",
"s... | null
|
200cb039-088b-4e78-8b1d-4c58b34ef335 | easy | erdavis0 | 2026-08-27 03:42:59.085242+00:00 | succeeded | f0f048421bab235b2c1c7f2958a7cff30fbcaa7cef983b31931d270fd5484080 | 31,301 | null | """Typed-prefix reversible transition B2: canonical sequence mixture.
Public delimiter syntax routes opaque N, X, and T fields into three strict
functional roles. X initializes only an evolving auxiliary tape, N initializes
only an invariant environment lane, and T reaches only a learned monotone
prefix controller. ... | {
"id": "200cb039-088b-4e78-8b1d-4c58b34ef335",
"created_at": "2026-08-27 03:42:59.085242+00:00",
"db_md5": "84a6cfec2208ae11b2788eb24985c9f5",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "47deccaa-5e30-4b13-a823-db5cfecdb1af",
"tier": "easy",
"dataset_id": "e7",
"status": "succeeded",... | {
"score": {
"mean_loss": 4.1975098848342896,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07736185565590858
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.618282794952393,
"example_count": 85,
... |
200f79f9-9fcb-4e8c-b6b4-7958dae1a78c | easy | c-nnr | 2026-08-04 11:46:32.201386+00:00 | succeeded | 332d65087f0ac021933b29ef382ba66bc1310639a8912c4e4c65034a91d6cb46 | 3,971 | null | """One Layer Deeper experiment: Two tied loops with identity-like zero-initialized residual outputs."""
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": "200f79f9-9fcb-4e8c-b6b4-7958dae1a78c",
"created_at": "2026-08-04 11:46:32.201386+00:00",
"db_md5": "71fd143347b2dc9ea71cc3943b107ed1",
"submitter": "c-nnr",
"github_login": "c-nnr",
"run_id": "8ba5918a-7e96-474e-aadd-4393f2b58a38",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.3132355213165283,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0283333333209157
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6845664978027344,
"example_count": 100,... |
201f3203-c9a5-451d-8224-43397d9801d1 | easy | lpbb | 2026-08-14 08:33:06.722162+00:00 | succeeded | 5ca6af37fd53148130e91cf86e2729ededdf5f27c94d309d8ee4d8dd89e60982 | 7,080 | null | """S4 — S2 with digit embeddings initialized as multi-frequency circle features
(the full Fourier basis of Z_10), tilted into the full width by a fixed random
rotation. Everything stays learnable; only starting values change."""
from __future__ import annotations
import math
import torch
import torch.nn.functional a... | {
"id": "201f3203-c9a5-451d-8224-43397d9801d1",
"created_at": "2026-08-14 08:33:06.722162+00:00",
"db_md5": "a636f0e52666b7aeb4bf085db83040e5",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "3e3e8698-46b8-4ec0-9383-249bb2f73e58",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 5.911176089506942,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00687500006519258
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.980215993406996,
"example_count": 800,
... |
202896bb-c361-4bb7-9794-e22e887d9c6c | easy | rohseh303 | 2026-08-09 08:17:45.641770+00:00 | succeeded | ce5aff9da56cc762aefabac58c10c6def507d15126ddea0f59dd0f2403111d62 | 7,090 | null | """Compact recurrent Transformer tuned for regularized rule learning.
One small block is reused for eight refinement rounds. The deliberately low
capacity and strong weight decay make memorizing the tiny Easy E1 training set
less attractive, while the small batches allow many updates in one minute.
"""
from __future... | {
"id": "202896bb-c361-4bb7-9794-e22e887d9c6c",
"created_at": "2026-08-09 08:17:45.641770+00:00",
"db_md5": "8d2c24d453d12ca8dbd528572f0d3c48",
"submitter": "Rohan Sehgal",
"github_login": "rohseh303",
"run_id": "f3a48ddd-7c60-42b8-8f5e-7eb0b4fe60de",
"tier": "easy",
"dataset_id": "e1",
"status": "suc... | {
"score": {
"mean_loss": 3.533276319503784,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999921768904
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.5635735988616943,
"example_count": 100,... |
202d842e-3470-43ae-90cd-3c408811a6a9 | easy | ehonig | 2026-08-20 10:04:10.110080+00:00 | succeeded | 1c193c4f3d2278ae3d0a4eb45840b2259bc53f27fdb69476da71befb7060f72d | 24,999 | null | """A plain Transformer for repeated modular squaring, and nothing else yet.
This is a deliberate restart. The previous submission accumulated a field
parser, a carry scan, a learned reciprocal, periodic features, a digit
bottleneck, a depth selector and two cell types, most of them tuned around a
memorisation table th... | {
"id": "202d842e-3470-43ae-90cd-3c408811a6a9",
"created_at": "2026-08-20 10:04:10.110080+00:00",
"db_md5": "76b86124de744621e7c00fe4a4591835",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "dc62e71b-a502-43ff-88b0-048d39cef8d5",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 1.9049361944198608,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016666666604578495
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8798168897628784,
"example_count": 10... |
202de380-4cdd-4af7-b08f-81255a85b9fd | easy | sankalp1999 | 2026-08-04 11:39:01.179067+00:00 | succeeded | 77cc6871190c1f9a648e3a9732ba3148078eb3b5cad75ff71f05b404cd5ade1d | 4,868 | 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": "202de380-4cdd-4af7-b08f-81255a85b9fd",
"created_at": "2026-08-04 11:39:01.179067+00:00",
"db_md5": "e6489669d403e75c3877811cdc307320",
"submitter": "Sankalp ",
"github_login": "sankalp1999",
"run_id": "35c6557c-680b-4677-b5b8-65a85d6090fa",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 8.433386325836182,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03833333309739828
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.728765487670898,
"example_count": 100,
... |
202e809b-8a9f-47ef-bd12-568ea3bc5ad1 | easy | gauravmishra | 2026-08-08 21:19:14.128324+00:00 | succeeded | b0b55fee7e56e6941e5a1bc53d69113790983b1be1b71bbe629ff7460b3da524 | 4,643 | 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": "202e809b-8a9f-47ef-bd12-568ea3bc5ad1",
"created_at": "2026-08-08 21:19:14.128324+00:00",
"db_md5": "161f77f076521ee63b89c3e3b8b443d3",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "07174bfc-56cd-4925-afb0-c40c331c1a4c",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.2214376199708328,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004166666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.23983568460952,
"example_count": 600,... |
202f6b62-4d60-4886-98db-d3e259554831 | easy | karanganesan | 2026-08-06 04:04:30.806104+00:00 | succeeded | 31af5240ebb58f7c0253ff326cdaee92503b783bee41e3d8272cd41bc2e5a5b8 | 23,810 | 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": "202f6b62-4d60-4886-98db-d3e259554831",
"created_at": "2026-08-06 04:04:30.806104+00:00",
"db_md5": "b1e053f3ebeb2f44c11b75c3f76c9711",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "c8992e24-16ed-4ecf-bf37-10a67e5c3623",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 8.093467950820923,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01666666753590107
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.6139655113220215,
"example_count": 100,... |
20314b01-9d42-4f9e-8427-939eba0b9295 | easy | jordanrubin | 2026-08-05 18:52:57.238443+00:00 | succeeded | c6ae474a482c73a95da12e2875a2516ea6d46717cb3d3ac3523df943516670c4 | 11,503 | null | """Continuous-state 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 times.
T controls o... | {
"id": "20314b01-9d42-4f9e-8427-939eba0b9295",
"created_at": "2026-08-05 18:52:57.238443+00:00",
"db_md5": "53d4eaeccd100dc64901c7682e141a66",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "4ab3fdbb-44aa-4eeb-b489-f5ce1a703acb",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.941238670234404,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.12020833429414779
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.355448905599242,
"example_count": 300,
... |
2035d71a-16ac-443b-b4d5-dec44f50d1d6 | easy | isaac0804 | 2026-08-05 08:38:50.958731+00:00 | succeeded | 0ab40eb7cfa42b9700f7e4a4f170eba07ae49a93a935f124b68d54c84d706635 | 9,365 | null | """Weight-tied looped Transformer (v6 core: damped gate, 2 sublayers/iteration)
plus explicit place-value embeddings for N's and x's digit tokens.
Motivation (see docs/findings.md, N-conditioning section): every architecture
tried so far treats N's (and x's) digits as an opaque span for generic
attention to notice imp... | {
"id": "2035d71a-16ac-443b-b4d5-dec44f50d1d6",
"created_at": "2026-08-05 08:38:50.958731+00:00",
"db_md5": "6c332bc632593f83d44af5125cddee51",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "a7267683-a80a-425d-88df-3efd7721114f",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.258008549585176,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003750000009313226
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.334939779731773,
"example_count": 800,... |
203da081-81b4-44fd-86a8-617a647fc2d8 | easy | DDanlov | 2026-08-12 23:05:12.898982+00:00 | succeeded | 88cac49667d55d1dcececf8243a1a64244f82f24a9dadb35998a7baed8af1f2d | 28,966 | null | from __future__ import annotations
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as cp
from torch.optim.optimizer import Optimizer
try:
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
... | {
"id": "203da081-81b4-44fd-86a8-617a647fc2d8",
"created_at": "2026-08-12 23:05:12.898982+00:00",
"db_md5": "b0cd394390a4839b67dc8fc2c7fa0cbc",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "4fda34a5-187b-471f-b52f-ab5029dc84d8",
"tier": "easy",
"dataset_id": "e2",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.3147868442672617,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3127248666566147,
"example_count": 300,
"exa... |
203fa22d-50c2-4da0-86ed-4ab8b0b6130c | easy | jordanrubin | 2026-08-14 01:29:23.898058+00:00 | succeeded | fc98447fc19a656845f48580dd88f53c0b2d7d00c35e32a8fe2d3dfff79a73f4 | 29,581 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "203fa22d-50c2-4da0-86ed-4ab8b0b6130c",
"created_at": "2026-08-14 01:29:23.898058+00:00",
"db_md5": "d08ba312cd6460ffc449c5d2476626a5",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "58f94fb5-c1a7-4c56-a421-2a4112c5bb5c",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.8649737620692743,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02249999949708581
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.231652056422469,
"example_count": 600,... |
204bb0be-1ea8-4931-a154-0a5c728e88ba | easy | jordanrubin | 2026-08-08 10:41:58.084721+00:00 | succeeded | 19e434c7e9fc9d9d601292a0f5c0a4083813bc7790caf87c0582fc1bbcc2cad8 | 16,909 | 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": "204bb0be-1ea8-4931-a154-0a5c728e88ba",
"created_at": "2026-08-08 10:41:58.084721+00:00",
"db_md5": "671c15c2db39c821999bae96a7defa51",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "1c19e553-5ed4-4f6c-a55b-b8cf9dee7b8f",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.978341693888233,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333091189463
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.9377526161114735,
"example_count": 600... |
2050d8c2-549e-47d3-958d-a4c62a93015a | easy | saketirl | 2026-08-04 04:28:00.841626+00:00 | succeeded | 8db534f75debd7465932eed8887b472b4ce422e8aaac4e40dcddcf02e0c246a1 | 7,810 | null | """S4D (diagonal structured state space) model trained with Adam.
Stack of bidirectional S4D blocks (S4D-Lin initialization, closed-form
Vandermonde kernel, FFT convolution) with gated MLPs between them. All
layers' kernels are built in one vectorized pass per forward.
The evaluator harness runs ~230 ms/step on e1 re... | {
"id": "2050d8c2-549e-47d3-958d-a4c62a93015a",
"created_at": "2026-08-04 04:28:00.841626+00:00",
"db_md5": "182e555ddb8b8c537482b6e28904a9c3",
"submitter": "saketirl",
"github_login": "saketirl",
"run_id": "0833313a-a432-4965-ba57-fe09e4173af7",
"tier": "easy",
"dataset_id": "e1",
"status": "succeede... | {
"score": {
"mean_loss": 2.19201397895813,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333302289248
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3850672245025635,
"example_count": 100,... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.