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 |
|---|---|---|---|---|---|---|---|---|---|---|
311c94d1-7db8-4882-8914-17d2cd29f806 | easy | jordanrubin | 2026-08-06 08:01:31.616408+00:00 | succeeded | 186588a96ac4f132146f63ec88637f77031ca1cf6339984ac830ee21352cdc66 | 14,772 | null | """Global-binding categorical looped Transformer for public Easy e1.
The recurrent boundary is exactly a 16-place by 10-symbol categorical state.
Each tied cell reconstructs a learned D=256 global residue fingerprint from
that entire state, carries it as the first token of a residual Transformer,
injects it into every... | {
"id": "311c94d1-7db8-4882-8914-17d2cd29f806",
"created_at": "2026-08-06 08:01:31.616408+00:00",
"db_md5": "ea4ae85a41233edad1a8e33cff81f4a3",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "c63438aa-efd3-4fff-9559-7dccb55f6a84",
"tier": "easy",
"dataset_id": "e1",
"status": "s... | {
"score": {
"mean_loss": 1.8589980006217957,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07999999821186066
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.7663612365722656,
"example_count": 100... |
311f1d0c-02af-4d02-8e2d-994a4ee3635a | easy | DDanlov | 2026-08-09 00:39:05.334422+00:00 | succeeded | 336dab7c7f688d23eb45ba40127ce89d509695692d183039e0957407b4d8d841 | 15,978 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with eigenvalue ma... | {
"id": "311f1d0c-02af-4d02-8e2d-994a4ee3635a",
"created_at": "2026-08-09 00:39:05.334422+00:00",
"db_md5": "1c1481c8c5835d8fc0f7de55a4a49de8",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "395adb4e-6617-4532-a4bc-c704e214979d",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 119.87782056835479,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0004166666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 122.31125293184289,
"example_count": 6... |
3126427e-cdf8-4e25-b180-7aa966cb459c | easy | shreyash-chonkie | 2026-08-26 18:25:32.489975+00:00 | succeeded | 5ccd3663213ddfc1b633ffaea5f6fa196bcc45ec2c9a18bf3cbcb044d2c69262 | 8,963 | null | """Narrow ternary spatial program with a prompt-closed local executor."""
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,
)
COMPILER_DI... | {
"id": "3126427e-cdf8-4e25-b180-7aa966cb459c",
"created_at": "2026-08-26 18:25:32.489975+00:00",
"db_md5": "9e30a1fdcb63dd10a5373fd1151c22c4",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "6c9bd87c-f1f4-4ffc-8b60-444b65e6c5b5",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 5.005200147628784,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05499999783933163
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.429709434509277,
"example_count": 100,
... |
312a3ac0-4327-4052-beb9-6f798aefb07f | easy | jordanrubin | 2026-08-08 10:39:14.286557+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": "312a3ac0-4327-4052-beb9-6f798aefb07f",
"created_at": "2026-08-08 10:39:14.286557+00:00",
"db_md5": "671c15c2db39c821999bae96a7defa51",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "001995da-1235-437a-a0c6-40f8e1c066d5",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.6454446975806816,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666517655055
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6715983646867523,
"example_count": 60... |
3134269b-a169-489c-9beb-e4df32e6533c | easy | briheet | 2026-08-07 01:46:17.352071+00:00 | succeeded | 96cf2f738d7911de387a6a21d648a7272ef717cabfe022aaa242be09a2ce6761 | 5,030 | null | """Basic single-pass Transformer with PyTorch AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 96
NUM_HEADS = 4
X_... | {
"id": "3134269b-a169-489c-9beb-e4df32e6533c",
"created_at": "2026-08-07 01:46:17.352071+00:00",
"db_md5": "3c362223c46ba6cd4b80e246e978d77e",
"submitter": "Aizenn",
"github_login": "briheet",
"run_id": "fba5ec40-9036-4964-8cb2-e6cc95216d32",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.4052846431732178,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.048333331476897
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.400663375854492,
"example_count": 100,
... |
31358a24-bbf0-4217-b4b3-6561d555a867 | easy | yashkant | 2026-08-12 04:05:47.396185+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": "31358a24-bbf0-4217-b4b3-6561d555a867",
"created_at": "2026-08-12 04:05:47.396185+00:00",
"db_md5": "520fd45ba52d67a06d79ff8be9c27218",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "74e40886-725e-4897-ab58-0f656d857067",
"tier": "easy",
"dataset_id": "e4",
"status": "succeed... | {
"score": {
"mean_loss": 2.0641446092680034,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013645833345750969
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.066723166678637,
"example_count": 120... |
313a5ef5-054b-4b3a-93bc-e9ecf25d1492 | easy | DDanlov | 2026-08-12 21:03:58.434099+00:00 | succeeded | 0286a59512605c7ac56e5e79cc82d56962200aedefda8053c0517b29c7419562 | 25,388 | 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": "313a5ef5-054b-4b3a-93bc-e9ecf25d1492",
"created_at": "2026-08-12 21:03:58.434099+00:00",
"db_md5": "181d48a2f4d63e21d76d94d1779a6fcb",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "c55a7d0e-6917-4602-b27e-081649ee4c0b",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.281911473497625,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.283824036474185,
"example_count": 600,
... |
3143caa9-4fd4-4d6c-9375-6b26902b1030 | easy | DDanlov | 2026-08-20 22:49:51.409248+00:00 | failed | e63234e753d9b96ecfa3b0231bbd6f9fc72f95604363273385e1228b43c9537c | 21,253 | null | from __future__ import annotations
import math
import time
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except Imp... | {
"id": "3143caa9-4fd4-4d6c-9375-6b26902b1030",
"created_at": "2026-08-20 22:49:51.409248+00:00",
"db_md5": "c289a8f52a18e75e2a2549d392d812ad",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "f5a5d295-6cc3-4e2c-b914-47a0dd82e14f",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
3144a226-37fa-4355-9658-9c18406634a0 | easy | lzy54 | 2026-08-29 08:27:56.983854+00:00 | succeeded | 2198c3e036cb63ce07ed516c32c16517b7cd5fafc3e2aa65b0b27ed0fb8ccf49 | 4,032 | null | """Phase 2 C1: four untied blocks with additive input recall."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 128
NUM_HE... | {
"id": "3144a226-37fa-4355-9658-9c18406634a0",
"created_at": "2026-08-29 08:27:56.983854+00:00",
"db_md5": "859baf35963cf1712b065b05e50b68ce",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "3259e4e5-3de4-40df-b951-d15498ce2617",
"tier": "easy",
"dataset_id": "e7",
"status": "succeeded",
... | {
"score": {
"mean_loss": 3.3657089471817017,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0441622119396925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.7145872116088867,
"example_count": 85,
... |
3149aee2-6b3b-4356-a019-fa1a7bdca902 | easy | erdavis0 | 2026-08-23 04:34:26.349963+00:00 | succeeded | b55593fc1495c8808062cd69554206bd2e9b09136f54bb7e28ab99edbd357bcd | 13,940 | null | """Field-aligned learned digit machine.
The fixed code in this model only exposes the public prompt grammar: delimiter
tokens select three fields and each field is right-aligned without interpreting
the value of any digit. All cross-field computation, local propagation,
workspace updates, and rendering are learned fr... | {
"id": "3149aee2-6b3b-4356-a019-fa1a7bdca902",
"created_at": "2026-08-23 04:34:26.349963+00:00",
"db_md5": "e47e69188792d13dc82454d0c65beec0",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "9122a500-5918-420f-b4b5-af299728a9a5",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",... | {
"score": {
"mean_loss": 7.346289868656732,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013125000009313227
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.474901374268041,
"example_count": 800,... |
314aff95-8a54-4e18-b330-2229c3f47b54 | easy | sapient-sapiens | 2026-08-07 08:05:20.645346+00:00 | succeeded | 7bd6ec64bda5f80f3263ec084b6d937d1a21a1e5a60409a987e7e4e4dbbad597 | 12,320 | null | """Digit-space recurrent operator for repeated modular squaring.
Nothing here converts a token into a decimal number. Digits stay as embeddings,
so what a digit is worth, and every carry and reduction, is learned. That also
sidesteps a hard numeric wall: at Hard's modulus scale `x^2` runs past the 53-bit
exact-integer... | {
"id": "314aff95-8a54-4e18-b330-2229c3f47b54",
"created_at": "2026-08-07 08:05:20.645346+00:00",
"db_md5": "39e0d378256dd1b37c17dbd0d34a82fa",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "53d6868f-f73d-4e78-97bd-c6aa4168b9e1",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 2.6238696404228783,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03999999995032946
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8474501371383667,
"example_count": 100... |
314d22fa-f8df-4888-b70e-953addfd4863 | easy | sapient-sapiens | 2026-08-14 17:59:44.276076+00:00 | succeeded | 28ba469aae3e42ca679270105f219b6bd47433fa80f83dc154b3c3e55e44c7c3 | 14,299 | null | """Digit-space recurrent operator for repeated modular squaring.
Nothing here converts a token into a decimal number. Digits stay as embeddings,
so what a digit is worth, and every carry and reduction, is learned. That also
sidesteps a hard numeric wall: at Hard's modulus scale `x^2` runs past the 53-bit
exact-integer... | {
"id": "314d22fa-f8df-4888-b70e-953addfd4863",
"created_at": "2026-08-14 17:59:44.276076+00:00",
"db_md5": "1efadda9e01ae84ec2857432a666c542",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "5f0dbb2a-f210-4e25-a591-fadcda0b7a47",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.524677844179389,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009166666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.7382781140740144,
"example_count": 600... |
315f171f-4781-4415-85a7-d79d951861aa | easy | karanganesan | 2026-08-05 09:56:53.600781+00:00 | succeeded | 78d820c26e876655daacd4066b392a356bd9566d7acd5d8166ad09b310b8279a | 22,082 | 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": "315f171f-4781-4415-85a7-d79d951861aa",
"created_at": "2026-08-05 09:56:53.600781+00:00",
"db_md5": "b73c579101bc1f7e2d798195fd973c2e",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "a9873057-e4fc-4fad-87a2-55d1d20ce1b2",
"tier": "easy",
"dataset_id": "e3",
"status": ... | {
"score": {
"mean_loss": 7.601091146469116,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006874999962747097
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.514750957489014,
"example_count": 800,... |
317868ca-1dbe-4242-b002-435de7c50be9 | easy | sapient-sapiens | 2026-08-15 20:06:21.104869+00:00 | succeeded | 79a82b8d2bf48b72382828a3ae6685a8e31df862b4b87fe5e1f6afd5995a6afd | 17,774 | null | """T-only adaptive chunk routing over a learned shared operator.
The controller sees only the categorical tokens inside the T prompt field and
is trained solely by final-answer loss. It never parses T numerically and T is
not used as an intermediate supervision target. Inputs use learned additive
token, semantic-gro... | {
"id": "317868ca-1dbe-4242-b002-435de7c50be9",
"created_at": "2026-08-15 20:06:21.104869+00:00",
"db_md5": "98fbb4c546d1ce23ff0be7df7febab69",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "54e654af-1b84-4896-8fea-cf50148e4519",
"tier": "easy",
"dataset_id": "e6",
"status": "su... | {
"score": {
"mean_loss": 5.972736290056412,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08851351649374575
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.065245151519775,
"example_count": 60,
... |
317bf351-65cc-40f4-8bc6-58da677d4bcc | easy | khushidahi | 2026-08-07 04:02:56.950980+00:00 | succeeded | f88d20032f5934305432fc7bbfa909869d3a27ba4d233dc25678d9b4cac044d2 | 17,795 | 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": "317bf351-65cc-40f4-8bc6-58da677d4bcc",
"created_at": "2026-08-07 04:02:56.950980+00:00",
"db_md5": "799d7714aaca0c2e6c359c2ec61df995",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "2e38f9b5-b847-4fca-8520-3f341b8340a3",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.171484351158142,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333507180214
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.170536518096924,
"example_count": 600,... |
317c16be-410a-4e65-9f3b-444288c71787 | easy | nikhilkumar26 | 2026-08-30 02:17:17.802853+00:00 | succeeded | 49c3f9c46a2477665158561a02d6703cb5aabc12b3e7c298c40679696858fde8 | 9,220 | null | """T-adaptive tiny recurrent model for One Layer Deeper.
T is parsed from the prompt and used only as the unroll count. The shared
residue step never sees T, so a learned one-step map can compose to T=64.
At eval the residue is discretized to digits after every step so errors do
not accumulate as latent drift.
"""
fr... | {
"id": "317c16be-410a-4e65-9f3b-444288c71787",
"created_at": "2026-08-30 02:17:17.802853+00:00",
"db_md5": "50c8b0aff0c37efbad1b44beda2540b8",
"submitter": "Nikhil Kumar",
"github_login": "nikhilkumar26",
"run_id": "fb2be147-dbc0-4daa-8c25-0933f49bb1f2",
"tier": "easy",
"dataset_id": "e6",
"status": ... | {
"score": {
"mean_loss": 1.880744216376788,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.060360362199512685
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.870038390159607,
"example_count": 60,
... |
317fba70-e97a-43ae-a0cc-aab78ffd11ed | easy | rishivg | 2026-08-04 09:35:35.546706+00:00 | succeeded | 539af6143e29cdb0d3b3b5db630d742da7b0ad4071669f33a85c9ed32d94c0b9 | 131,327 | null | """Looped transformer with place-value coordinates, Muon, and a clock-paced LR.
Design notes, in rough order of expected value:
1. Weight-tied recurrence with input injection. The task is function
composition of depth ``T``, so one block applied ``LOOPS`` times is a much
better prior than a single feed-forw... | {
"id": "317fba70-e97a-43ae-a0cc-aab78ffd11ed",
"created_at": "2026-08-04 09:35:35.546706+00:00",
"db_md5": "f44dbee84791e3877f32445b604d6bf7",
"submitter": "Rishi Gottumukkala",
"github_login": "rishivg",
"run_id": "86fe6fe6-b21c-4acf-a86c-91e05a0f8ca3",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 16.219677448272705,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.012500000186264515
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 15.837912559509277,
"example_count": 60... |
31933604-7b7d-41c1-9659-db2a5ef0835f | easy | AFeinfield | 2026-08-19 23:16:43.788878+00:00 | succeeded | bf333e291b7fa2382a52de2b2dcc4dd185d0f5064ad3b957aa01a16f817e9696 | 19,260 | null | """092: INSTRUMENTED BASE — two measurement fixes, no architecture change.
(1) Step-indexed LR schedule with a fixed screening step budget. The old
schedule decayed on WALL CLOCK, so throughput jitter changed the
LR-per-step trajectory: identical files on e2 gave 1470 vs 1025 steps
and train acc 0.725 vs 0... | {
"id": "31933604-7b7d-41c1-9659-db2a5ef0835f",
"created_at": "2026-08-19 23:16:43.788878+00:00",
"db_md5": "2dcbfc96ca5a2d30e8fddecf166d77d7",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "5de9aaf7-8ed0-4c5e-aa7a-9aa99d4172ac",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 5.530302286148071,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011458334047347307
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.497442722320557,
"example_count": 300,... |
3196d222-bc0b-4d01-8314-87e90acfefc0 | easy | akkolli | 2026-08-04 18:23:09.082330+00:00 | succeeded | cee61768b8bdcd6f602a60e76a303bf448a4dd31d549ed5b062c3a6357778db5 | 51,270 | null | """One learned arithmetic and learned-width-reference trajectory mixture.
The decimal step count drives three independent recurrent trajectories with
shared learned parameters. One applies a two-factor Neural Multiplication
Unit directly, one applies that arithmetic before a learned length-relative
digit transport, a... | {
"id": "3196d222-bc0b-4d01-8314-87e90acfefc0",
"created_at": "2026-08-04 18:23:09.082330+00:00",
"db_md5": "de30a82b636bf22c54e476d4a50c5f85",
"submitter": "Akshay Kolli",
"github_login": "akkolli",
"run_id": "006f4277-55e9-42af-96a8-f7661faf5a39",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 0.10875302404039414,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.1087530255317688,
"example_count": 600,
"ex... |
3197ce46-f28b-4499-9825-cefa9f32cb9c | easy | khushidahi | 2026-08-21 17:22:21.409653+00:00 | succeeded | a2818ebe65bfe6b23892ade07a4366819b7ba3aa6afceeb39a76fb6fc6033ec2 | 29,576 | null | """R342: update-matched four-visit inner-recurrence control from R340.
This submission tests the strongest rules-safe generic baseline suggested by
the public competition discussion. It does not implement squaring, modular
reduction, or any other task solver. Numeric prompt fields are only rearranged
onto equal-leng... | {
"id": "3197ce46-f28b-4499-9825-cefa9f32cb9c",
"created_at": "2026-08-21 17:22:21.409653+00:00",
"db_md5": "69ed14f9d943d160e04881346a5ba36b",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "abb70af2-089c-4722-9ee0-7e43bc1450a7",
"tier": "easy",
"dataset_id": "e6",
"status": "succ... | {
"score": {
"mean_loss": 0.7727893590927124,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.8200450837612152
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.5301251411437988,
"example_count": 60,
... |
319b80bb-cfdc-4b22-956a-bb08b5d9d8b6 | easy | chad-atexpedient | 2026-08-20 15:28:08.207762+00:00 | succeeded | 76916fcf11bccbce8691a6f15b5bb68cc34c2ad350057a82b955bdc2504cf6bb | 23,830 | null | """X113: X105 with an RNG-parity-fixed ordinal digit decoder.
X107's ordinal loss left the unconstrained output head unchanged. X113 adds a
small residual decoder whose ten digit logits are generated from learned latent
center and sharpness coordinates, imposing smooth ordered output geometry. X105's
shared tensors re... | {
"id": "319b80bb-cfdc-4b22-956a-bb08b5d9d8b6",
"created_at": "2026-08-20 15:28:08.207762+00:00",
"db_md5": "f8d3e2b042e4530253756ca85b0ccfb7",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "b00eca56-7a1e-46ab-a8f5-71ca748ac888",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.30795396300737,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3563936644070886,
"example_count": 600,... |
31a4a407-ea50-4e73-bbad-e1d9a3bc3d36 | easy | Maxwells-Demons | 2026-08-11 20:22:26.898005+00:00 | succeeded | 3d5d7a13cfa67e34369aa472e02cc1d52461af61f34261037e9a10ed61a8930c | 73,786 | null | """One shared digit operator, its two free neighbours of x, and two learned maps.
The backbone is unchanged from v47/v96: the state IS the digit one-hot, one
application of the operator is
(state digits, N digits) -> next state digit logits
and the straight-through code of those logits is the whole state handed to... | {
"id": "31a4a407-ea50-4e73-bbad-e1d9a3bc3d36",
"created_at": "2026-08-11 20:22:26.898005+00:00",
"db_md5": "fbfa6e6586580fcad2856779feb5c9ed",
"submitter": "Maxwells_Ayakashi",
"github_login": "Maxwells-Demons",
"run_id": "6f00682c-8338-496e-a54c-8296d88728b7",
"tier": "easy",
"dataset_id": "e4",
"st... | {
"score": {
"mean_loss": 1.7355441431175387,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.45229166716337205
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5946724197433313,
"example_count": 120... |
31a7e46e-08d9-47ac-9101-940884988e60 | easy | DDanlov | 2026-08-20 11:40:30.902378+00:00 | failed | 0c4542096f9e1c6e2d60058900aa683dd7f789aa7bf7d9cc3239635a1e696c2f | 20,762 | null | from __future__ import annotations
import math
import time
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except Imp... | {
"id": "31a7e46e-08d9-47ac-9101-940884988e60",
"created_at": "2026-08-20 11:40:30.902378+00:00",
"db_md5": "4fbf6a2f446e39fe61f3e9e8ab1ff7ac",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "2b1eaecb-1631-4887-824c-ea5607d9d3ca",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
31a96e69-8497-4bbe-9e34-0e2ff9057a40 | easy | kushasareen | 2026-08-10 22:16:42.702409+00:00 | succeeded | afe41c2712aacf8a2ef69de16a774caf58f4abaae65c3e567edb7e9fcd9b2ddd | 5,884 | null | """T-conditioned recurrent Transformer.
A tied-weight block is applied once per requested squaring step, so compute
scales with T and the model can extrapolate to depths never seen in training.
Untied pre-blocks encode the prompt and an untied post-block decodes the
answer digits.
"""
from __future__ import annotatio... | {
"id": "31a96e69-8497-4bbe-9e34-0e2ff9057a40",
"created_at": "2026-08-10 22:16:42.702409+00:00",
"db_md5": "83a84974ed856e7e41dd4a9148a00ef6",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "549be51c-4985-4983-8ddc-01a662f5bed7",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 7.739135585078465,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005000000012417634
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.045222718619446,
"example_count": 600,... |
31acaf76-f777-405a-9da2-ec099286af0e | easy | kushasareen | 2026-08-27 03:56:53.201790+00:00 | succeeded | bee10682f835f6d7b299b07b010530e7f8b1ed990a2920b62722416900b18fa9 | 10,901 | null | """loop3_noT: digitcomp (4 blocks) where blocks 2..4 repeat T times; T token dropped.
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 s... | {
"id": "31acaf76-f777-405a-9da2-ec099286af0e",
"created_at": "2026-08-27 03:56:53.201790+00:00",
"db_md5": "ab34578db948e86897d03214756718df",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "5d8bcbce-7f20-4190-89c3-4fe5a250d815",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 2.8126740753650665,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.40791669115424156
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.099938869476318,
"example_count": 300,... |
31adcc6e-b0ee-49fe-9f11-4d98cfc96409 | easy | jordanrubin | 2026-08-11 01:34:23.880946+00:00 | succeeded | 67982e9f505a2cf74a8823191467477f81e0c11b7660f6c116f0fafa565fb3eb | 21,669 | 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": "31adcc6e-b0ee-49fe-9f11-4d98cfc96409",
"created_at": "2026-08-11 01:34:23.880946+00:00",
"db_md5": "5f41f74e5ab605af5d290f14916e4deb",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "331be5f9-e0eb-4340-9aaa-7c5285b2c279",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.3760574310996923,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007499999832361937
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1641744980630317,
"example_count": 60... |
31b1186d-85fb-464f-81c8-fb5234aedbe1 | easy | gauravmishra | 2026-08-08 00:07:37.889219+00:00 | succeeded | f44c55a78e66d2aa7ef0d32760b6388b9c77753aa3d0409dee5cf627bc2e8f34 | 3,755 | null | """Four-pass Transformer with four independently parameterized blocks."""
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 = 12... | {
"id": "31b1186d-85fb-464f-81c8-fb5234aedbe1",
"created_at": "2026-08-08 00:07:37.889219+00:00",
"db_md5": "02e726bbede87b25cc32e4046aae607a",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "72489c9c-7de8-4677-8338-1f74a8a5639f",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 3.7466593621937463,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005416666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.066917050579738,
"example_count": 600... |
31b1560b-6105-4e40-b60b-18bb9941ed85 | easy | lpbb | 2026-08-14 15:32:51.684957+00:00 | succeeded | ac63bbfb9e65ee6fe157d2136cdf2e620e307cf9128a580cb52512d566ca1ed1 | 11,432 | null | """CLEAN S14 — S12 hygiene base with the optimizer swapped to MUON
(orthogonalized momentum on the cell's 2D matrices, AdamW elsewhere).
Question: does a second-order-flavored step-efficiency unlock (diagnosis
option b; grokfast falsified in L1d/s9) move clean e3 where additive
tricks did not? One variable changed vs S... | {
"id": "31b1560b-6105-4e40-b60b-18bb9941ed85",
"created_at": "2026-08-14 15:32:51.684957+00:00",
"db_md5": "cbb9dcee8eb1411dd826dd06276f953d",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "a4096a96-021f-44fd-9cda-306af81df00e",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 5.166506460396702,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011250000000000001
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.07291735026031,
"example_count": 800,
... |
31b9a181-be73-4e0c-a731-ec168c74b1b1 | easy | TheYkk | 2026-08-04 12:44:06.846932+00:00 | succeeded | d611dba1e9ccd40f7025700024536d04ec0ea0629d0b4950053f46f85432ff3b | 11,565 | null | """Encoder / iterated-operator / positional-decoder composition model.
Why this shape
--------------
The label is a composition ``S^T(x)``: one unknown step map applied a
prompt-dependent number of times. A transformer stack -- looped or not --
spreads that map across per-layer parameters and per-position activations... | {
"id": "31b9a181-be73-4e0c-a731-ec168c74b1b1",
"created_at": "2026-08-04 12:44:06.846932+00:00",
"db_md5": "bef360f893dc2c90378cf80cf5fadf26",
"submitter": "Kaan Karakaya",
"github_login": "TheYkk",
"run_id": "13f61edb-810b-4aff-acef-1691abda7f69",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 7.802098989486694,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.026666666381061077
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.523525714874268,
"example_count": 100,... |
31b9ea1b-63f8-4a69-9bb8-17fd713c5d0f | easy | amirmeisami | 2026-08-17 17:44:56.589365+00:00 | succeeded | b8f38bbf52c044fb9a4b28b564eef4b43e4ccd75eff03b887c66157cce24eb74 | 15,723 | null | """T021 Arm A: sparse recurrent memory."""
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,
)
D_MODEL =... | {
"id": "31b9ea1b-63f8-4a69-9bb8-17fd713c5d0f",
"created_at": "2026-08-17 17:44:56.589365+00:00",
"db_md5": "47f809c028a146dc06f08e619fea732e",
"submitter": "Amir Meisami",
"github_login": "amirmeisami",
"run_id": "d8a70e20-8aee-4ce2-9b7c-59b3ba72e922",
"tier": "easy",
"dataset_id": "e7",
"status": "s... | {
"score": {
"mean_loss": 4.829255970716015,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.12112299509067587
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.646028393070872,
"example_count": 85,
... |
31be076c-d1dd-47a9-9ba3-556d7c3ec50d | easy | Maxwells-Demons | 2026-08-06 10:05:10.884801+00:00 | succeeded | 0b1352a5db45837363938726eb7147773ac6a77b38e74f4c6b2b4f53efdb2988 | 14,056 | null | """v7: value-bottleneck recurrence.
The recurrent state is forced to be a soft distribution over the decimal
digits of the current residue (SLOTS x 10). The shared transition g must
therefore implement "digits of v -> digits of v^2 mod N"; per-example
memorization is impossible inside the bottleneck, so trajectories o... | {
"id": "31be076c-d1dd-47a9-9ba3-556d7c3ec50d",
"created_at": "2026-08-06 10:05:10.884801+00:00",
"db_md5": "1c4b4844498884827358c043a10b7168",
"submitter": "Maxwells_Ayakashi",
"github_login": "Maxwells-Demons",
"run_id": "5cf73091-7178-4849-9172-f699a88efd24",
"tier": "easy",
"dataset_id": "e3",
"st... | {
"score": {
"mean_loss": 21.187917709350586,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06125000026077032
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 20.390880584716797,
"example_count": 800... |
31be089c-5e75-47de-b890-f867c505aaed | easy | erdavis0 | 2026-08-25 01:47:11.983572+00:00 | succeeded | 362ea16f0171e3e2ad492a460b5ba6fbdde70a64eea30dea9a6544b2214be1be | 24,120 | null | """Matched categorical-boundary panel: C4 ST oracle plus commitment.
Public delimiter syntax is used only to route opaque token identities onto
separate right-aligned tapes. One T-blind learned macro maps a categorical X
tape and a digitwise N tape to a new categorical tape through freshly reset
interaction, carry, a... | {
"id": "31be089c-5e75-47de-b890-f867c505aaed",
"created_at": "2026-08-25 01:47:11.983572+00:00",
"db_md5": "19b07a0124d2461910b21f1195de2485",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "f17e20b0-8ea4-4c42-85ab-5b8e6e53bbc0",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.074067702075858,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008020833345750968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.075357832181614,
"example_count": 1200... |
31c281df-9270-4bd1-a547-5503e65468cb | easy | kushasareen | 2026-08-10 22:45:19.562383+00:00 | succeeded | 42a726f13a24618c0b2f92cdf7b9c0970c5cc54971c57d8221e23c057a6a9808 | 6,472 | null | """T-conditioned recurrent Transformer.
A tied-weight block is applied once per requested squaring step, so compute
scales with T and the model can extrapolate to depths never seen in training.
Untied pre-blocks encode the prompt and an untied post-block decodes the
answer digits.
"""
from __future__ import annotatio... | {
"id": "31c281df-9270-4bd1-a547-5503e65468cb",
"created_at": "2026-08-10 22:45:19.562383+00:00",
"db_md5": "0468482b4e7fa17719bb8589be384fa8",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "b8da460f-c9cb-48cb-8ad7-3ad1cd1f7503",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 9.348401420963695,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004583333345750968
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.579795418298833,
"example_count": 600,... |
31c6a365-6070-4eb5-b962-c09ea8196f1a | easy | Dandandan | 2026-08-07 03:13:07.390092+00:00 | failed | f60b5998caa8e53a4998a289f36c054b0a66fbd7f2bc7c00253d2e0450c93e60 | 98,978 | null | """Generalized exact recursive learner for One Layer Deeper.
The learned transition is a population of ordinary modular neural units with
several generic arithmetic activations. A single learned evidence selector
chooses one transition, which is then reused for every input-requested outer
step. The population covers ... | {
"id": "31c6a365-6070-4eb5-b962-c09ea8196f1a",
"created_at": "2026-08-07 03:13:07.390092+00:00",
"db_md5": "2613f5d358127dede9d692df8c2ad368",
"submitter": "Daniël Heres",
"github_login": "Dandandan",
"run_id": "d94db894-9682-425c-a5c2-9b9f18261562",
"tier": "easy",
"dataset_id": "e1",
"status": "fai... | null
|
31c9c90d-c2be-470a-a9a9-162f5cd74fd2 | easy | ddoan | 2026-08-14 12:24:16.974345+00:00 | succeeded | 9cf7afce109d6672ab3df424e8dc94a0e20903d75353c510cf8b3bdb68da1f5f | 9,939 | null | """Pinned numeral codec with a directly supervised tied transition."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
BatchReuseContext,
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,... | {
"id": "31c9c90d-c2be-470a-a9a9-162f5cd74fd2",
"created_at": "2026-08-14 12:24:16.974345+00:00",
"db_md5": "64d6ee30811bbb81e65bb88789f8996b",
"submitter": "Doug Doan",
"github_login": "ddoan",
"run_id": "d71cb208-663d-40a4-9646-4cdb0ee792b1",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.954266164733119,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.9499992522423577,
"example_count": 600,
"exa... |
31cfaa51-2446-4296-aff9-58ea6ef50346 | easy | rishivg | 2026-08-10 21:38:35.299751+00:00 | succeeded | 99f6c766177c6a337ca4bc61970c287e6b32d0fa74517dcbe69dd0226b0828dd | 138,757 | null | """Looped transformer with place-value coordinates, Muon, and a clock-paced LR.
Design notes, in rough order of expected value:
1. Weight-tied recurrence with input injection. The task is function
composition of depth ``T``, so one block applied ``LOOPS`` times is a much
better prior than a single feed-forw... | {
"id": "31cfaa51-2446-4296-aff9-58ea6ef50346",
"created_at": "2026-08-10 21:38:35.299751+00:00",
"db_md5": "61ee1154011f54500edf97dcd3595d71",
"submitter": "Rishi Gottumukkala",
"github_login": "rishivg",
"run_id": "6c037447-f561-4636-a252-8382abcb6ffc",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 14.19094467163086,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01583333406597376
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 14.06486701965332,
"example_count": 600,
... |
31d3fc9b-438a-4d9b-bc7c-2ea45d554d0c | easy | jordanrubin | 2026-08-06 19:48:50.120524+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": "31d3fc9b-438a-4d9b-bc7c-2ea45d554d0c",
"created_at": "2026-08-06 19:48:50.120524+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "569d1fc8-c885-4d78-9e62-3da2ae3fc0fc",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.408684878394094,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.2270833283352355
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.550953182747693,
"example_count": 300,
... |
31d85f5e-389b-4f69-ba2a-412ae8ed19bd | easy | repst | 2026-08-23 23:54:42.583733+00:00 | succeeded | 0bf626f616ae5a260342167a04ac963eba78f576e1653c0963768abec86be063 | 90,978 | null | # R547: exposure-aware recurrent program-space machine.
#
# ARCHITECTURE/VOCABULARY (analogous to an LM): generic modular register VM;
# bases {0,1,-1,x,r0,r1}; ADD/SUB/MUL expression grammar and semantic quotient;
# exact recurrent execution; decimal tokenization; MDL/dataflow-generated search
# views; and a balanced ... | {
"id": "31d85f5e-389b-4f69-ba2a-412ae8ed19bd",
"created_at": "2026-08-23 23:54:42.583733+00:00",
"db_md5": "b2f10aaf928c4887737a03aeccdf37e6",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "5070c471-9dc7-4eea-9260-58b87317a737",
"tier": "easy",
"dataset_id": "e1",
"status": "succeed... | {
"score": {
"mean_loss": 0.002199022565037012,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0021479646675288677,
"example_count": 100,
... |
31da5730-eb47-409d-879a-c8a552812dbb | easy | Mantissagithub | 2026-08-19 11:34:46.339871+00:00 | succeeded | 1c7b1d52f346759d4c742fa3f1096c9baa49fe43667f68db87a4493364b4339f | 32,720 | null | """attempt_035 + a contentless T field (depth enters only as iteration count).
ADDENDUM (attempt_062_deeprescale): attempt_057_bs256 + a depth-only rescale of
the captured read-out. Designed to be provably free on test.
The problem, measured rather than guessed. mean|state| per loop iteration in
this file is 11.1, 28... | {
"id": "31da5730-eb47-409d-879a-c8a552812dbb",
"created_at": "2026-08-19 11:34:46.339871+00:00",
"db_md5": "b10d1093f0a66623bac38e9e5570d6bf",
"submitter": "Pradheep P",
"github_login": "Mantissagithub",
"run_id": "ae6282d3-d7b2-4260-8c9b-137eb2442dfb",
"tier": "easy",
"dataset_id": "e2",
"status": "... | {
"score": {
"mean_loss": 8.791507720947266,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.37416670052334666
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 14.960930824279785,
"example_count": 300,... |
31df7024-add8-4b4c-9e7e-f287fcb727fe | easy | DDanlov | 2026-08-07 01:35:11.823557+00:00 | succeeded | f29c2da9f5e6751b8e362f53155142b62439f690455098261d445082c66a0fb6 | 8,781 | null | """
Official Competition Submission:
1-to-1 Sequence-Aligned Deep Energy-Based Reasoning Model.
Features:
- Nesterov-Accelerated SafeShampoo (2nd-Order Kronecker Preconditioning + NAG momentum)
- Pure Single-Path Refined Loss (Dual-Path compete signal removed)
- SNR-Bounded Latent Noise Injection (alpha=0.03)
- Analyti... | {
"id": "31df7024-add8-4b4c-9e7e-f287fcb727fe",
"created_at": "2026-08-07 01:35:11.823557+00:00",
"db_md5": "2c0e2cf96e7851fc2d5bb36b4489f034",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "4f8e2120-7014-412f-ab36-44644e4867af",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.4853737354278564,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0333333327434957
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.609774351119995,
"example_count": 100,
... |
31e6eddc-ccf6-4348-ba3f-7bb29fb47b0a | easy | chuk-uzowihe | 2026-08-12 17:10:28.358396+00:00 | failed | 58ec1c662a679e0759253507283e7cc7afccfd8df35e872b571dc9a97fcb5e14 | 20,514 | null | """v10: dense fixed-loop recurrence with deep supervision.
Lineage: FPRM structure (alpha/beta-tied residual scaling, pre-rms_norm,
input injection every iteration) + the v8 2D digit lattice / RoPE task
encoding. Three deliberate departures from the v9 MoEUT line:
1. **Full BPTT.** The loop runs a fixed N_ITER iterat... | {
"id": "31e6eddc-ccf6-4348-ba3f-7bb29fb47b0a",
"created_at": "2026-08-12 17:10:28.358396+00:00",
"db_md5": "f40f569d019188e4b032a644b15b3a5f",
"submitter": "Chuk Uzowihe",
"github_login": "chuk-uzowihe",
"run_id": "e4d997a9-b409-4a81-b62d-34b91b6ea70d",
"tier": "easy",
"dataset_id": "e1",
"status": "... | null
|
31ec749a-869c-4f72-b10d-d1c875e41f7b | easy | DDanlov | 2026-08-11 19:33:03.762935+00:00 | failed | 442191d1eba4d9ccae2ebe440a003b2b0069096c2c7ec8777121414ea3c60fd8 | 19,092 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with bias-correcte... | {
"id": "31ec749a-869c-4f72-b10d-d1c875e41f7b",
"created_at": "2026-08-11 19:33:03.762935+00:00",
"db_md5": "7f172a0abe2423947869728331505684",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "6a30abb1-9ded-4886-a5b4-1d9e3fae2116",
"tier": "easy",
"dataset_id": "e1",
"status": "failed",
... | null
|
31ef4343-2bc3-4fc6-81d4-3112e058232d | easy | akkolli | 2026-08-04 07:07:31.461499+00:00 | succeeded | e5b7349f255e6951e3b601b3bf2fee9f5114b70c1ec32e0eb1dc0f1a48c2e0e4 | 86,246 | null | """Lazily routed union guard with an ordered-context generic fallback.
The exact branch is one randomly initialized CE-trained selector over 7,056
two-register instruction DAGs and 3,125 full quartic modular recurrences. A
separate learned dataset gate mixes that exact branch with a generic ConvGRU
whose shared recur... | {
"id": "31ef4343-2bc3-4fc6-81d4-3112e058232d",
"created_at": "2026-08-04 07:07:31.461499+00:00",
"db_md5": "06fba819a85023cfc0bfd37535e5826e",
"submitter": "Akshay Kolli",
"github_login": "akkolli",
"run_id": "cd54c9f2-c482-4f64-83ed-85038e90b4b7",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 0.0,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0,
"example_count": 800,
"exact_accuracy": 1.0,
"... |
31ef9f13-d51d-45ce-aa17-27b58480f023 | easy | khushidahi | 2026-08-20 04:32:03.904765+00:00 | succeeded | b7a13d8f2a9d8b5848f33cd4d268740114167f5f0602909455ce2dd36d09d22d | 35,831 | null | """R311: gentle endpoint-product semantic functional root.
This candidate is deliberately diagnostic. It trains on evaluator-owned final
labels at several depths so that one shared transition must compose. The model
never converts N or X to native numbers and never implements multiplication or
modular reduction.
Th... | {
"id": "31ef9f13-d51d-45ce-aa17-27b58480f023",
"created_at": "2026-08-20 04:32:03.904765+00:00",
"db_md5": "1c842a5547fc645567aa14d2cf25d98a",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "0a7bde4e-d55b-4e82-8c11-ef74a6cfbc99",
"tier": "easy",
"dataset_id": "e3",
"status": "succ... | {
"score": {
"mean_loss": 2.254761011783491,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011875000060535967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.240187883978889,
"example_count": 800,... |
31f18e93-cae2-4aee-9c03-8f837817b957 | easy | blake-camp-surge | 2026-08-06 01:00:00.280898+00:00 | succeeded | a9cb8601a7fde9fe1a8541e375819319a4256d4d8c754205ffdabd8d2fb929ba | 10,097 | null | """Deep-equilibrium reasoner with DAN-generated linear attention."""
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": "31f18e93-cae2-4aee-9c03-8f837817b957",
"created_at": "2026-08-06 01:00:00.280898+00:00",
"db_md5": "57fb0b68eab6fd2ea025fed4468f6aca",
"submitter": "Blake Camp",
"github_login": "blake-camp-surge",
"run_id": "dcc5c714-751b-4833-8951-aacdcd74782e",
"tier": "easy",
"dataset_id": "e3",
"status":... | {
"score": {
"mean_loss": 2.2912198671085866,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005625000055879354
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.356462544034798,
"example_count": 800... |
31f7a61d-28d7-419e-baee-c7c85119e876 | easy | blake-camp-surge | 2026-08-05 18:03:14.485916+00:00 | succeeded | 8735b2647d587725bd7ee42af06c7196afc3209abe150b7ab0d29a1a37d2f7f3 | 17,346 | null | """Width-16 linear-memory DAN with differentiable scratch nodes."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_m... | {
"id": "31f7a61d-28d7-419e-baee-c7c85119e876",
"created_at": "2026-08-05 18:03:14.485916+00:00",
"db_md5": "5cc8b689dab16bcac0fcfa1643cae149",
"submitter": "Blake Camp",
"github_login": "blake-camp-surge",
"run_id": "f21c43c4-d820-42d8-a18c-88c4258fd890",
"tier": "easy",
"dataset_id": "e3",
"status":... | {
"score": {
"mean_loss": 2.1537427237194695,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011875000023283065
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1535414307369223,
"example_count": 80... |
31fe0c8d-3ad9-46e0-a912-ca06649d3702 | easy | Bananafly | 2026-08-12 09:24:22.421451+00:00 | succeeded | 3bff47f433ff8607a76979ff72b724ac8d397b699be1bdd408ae0bb70c208ac9 | 11,651 | null | """T-aligned soft residue model with the Easy T<=3 training cap."""
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_MO... | {
"id": "31fe0c8d-3ad9-46e0-a912-ca06649d3702",
"created_at": "2026-08-12 09:24:22.421451+00:00",
"db_md5": "ce562269be07d047c71d0e6de2ccd6ac",
"submitter": "Andre Kreidemann",
"github_login": "Bananafly",
"run_id": "b193d968-0c7b-4ca2-bc15-71e85546bb2e",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.6183664798736572,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6144089698791504,
"example_count": 100,
"exa... |
3200f631-05f9-4616-ab9f-8ad2988e0942 | easy | erdavis0 | 2026-08-18 03:58:46.125060+00:00 | succeeded | af9d2cbf680b1c6428593ea121a7d9b549fe45e5180c1c7ceaa1b6f5dd7020b2 | 34,092 | null | """Canonical token-register transition learner, version 12.
The model deliberately learns token semantics and field roles. Its recurrent
cell receives only a canonical categorical state and static context, while the
wall-clock curriculum controls how many copies of that same cell are composed.
Version 12 retains the... | {
"id": "3200f631-05f9-4616-ab9f-8ad2988e0942",
"created_at": "2026-08-18 03:58:46.125060+00:00",
"db_md5": "996305230845ceb7ab6d0495f133d4ef",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "d1ae6048-f2b1-4bc1-9d80-906b08349fb0",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.250003481087308,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2548313640692843,
"example_count": 600... |
32078b0a-1aa2-4d31-a5cb-26c8d1a8bf77 | easy | DESU-CLUB | 2026-08-30 22:09:32.384557+00:00 | succeeded | bf3e7bf4b7fca15820ae57cf5004ec055c0ff9307c9de8730675bcc02459080e | 17,051 | null | """Discrete-state loop over a place-value tape, trained with orthogonalize-only Muon.
The recurrent state carried between squarings is the *digit string*, not a
residual stream. Each outer iteration reads (N, x) off an LSB-first place-value
tape, emits digit logits for one squaring, hard-argmaxes them (straight-throu... | {
"id": "32078b0a-1aa2-4d31-a5cb-26c8d1a8bf77",
"created_at": "2026-08-30 22:09:32.384557+00:00",
"db_md5": "d5b91367ab54d9c6314ec0cf83a2f5b4",
"submitter": "Low Keng Hoong, Warren",
"github_login": "DESU-CLUB",
"run_id": "7fe4fcf8-7cbf-4e3d-89ef-b490e88cf08b",
"tier": "easy",
"dataset_id": "e6",
"sta... | {
"score": {
"mean_loss": 1.1329495927242383,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.7227477640719027
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.6142023801803589,
"example_count": 60,
... |
32081771-5204-4fdd-9170-0aa39839a5a2 | easy | andreamiele | 2026-08-06 09:01:45.101419+00:00 | succeeded | deef77a6961933b4dceed7fc02c84d30b0c917d6575ad430ce72378cb6c15419 | 22,805 | null | """Loop model with recall, composition consistency, and deep supervision.
Motivated by what the easy-to-hard / latent-reasoning literature converges on
and which earlier attempts here lacked:
* **Recall (input injection).** Deep Thinking (Schwarzschild et al. 2021;
Bansal et al. 2022) found iterating past the trai... | {
"id": "32081771-5204-4fdd-9170-0aa39839a5a2",
"created_at": "2026-08-06 09:01:45.101419+00:00",
"db_md5": "fed9ae064d2e37d2768c8fa51e77931c",
"submitter": "Andrea Miele",
"github_login": "andreamiele",
"run_id": "23fe40f6-a52e-48d8-9506-8f2a4e45380e",
"tier": "easy",
"dataset_id": "e1",
"status": "s... | {
"score": {
"mean_loss": 1.6511309148769606,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.7249999996026357
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.3287631844851508,
"example_count": 100,... |
320a8654-7887-40cf-912b-361f75a966a4 | easy | sapient-sapiens | 2026-08-26 11:08:46.233699+00:00 | failed | 7a1e2c5bd8f30916646bebf2aac6ab296f682a5acb81ea5518c58e4bbe0e3ac8 | 24,728 | null | """Standalone nested inner/outer recurrence candidate.
Representation is categorical only: ``E_token + E_group + E_leftrel``. No
field is decoded into a number, no field value sets depth, halting, or any
mixing weight, and the only supervised target is the final answer.
Structure
---------
One prompt encoder produce... | {
"id": "320a8654-7887-40cf-912b-361f75a966a4",
"created_at": "2026-08-26 11:08:46.233699+00:00",
"db_md5": "12920208fac6929b3d03b9428d490d51",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "e14b9a82-233b-4828-8420-79c66b6d6e92",
"tier": "easy",
"dataset_id": "e5",
"status": "fa... | null
|
320b412a-4d07-408a-b36c-3bee4451c088 | easy | shreyash-chonkie | 2026-08-26 11:00:52.948180+00:00 | succeeded | b919330e032745f18cd8e641f2c185a8151e68fed7c2f8ccf426560cb2f24eb8 | 16,034 | null | """Soft compiler with a factorized codebook around one shared executor."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
a... | {
"id": "320b412a-4d07-408a-b36c-3bee4451c088",
"created_at": "2026-08-26 11:00:52.948180+00:00",
"db_md5": "d7a65763756adac88c9a8ab9596a9567",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "06f82d35-a308-4908-9375-29697e28909c",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 2.014362156391144,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03833333496004343
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9620031118392944,
"example_count": 100,... |
320d839c-4da9-4b21-878b-ef10cc63f038 | easy | nikolageorgiev2000 | 2026-08-04 10:59:34.781674+00:00 | succeeded | 4fbc32e34da472bb8b82eab6477d6e20777ad18b6f61c1ed846753e5e96896dd | 8,615 | null | """T-composed categorical-digit reasoner with fully learned pair/reduction maps."""
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,
asser... | {
"id": "320d839c-4da9-4b21-878b-ef10cc63f038",
"created_at": "2026-08-04 10:59:34.781674+00:00",
"db_md5": "f5f08bd2e1da3b5b1fc6996586bf650c",
"submitter": "Nikola Georgiev",
"github_login": "nikolageorgiev2000",
"run_id": "b723ef6e-7ae4-484f-8081-6413b56dd337",
"tier": "easy",
"dataset_id": "e1",
"s... | {
"score": {
"mean_loss": 3.7405648231506348,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3633333295583725
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.036133289337158,
"example_count": 100,
... |
32164e30-74d0-43e8-9964-a2d4b53dc643 | easy | ehonig | 2026-08-20 04:35:48.198351+00:00 | succeeded | 6182105adbdca7263a66437859c11939d87dd16e7a91abdba411a6bbe45bebb6 | 22,766 | 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": "32164e30-74d0-43e8-9964-a2d4b53dc643",
"created_at": "2026-08-20 04:35:48.198351+00:00",
"db_md5": "5bd80ae2546989f4a991f582a987e805",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "d7fdd8b9-9315-48b2-a815-8e87f207d57c",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 1.900057315826416,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666716337204
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.867453932762146,
"example_count": 100,... |
3216ec3e-e4b6-4c8d-93dc-bd0cb6bf2fc8 | easy | newjordan | 2026-08-09 15:52:08.407179+00:00 | succeeded | 0abb18957e898cc274f9e7fd06a4dd88cb9eac69d3a408e1bfb03d0fa365d271 | 93,405 | null | """Helix-Crawler — a recursive-weight-shared "crawler" ALU for modular squaring.
The task is r_{i+1} = r_i^2 mod n, chained T times: a *recurrence* over a single
operator. This build embodies the crawler / Frugendorff lineage (spark:~/sota_crawler)
mapped onto that structure. n never enters as a table row (unseen modu... | {
"id": "3216ec3e-e4b6-4c8d-93dc-bd0cb6bf2fc8",
"created_at": "2026-08-09 15:52:08.407179+00:00",
"db_md5": "7fc5b712a1a89f3a0ad2e8918daeb855",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "56ede4bb-9e62-4f9d-88fd-23e31b076da3",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 4.846651359142198e-06,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.845738988686843e-06,
"example_count": 600,
... |
321b4929-da13-4888-94e2-57ed942998ef | easy | DDanlov | 2026-08-14 20:18:06.422634+00:00 | succeeded | ed3b2d23cfb1581e9dc4bee76ac8cace96e1db2fd2fbafe1132e7dafda93662a | 21,983 | 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": "321b4929-da13-4888-94e2-57ed942998ef",
"created_at": "2026-08-14 20:18:06.422634+00:00",
"db_md5": "18f5a2badb0aa94b30c280ca7f8adef4",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "1f693005-b980-4ee3-ba4d-b19d8ef6426e",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.366318163003226,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.368761746872701,
"example_count": 600,
... |
32244071-16e8-4df2-b969-9a656102dc73 | easy | DDanlov | 2026-08-20 22:41:54.915568+00:00 | failed | 4c203ab753b1e7844f3a4fab7ad103c37fedd97977bf3cfcba105a1515e05eb9 | 21,249 | null | from __future__ import annotations
import math
import time
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
except Imp... | {
"id": "32244071-16e8-4df2-b969-9a656102dc73",
"created_at": "2026-08-20 22:41:54.915568+00:00",
"db_md5": "d29246c49751f7ea3be017723c4c214d",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "3bd7f28c-5f5e-478f-bf33-732982ba0d3a",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
3225fbf7-b4af-4f15-97bf-d3b5e600f910 | easy | kushasareen | 2026-08-10 21:27:04.252931+00:00 | succeeded | a274410e75a52bf45ea521650451d29309753d0704da3763a0c1aa40ca8ba369 | 5,876 | null | """T-conditioned recurrent Transformer.
A tied-weight block is applied once per requested squaring step, so compute
scales with T and the model can extrapolate to depths never seen in training.
Untied pre-blocks encode the prompt and an untied post-block decodes the
answer digits.
"""
from __future__ import annotatio... | {
"id": "3225fbf7-b4af-4f15-97bf-d3b5e600f910",
"created_at": "2026-08-10 21:27:04.252931+00:00",
"db_md5": "4811307f3468ebc35c6ff8641aace1a2",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "68bbf481-0b90-411d-8e89-031a667e5d1a",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.211232511332465,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.5279168088339903,
"example_count": 600,
"ex... |
32274468-c66d-498f-8e1c-1f1b1ad3effa | easy | epicosy | 2026-08-20 13:22:58.901049+00:00 | succeeded | 271f11a54caa1b23d6ec3ed7999d93aa33bd02b537c0a34abea07c3fd104c349 | 4,460 | null | """Shared-refinement Transformer baseline for the Easy 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,
)
D_MODEL = 128
NUM_HEAD... | {
"id": "32274468-c66d-498f-8e1c-1f1b1ad3effa",
"created_at": "2026-08-20 13:22:58.901049+00:00",
"db_md5": "81bdf8de54eee0924b64d79557021ff0",
"submitter": "Eduard Pinconschi",
"github_login": "epicosy",
"run_id": "e7492c01-3382-47b6-beab-9ed67fc3c2ad",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 2.0440818071365356,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333334140479565
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.119250774383545,
"example_count": 10... |
32301730-5e93-4fdc-b7af-92c7b95bfbd2 | easy | sapient-sapiens | 2026-08-07 08:10:03.781081+00:00 | succeeded | 6689b33057c4522b912c24c8302d89f2fa903d1ed61ad52548b8f70d60d79215 | 12,464 | null | """Digit-space recurrent operator for repeated modular squaring.
Nothing here converts a token into a decimal number. Digits stay as embeddings,
so what a digit is worth, and every carry and reduction, is learned. That also
sidesteps a hard numeric wall: at Hard's modulus scale `x^2` runs past the 53-bit
exact-integer... | {
"id": "32301730-5e93-4fdc-b7af-92c7b95bfbd2",
"created_at": "2026-08-07 08:10:03.781081+00:00",
"db_md5": "a30cbeab0811362a9107bf42db0134aa",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "e642cc6e-2e5b-49f0-8dfe-2c9d593c32f7",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 4.372228920905668,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.029999999875823656
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.6747283935546875,
"example_count": 100... |
323095af-ebfa-47e0-856a-8bfb9430868e | easy | evanatyourservice | 2026-08-04 14:26:54.165643+00:00 | succeeded | 8b94cfebd9722acaaf911bbc1bd9d36677782b4e6bafd5f158eb3890a0011afa | 72,261 | null | from dataclasses import dataclass, field
K3: tuple[int, ...] = (3, 4, 5, 7, 8, 9, 11, 13, 16, 17, 19, 23, 25, 27, 29, 31)
@dataclass
class K13:
vocab_size: int = 17
max_seq_len: int = 64
d_model: int = 256
n_heads: int = 4
ffn_mult: float = 3.0
cell: str = 'mixer'
dropout: float = 0.0
e... | {
"id": "323095af-ebfa-47e0-856a-8bfb9430868e",
"created_at": "2026-08-04 14:26:54.165643+00:00",
"db_md5": "811146e7000a4b615c0dd2831942ccf1",
"submitter": "Evan Walters",
"github_login": "evanatyourservice",
"run_id": "1a5b26b6-4b5e-4995-9dac-1486bb03e6fd",
"tier": "easy",
"dataset_id": "e2",
"statu... | {
"score": {
"mean_loss": 2.3687503337860107,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333972841501
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4031691551208496,
"example_count": 30... |
3231f32e-3182-41ab-800b-a3b85dfed26e | easy | viridale | 2026-08-06 07:56:52.455076+00:00 | succeeded | 913ed7084aeb9de53b63282cf47a23c2c930d0e5f3f8ca5d631e8f0aedb0ad71 | 7,944 | null | """Clean global learned-weight sphere on the exact v177 route.
hypothesis: the universal train-fit/held-out gap is a norm-mediated grokking delay;
constraining one global random-weight radius from initialization may prevent the
high-norm memorizing phase while preserving the calibrated learned recurrence.
axis... | {
"id": "3231f32e-3182-41ab-800b-a3b85dfed26e",
"created_at": "2026-08-06 07:56:52.455076+00:00",
"db_md5": "95710a87757b741254ded72747e28dce",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "d585361f-0737-419e-993d-1fb151e42acb",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 2.1818689107894897,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.195966958999634,
"example_count": 100,
... |
3232fa7b-4094-48d6-b056-4aa704012d73 | easy | ehonig | 2026-08-14 01:15:42.997609+00:00 | succeeded | bed7b46c989217f150e8f32594973831db97cd435e186415766922012faf890c | 48,079 | 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": "3232fa7b-4094-48d6-b056-4aa704012d73",
"created_at": "2026-08-14 01:15:42.997609+00:00",
"db_md5": "8a71b19028930edd1dd4facf7e650670",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "44287a70-8ae0-40a5-a595-3da9e8cddcd4",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 6.010462999343872,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.17999999970197678
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.578084945678711,
"example_count": 100,
... |
323765d3-65e2-48ad-b7e7-aa119f59379c | easy | amirmeisami | 2026-08-17 23:25:26.689067+00:00 | succeeded | 00a60383558f98cd12d4f2d57fdf3843ac24931fc5ad0ebd263aa1c2c62a16cd | 13,092 | null | """T023 P3: D with compositional N-conditioned FiLM."""
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
D_MODEL, POLY_WIDTH, INNER_ST... | {
"id": "323765d3-65e2-48ad-b7e7-aa119f59379c",
"created_at": "2026-08-17 23:25:26.689067+00:00",
"db_md5": "6e419406fb2757f40eb4e8f7516b73ea",
"submitter": "Amir Meisami",
"github_login": "amirmeisami",
"run_id": "041bb80c-38ba-4ba4-8c88-5b8d38a1ac54",
"tier": "easy",
"dataset_id": "e3",
"status": "s... | {
"score": {
"mean_loss": 5.156715846632,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.271208133023478,
"example_count": 800,
"exact_a... |
3239a729-d811-459e-8d94-e049a7489f5b | easy | liam-gb | 2026-08-13 13:07:11.381112+00:00 | succeeded | c98994ef0a17cbbc563e971c85fd060c426e82414551fdb5ce599d6575496cbc | 15,106 | null | """rns_pm_gather_ct_widet_pt: widet + per-prime temperature (16-prime bank).
The wide readout is what makes long answers reachable; the trim is what pays
for it, by not running loop iterations whose update is multiplied by zero.
Base docstring:
Single change vs rns_pm_gather_ct: the readout carries one
distance-from... | {
"id": "3239a729-d811-459e-8d94-e049a7489f5b",
"created_at": "2026-08-13 13:07:11.381112+00:00",
"db_md5": "f909714a31387e09614b38df91ccc3f4",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "b4c313cb-1dcf-4ce1-8d4e-8381759aecf5",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 5.185023605823517,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.486666664481163
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.129050254821777,
"example_count": 100,
... |
3248ae8e-56a8-4e40-8a0d-254bef56be04 | easy | sapient-sapiens | 2026-08-30 22:41:23.176389+00:00 | succeeded | 6b3c32a4adf79726f32ea6df75c35b1dbf90f5a0d5a0d2bdc9ce8bc1af2d1f0b | 19,450 | null | """Eligible E5 direct-Horner model with detached same-final suffix credit.
The ordinary eight-square branch retains full BPTT and weight 0.9. A T-aware
selector chooses a detached predecessor from X/A1--A7, and one fresh tied
square receives weight 0.1 against the same final answer labels. No
intermediate value, depth... | {
"id": "3248ae8e-56a8-4e40-8a0d-254bef56be04",
"created_at": "2026-08-30 22:41:23.176389+00:00",
"db_md5": "47cc921b7a5378961bb31000aa7fe63f",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "4024c028-b961-40fe-a2f9-52fc2b86e34f",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.1822677019558974,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333345750967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1809668327126266,
"example_count": 60... |
32490466-2deb-425d-9fdb-783bc43fab94 | easy | karanganesan | 2026-08-06 04:06:03.585335+00:00 | succeeded | dde1f5143796bb3b7f1bd5be69d4182743656898eda5fa57401966ae4f34adda | 23,811 | 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": "32490466-2deb-425d-9fdb-783bc43fab94",
"created_at": "2026-08-06 04:06:03.585335+00:00",
"db_md5": "c98723877a9d412dcf2858d4393b1fc5",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "cde34490-736d-4aff-8d02-ea0bd2f3bfce",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 9.129137516021729,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.041666666977107525
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.51321792602539,
"example_count": 100,
... |
32498641-41df-497f-a05e-da08f856fd9e | easy | sapient-sapiens | 2026-08-14 17:02:54.835569+00:00 | succeeded | 28ba469aae3e42ca679270105f219b6bd47433fa80f83dc154b3c3e55e44c7c3 | 14,299 | null | """Digit-space recurrent operator for repeated modular squaring.
Nothing here converts a token into a decimal number. Digits stay as embeddings,
so what a digit is worth, and every carry and reduction, is learned. That also
sidesteps a hard numeric wall: at Hard's modulus scale `x^2` runs past the 53-bit
exact-integer... | {
"id": "32498641-41df-497f-a05e-da08f856fd9e",
"created_at": "2026-08-14 17:02:54.835569+00:00",
"db_md5": "1efadda9e01ae84ec2857432a666c542",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "2aa7e971-2b80-45c9-86cf-f73e315da157",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 3.6421984211167793,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.16999999893208345
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.194008827209473,
"example_count": 100,... |
325215cc-9d46-4e89-bd4e-0cdbb0ed4bfe | easy | karanganesan | 2026-08-04 02:04:35.243687+00:00 | succeeded | 2789252f6dea1aab9eafd26edd1457b106239034b193bdab563b9da52f625a7e | 11,979 | 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": "325215cc-9d46-4e89-bd4e-0cdbb0ed4bfe",
"created_at": "2026-08-04 02:04:35.243687+00:00",
"db_md5": "a7d24de876f0bd4e84232cb8027c69bf",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "bdb86765-7280-4555-a59b-dcc2a536ad31",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.0600165724754333,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03333333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.962497591972351,
"example_count": 100,... |
32548471-8ebe-41f5-a7d5-321623358bfe | easy | mhlov000111 | 2026-08-14 17:59:00.833502+00:00 | failed | 757a0256b2c21619c98df86b80eaf2c2a9aeba1491bc83120aec11cd1648e4ee | 3,701 | 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": "32548471-8ebe-41f5-a7d5-321623358bfe",
"created_at": "2026-08-14 17:59:00.833502+00:00",
"db_md5": "a48d99b8dad27072b6a00414651e99a4",
"submitter": "mhlov000111",
"github_login": "mhlov000111",
"run_id": "3403b9fe-976c-421e-9ef1-b5d0c5ce844f",
"tier": "easy",
"dataset_id": "e1",
"status": "fa... | null
|
3254fc8d-5103-4241-a3c0-eadef1b98558 | easy | nikhilkumar26 | 2026-08-29 18:05:01.530890+00:00 | succeeded | 06a8315650eb84764c1b082273e20e1ca4d14352c5fc721c17bc0ef091382461 | 6,017 | null | """
One Layer Deeper Easy E1:
direct T-conditioned MLP from packed x/N digits to answer digits.
No identity residual (that stuck on copying x). No attention.
T is a feature, not a loop count, so T=1 can be its own map.
No modular arithmetic in Python.
"""
from __future__ import annotations
import torch
import torch.... | {
"id": "3254fc8d-5103-4241-a3c0-eadef1b98558",
"created_at": "2026-08-29 18:05:01.530890+00:00",
"db_md5": "4819125a321431ad1cbfd9b4c956f399",
"submitter": "Nikhil Kumar",
"github_login": "nikhilkumar26",
"run_id": "9a3ef849-e0af-4102-84cb-2c54f93f72a9",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 17.220815181732178,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999828636646
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 19.979307174682617,
"example_count": 100... |
3255ada2-a58e-4e32-90ae-d32f7cbc32fc | easy | isaac0804 | 2026-08-06 15:14:30.252158+00:00 | succeeded | dfdf066f992f0dc6d8ae202a6e27d0fc07eff063af5141d9b5c756b808fc4ddd | 8,076 | null | """E3 candidate #4: isolate the six 10-bit moduli (half of E3's real held-out
set, ~4x more training data each) for the full training budget.
submissions/e3_single_n_isolation showed one modulus trained alone reaches
~7-9% raw held-out accuracy, but with only ~150 reachable rows out of 1600
that's diluted to a ~0.7-0.... | {
"id": "3255ada2-a58e-4e32-90ae-d32f7cbc32fc",
"created_at": "2026-08-06 15:14:30.252158+00:00",
"db_md5": "5a8b85c0dfd2e66e565f20f0ec1dfd11",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "e633b329-e16b-427a-9f54-b7dc93ccf599",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 13.710800919002128,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010625000004656614
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 14.08648749083895,
"example_count": 800... |
325a068d-368e-492b-bcde-3a26da7bd65b | easy | khushidahi | 2026-08-12 21:15:06.235105+00:00 | succeeded | d75ebe1b1033c6c4348e14b952237f6215909fb8b4cc213de52ee63c772b57bb | 16,711 | 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": "325a068d-368e-492b-bcde-3a26da7bd65b",
"created_at": "2026-08-12 21:15:06.235105+00:00",
"db_md5": "8294b9d5386634e1f3e6103a29bde8d9",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "b860b01a-ee1a-45ab-a74d-67e0bf388118",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.3077492713928223,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01000000024214387
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2933859825134277,
"example_count": 600... |
32629e2b-511f-4ec4-9e5f-3d862ecf3efd | easy | ehonig | 2026-08-13 19:58:25.233082+00:00 | succeeded | 0d74b0a6c1350b833c8c6a976d07c8519a44d39a3f47506337aa7456f8ecba18 | 38,708 | 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": "32629e2b-511f-4ec4-9e5f-3d862ecf3efd",
"created_at": "2026-08-13 19:58:25.233082+00:00",
"db_md5": "8b7f3852889c325824c5c4db0f13bbf7",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "b9638951-bda0-4926-aa5d-05358a417398",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 4.2439093589782715,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999828636646
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.1934967041015625,
"example_count": 100... |
3264077d-cbc5-4687-8a51-dbeb607c9a16 | easy | khushidahi | 2026-08-21 20:38:09.793368+00:00 | succeeded | cadc3206cff31b085c166e9367d2f313a7e39b76a93688b16228e12a0dad6b46 | 26,236 | null | """R358: learned pair-product workspace plus an ordered recurrent reducer.
This candidate is deliberately diagnostic. It trains on evaluator-owned final
labels at several depths so that one shared transition must compose. The model
never converts N or X to native numbers and never implements multiplication or
modula... | {
"id": "3264077d-cbc5-4687-8a51-dbeb607c9a16",
"created_at": "2026-08-21 20:38:09.793368+00:00",
"db_md5": "5a87b031452eb71be1bf332bc1678936",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "3fe485b7-11b0-4c13-885e-890be09b9c18",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 3.06973936863941,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04958333334575097
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.1679189707666233,
"example_count": 600,
... |
3271cc42-e138-4e73-8d6c-af83824f2da1 | easy | khushidahi | 2026-08-22 01:36:12.922011+00:00 | failed | fda551d74b63f3a0cd85ea909c5b2c9f00593a40a0d212a8308ae654fc7ff1d0 | 19,190 | null | """R395: minimal Universal-Transformer functional-root control for M5.
This candidate adapts the released Universal Transformer to the competition's
separate-output interface. Numeric prompt symbols are only rearranged onto
equal-length, right-aligned categorical N and X tapes. A third tape supplies
parallel learned... | {
"id": "3271cc42-e138-4e73-8d6c-af83824f2da1",
"created_at": "2026-08-22 01:36:12.922011+00:00",
"db_md5": "04b52418182fef5f933830f7207aa6e5",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "375dc596-55c3-485a-89d7-8fdb659f83da",
"tier": "easy",
"dataset_id": "e5",
"status": "fail... | null
|
3274a62f-0da2-4396-b00c-bd5805783bb4 | easy | sapient-sapiens | 2026-08-16 19:46:44.504324+00:00 | succeeded | 64bf8862ecd9345b1be87ecccf90025f8e8e9fe2a3fb75d5af62befd9734861c | 21,856 | null | """E6 SAM-family variant: frozen v55 with a one-variable SAM change.
Architecture, representation, HybridMuon split, rho=0.02, batch, wall-clock
schedule, and loss are unchanged from act64_e6_w1024_dh4_L6_v55. T is
diagnostic-only.
"""
from __future__ import annotations
import math
import time
import torch
import t... | {
"id": "3274a62f-0da2-4396-b00c-bd5805783bb4",
"created_at": "2026-08-16 19:46:44.504324+00:00",
"db_md5": "9c3fe8c09794630ba52311f2af7a9df6",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "ba9146a8-545c-438f-89d9-e185d9c19dd8",
"tier": "easy",
"dataset_id": "e6",
"status": "su... | {
"score": {
"mean_loss": 6.267295125412614,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.1270270360180655
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.668114185333252,
"example_count": 60,
... |
327d53d5-25e9-4378-a48c-9568bbd89c06 | easy | oupadhyay | 2026-08-29 17:39:40.769695+00:00 | succeeded | 5259c3ed82c71e7299d1fcb18f73a41e77662cea15459a10106aac2de747f1d4 | 3,991 | null | """Conditional optimizer screen: warm64."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,assert_model_state
W,D,H=4,64,32
class C:
def __init__(s,vocab_size,max_seq_len):s.vocab_size... | {
"id": "327d53d5-25e9-4378-a48c-9568bbd89c06",
"created_at": "2026-08-29 17:39:40.769695+00:00",
"db_md5": "e6869187b94adc2273e8f72a54b761db",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "4578b452-d0c9-4f31-b2ce-9dd6f3035ecb",
"tier": "easy",
"dataset_id": "e6",
"status": "s... | {
"score": {
"mean_loss": 1.892147183418274,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.036936938762664795
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8545944690704346,
"example_count": 60,... |
3284f82f-03ef-4e25-beab-d0707d611213 | easy | alirezashirvani-jr | 2026-08-23 19:41:03.204292+00:00 | failed | f687f8b5b6d35742c90ebc74cd65a2dec1f2ec54fcfc10710a84fa5c7ec43290 | 38,404 | null | """Looped digit machine whose transition is assembled by learning.
The recurrent state is the decimal digit string of the current value; one
loop application applies one learned transition chosen by a gate over an
enumerated polynomial family (all mod N): pure powers x^k for k in 1..9,
x^2+c and x^3+c for c in 1..9, t... | {
"id": "3284f82f-03ef-4e25-beab-d0707d611213",
"created_at": "2026-08-23 19:41:03.204292+00:00",
"db_md5": "14c8a7ce951b95dea1fe85f86155e74f",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "80dbecf9-7fe0-4c53-90ce-bf0d940aad83",
"tier": "easy",
"dataset_id": "e1",
... | null
|
328ab284-58a7-40de-85ed-6900cbd134b6 | easy | erdavis0 | 2026-08-26 03:10:47.183958+00:00 | succeeded | b063dd87e5c8ed86565b7dc9bb9038c969fb662780579ead5b62238da52e334a | 30,346 | null | """Latent-role continuous REV control C0: ordinary evaluator token CE.
Public delimiter syntax only routes opaque token categories into distinct,
right-aligned N, X, and T tapes. Learned local cells compile separate
interaction, carry, and modulus-conditioned workspaces. N is isolated from
interaction and carry and ... | {
"id": "328ab284-58a7-40de-85ed-6900cbd134b6",
"created_at": "2026-08-26 03:10:47.183958+00:00",
"db_md5": "b092954054ec5242242d7829dc8330c1",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "c25a6b03-ca33-4aa3-a9e2-6c9dc761eee9",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",... | {
"score": {
"mean_loss": 3.4800833301911487,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016666666753590107
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.4634572711226115,
"example_count": 60... |
3292086e-1ce5-4be8-877c-536f039dd349 | easy | chuk-uzowihe | 2026-08-09 19:21:22.202946+00:00 | succeeded | a76e9e85eac62a02d873ee353e6059b2eccaf2203a716e2542a1d16223f86312 | 36,586 | null | """One Layer Deeper submission: depth-routed MoEUT over a 2D digit lattice.
v9: the recurrent block is a single MoEUT layer (refs/moeut) — SwitchHead
attention + sigma-MoE feedforward — whose routers are conditioned on the
loop depth (Fourier features of the iteration index) so one shared bank can
express phase-depend... | {
"id": "3292086e-1ce5-4be8-877c-536f039dd349",
"created_at": "2026-08-09 19:21:22.202946+00:00",
"db_md5": "25707291aae08f565bb2ff223de6b58e",
"submitter": "Chuk Uzowihe",
"github_login": "chuk-uzowihe",
"run_id": "23fd3a2b-9577-42a2-911a-c34d0d0f6842",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 6.6963605880737305,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004999999888241291
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.059166431427002,
"example_count": 100... |
329606f2-c142-4d14-bc36-a22d475ed99d | easy | alirezashirvani-jr | 2026-08-30 01:23:35.683712+00:00 | succeeded | 8a2a4c8c8062f383d5f21e7b140c34861f6973afac9e18322228479f7a7a6959 | 30,005 | null | from __future__ import annotations
import math
import time
from contextlib import nullcontext
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
MODEL_... | {
"id": "329606f2-c142-4d14-bc36-a22d475ed99d",
"created_at": "2026-08-30 01:23:35.683712+00:00",
"db_md5": "643a70415f53a03241cccb340197a6a8",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "fb9c5528-23c8-48d2-b7ba-88694d940305",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 2.766132621950394,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.019166666883975267
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.786709785461426,
"example_count": 600,... |
329c5045-ca73-4710-8040-49b2d96be368 | easy | christopherhwood | 2026-08-05 13:33:28.608373+00:00 | succeeded | 2617dc6b6f2e0909eeff898cbf32fdb2140f02ebe4b9e63c0bd7309efd7f07cc | 9,724 | null | """Recurrent modular-squaring model for One Layer Deeper.
Mathematical premise
--------------------
Every target is result = x^(2^T) mod N with the one-step recurrence
result(T+1) = result(T)^2 mod N , result(0) = x .
So the whole certified-depth ladder T = 1,2,4,8,16,32,64 is generated by
iterating a... | {
"id": "329c5045-ca73-4710-8040-49b2d96be368",
"created_at": "2026-08-05 13:33:28.608373+00:00",
"db_md5": "e006781bf75d19051f0d3605aba851fb",
"submitter": "Chris Wood",
"github_login": "christopherhwood",
"run_id": "364ef4e6-1ce2-4188-a0e9-2daf5b7d16bd",
"tier": "easy",
"dataset_id": "e1",
"status":... | {
"score": {
"mean_loss": 3.2159193754196167,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0716666653752327
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.853339910507202,
"example_count": 100,
... |
329f165a-be70-46e6-9e5f-dc880773a90c | easy | khushidahi | 2026-08-02 06:04:22.720951+00:00 | succeeded | 07e0401d081cc53919b68a3ab3a4efcb750919cc23da183c3cce92d1ab84f7b7 | 8,805 | null | """R4 exactness-aligned-loss Easy E5 probe.
The model, optimizer, stochastic recurrent-depth curriculum, evaluation depth,
batches, and 200-step checkpoint match the R3 Easy control. The only change is
a boundary-preserving TokenLossBatch objective that blends per-example mean
cross-entropy with a smooth maximum over ... | {
"id": "329f165a-be70-46e6-9e5f-dc880773a90c",
"created_at": "2026-08-02 06:04:22.720951+00:00",
"db_md5": "b3ab49f13b57942de9881b1225412b37",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "1885588e-ba05-4316-aa5a-5704e7580691",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.157239079475403,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006249999976716936
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1617157459259033,
"example_count": 600... |
32a3faad-1625-48f7-af36-812601c3972a | easy | sapient-sapiens | 2026-08-19 19:29:15.223647+00:00 | succeeded | daa4fc1f1821b95b049e62d1602f10121592d66b9efbbb3e78e141bca4403b2b | 33,252 | null | """Depth-quantized recurrence with a mixture-over-depth objective.
The recurrent state is a bank of right-aligned digit slots. Every operator
application is followed by a soft quantization back onto the token simplex, and
the digit logits produced by that quantization are the answer logits at that
depth. There is no s... | {
"id": "32a3faad-1625-48f7-af36-812601c3972a",
"created_at": "2026-08-19 19:29:15.223647+00:00",
"db_md5": "2553f964fd1803486430c2690edf5c49",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "c70989f4-9fe2-47aa-a55b-05d0832f69aa",
"tier": "easy",
"dataset_id": "e6",
"status": "su... | {
"score": {
"mean_loss": 4.762935870314298,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07702703154771715
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.433041572570801,
"example_count": 60,
... |
32a923bd-9598-4e5c-885c-b25fed9cb91d | easy | TZZheng | 2026-08-07 05:23:45.386075+00:00 | succeeded | e59ec39ffa3b704b0493cfb69cb8f2f581cc3cdf99e395e73b569692aea97814 | 12,532 | null | """Learned recurrent modular-transition model.
The mutable digit state is initialized from x. A learned square-oriented
self-mixing network and an N-conditioned reduction network form one transition,
and exactly the same transition parameters are reused T times. No arithmetic
operation is hard-coded; only the public p... | {
"id": "32a923bd-9598-4e5c-885c-b25fed9cb91d",
"created_at": "2026-08-07 05:23:45.386075+00:00",
"db_md5": "7b2fb2f77f32883e47780f3ecfb50f58",
"submitter": "TZZheng",
"github_login": "TZZheng",
"run_id": "37fe4752-2c0d-4dcc-94c9-6fce3f0bc252",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 6.501774311065674,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04166666604578495
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.616199016571045,
"example_count": 100,
... |
32ad566c-bd5b-4a4e-9718-07f58b1f89c8 | easy | sapient-sapiens | 2026-08-21 09:05:52.100404+00:00 | succeeded | 11c1f120fe938e26d0e3e680e53beb6e416f9d2fcea67fcfbaefbce9944559fe | 36,841 | null | """Hierarchical learned digit-mixing recurrence with an answer-only objective.
The recurrent state is a bank of right-aligned digit slots. Every operator
application is followed by a soft quantization back onto the token simplex, and
the digit logits produced by that quantization are the answer logits at that
depth. T... | {
"id": "32ad566c-bd5b-4a4e-9718-07f58b1f89c8",
"created_at": "2026-08-21 09:05:52.100404+00:00",
"db_md5": "ad6ff76a87ec0fee4fbe97989e662521",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "9f074453-baa0-49fb-8434-10155757076c",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.1419773012831906,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0025
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1432431207911318,
"example_count": 600,
"... |
32aec6a1-7455-47e8-8a02-ed094ab4d933 | easy | shirvani-jr | 2026-08-28 07:45:57.385410+00:00 | succeeded | e87740b88641d860c1f7fff578bd38e2c26a3cdf605ce7155b8001c72747066b | 39,517 | null | """Adaptive learned discrete recurrent fabric for One Layer Deeper.
A gradient-trained, operation-free recurrent computational machine. The
model learns, from the competition's endpoint supervision: how to parse the
prompt into a bank of categorical digit registers, a reusable transition
circuit applied serially, a l... | {
"id": "32aec6a1-7455-47e8-8a02-ed094ab4d933",
"created_at": "2026-08-28 07:45:57.385410+00:00",
"db_md5": "74f70b0c890ec2ef90620d41f0e8936f",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "64eb91a0-b428-4b74-8069-8157f96531dc",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.0071306824684143,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018333333544433117
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.992536187171936,
"example_count": 100... |
32b2a27e-b88e-4ee2-8c57-a3632957d9a4 | easy | jayyala | 2026-08-16 14:51:02.421193+00:00 | succeeded | cc2907731b12e4ec1443b0cab577d4bf983e7e8f537dc04881fa6c720f2decca | 13,881 | null | """A recurrent machine over structurally encoded byte slots.
Integers use little-endian byte slots for compactness, but a byte is never an
atomic learned symbol: both encoding and decoding factor through its eight
bits. One tied transition cell is applied exactly T times. Larger integers
add slots and recurrent comp... | {
"id": "32b2a27e-b88e-4ee2-8c57-a3632957d9a4",
"created_at": "2026-08-16 14:51:02.421193+00:00",
"db_md5": "d494ebb3966b31f3f10c8605e982bada",
"submitter": "jayyala",
"github_login": "jayyala",
"run_id": "04bf8dec-8e35-4f53-9fe2-77d9371803bf",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 36.432124077199816,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333457509677
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 36.63677215576172,
"example_count": 60... |
32b6a664-add4-4160-8146-45ffb67e14d3 | easy | yashkant | 2026-08-14 16:31:18.204524+00:00 | succeeded | b16837e0c672688c0ed3c5da54128810aaf93fdd210f5212f585458e89ce4742 | 34,010 | null | """Round1795 parsed-T conditional mean-free Dykstra corrections.
Both arms use the strongest Round1792 AdamW5/EMA.995 W8/D2 carrier with
strength-2 contrastive set codes and eagerly compute the same raw and
mean-free correction branches. The candidate keeps Round1792's mean-free
corrections only for public T=2 and se... | {
"id": "32b6a664-add4-4160-8146-45ffb67e14d3",
"created_at": "2026-08-14 16:31:18.204524+00:00",
"db_md5": "ee5e787a5c02ccdb8d94b8fb05fe6f02",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "b89bcdda-d0d8-4152-8ffb-23e926b40970",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 3.6306509971618652,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.19729167222976685
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.1369829177856445,
"example_count": 300... |
32be355e-5a23-4855-ab2f-e5f570c03506 | easy | himalalps | 2026-08-08 04:40:19.806228+00:00 | succeeded | ebf51739a09f9e77d8cf943b5c76acc3d1c2ffb0b256672e6db2eec56ea26fed | 71,126 | 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": "32be355e-5a23-4855-ab2f-e5f570c03506",
"created_at": "2026-08-08 04:40:19.806228+00:00",
"db_md5": "97b06e52720df955cecc637a7723f4b7",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "86d18b36-474d-4cb8-a9ad-2361ccd8223b",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 2.742419707278411,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333233992258
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2519426345825195,
"example_count": 600... |
32c121b5-ed06-4ba8-a19f-2e8dfd8f9803 | easy | harmya | 2026-08-01 22:43:11.487767+00:00 | succeeded | 4a7d4bd7da6eb51ac6d978b0e06235858d6495f1e0b223088b6f4b26eaf2af30 | 29,979 | null | """Learned multiplicative characters on Z_N.
Each modulus seen in training gets K angle embeddings over its residues,
trained by a label-free consistency loss: angle(a) + angle(b) = angle(ab mod N).
Squaring is angle doubling; y = x^(2^T) mod N is T doublings, each snapped to
the nearest embedding. Converged tables ar... | {
"id": "32c121b5-ed06-4ba8-a19f-2e8dfd8f9803",
"created_at": "2026-08-01 22:43:11.487767+00:00",
"db_md5": "e1c45211fe3ff0574f67f28f0fe30a72",
"submitter": "Harmya Bhatt",
"github_login": "harmya",
"run_id": "3457e4d1-4c0b-450a-b872-f74f826391d7",
"tier": "easy",
"dataset_id": "e2",
"status": "succee... | {
"score": {
"mean_loss": 0.0,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0,
"example_count": 300,
"exact_accuracy": 1.0,
"... |
32c3186b-240d-406e-8add-a64adca059a7 | easy | achyutreddy24 | 2026-08-04 23:24:27.620411+00:00 | succeeded | db4796c97c70b95054cb4d48740fff6bbf32dc318cf7f0aef288d894760582a4 | 12,571 | null | """A looped transformer with learned, input-conditioned depth.
WHAT THIS IS
============
This submission contains no task-specific arithmetic: no modular structure, no
hand-set initialization, no token-map constants. It is a single generic
transformer block applied repeatedly to a latent state, with a learned halting... | {
"id": "32c3186b-240d-406e-8add-a64adca059a7",
"created_at": "2026-08-04 23:24:27.620411+00:00",
"db_md5": "ee5fd363e0ab989e41e0a043e087d68b",
"submitter": "Achyut Reddy",
"github_login": "achyutreddy24",
"run_id": "854b2f6e-26c1-4030-9717-d4486524a6dd",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.1305837631225586,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05000000074505806
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.194573163986206,
"example_count": 100,... |
32c8258d-c42c-4317-a778-d3d331977683 | easy | khushidahi | 2026-08-20 03:37:11.004557+00:00 | succeeded | 1015421bde4b81c63df7f5af90d1d1f37b7b3c931def1924432509bb9b4ab4e4 | 33,722 | null | """R299: dual-credit balanced causal categorical functional root.
This candidate is deliberately diagnostic. It trains on evaluator-owned final
labels at several depths so that one shared transition must compose. The model
never converts N or X to native numbers and never implements multiplication or
modular reducti... | {
"id": "32c8258d-c42c-4317-a778-d3d331977683",
"created_at": "2026-08-20 03:37:11.004557+00:00",
"db_md5": "ba5c8c183f8eeacc9cee02943fe08ba7",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "03f1e095-db79-4498-ad97-2f5f23293bb2",
"tier": "easy",
"dataset_id": "e3",
"status": "succ... | {
"score": {
"mean_loss": 2.3235498615781904,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006250000009313226
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3835446029136698,
"example_count": 80... |
32cb62dd-cf22-4ba9-903e-71df6c43b876 | easy | ehonig | 2026-08-19 13:37:51.120736+00:00 | succeeded | 49cac13e3b6dc56a2eb81acb605dd5571ece2e54245181f91d54ae9cab438b16 | 16,046 | 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": "32cb62dd-cf22-4ba9-903e-71df6c43b876",
"created_at": "2026-08-19 13:37:51.120736+00:00",
"db_md5": "1e9df395b20f81e491ca0d091728444c",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "5ba89541-8a46-4935-a799-7c71ae21a0ac",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 4.150514125823975,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.036666667088866234
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.979248046875,
"example_count": 100,
... |
32cc3b61-d801-4847-9bf4-0bad6b888ccd | easy | velocizapkar | 2026-08-11 21:26:04.846741+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": "32cc3b61-d801-4847-9bf4-0bad6b888ccd",
"created_at": "2026-08-11 21:26:04.846741+00:00",
"db_md5": "eaac69fe345f82973fcf392c504a40ff",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "5722c04f-ead4-4389-935b-c14f4e0e859c",
"tier": "easy",
"dataset_id": "e2",
"statu... | {
"score": {
"mean_loss": 2.7604352235794067,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00541666685603559
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.0651791095733643,
"example_count": 300... |
32cc741f-6f0a-4a73-bd1e-db0a054f2127 | easy | himalalps | 2026-08-07 02:48:00.492516+00:00 | succeeded | 8af717b7c876adf9bb888628d3a07362b9a96b68fb399684a1f0f65f5b3effa2 | 67,096 | 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": "32cc741f-6f0a-4a73-bd1e-db0a054f2127",
"created_at": "2026-08-07 02:48:00.492516+00:00",
"db_md5": "9883a5898a8b75406130229c3aa0087d",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "72cacdcf-2e3b-4c5c-b96d-1c52d58e3ac8",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 2.6935141298684986,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006250000040357312
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.7836976051330566,
"example_count": 60... |
32ce0e71-bbf3-41b7-917e-ec53436fd1af | easy | khushidahi | 2026-08-06 05:24:39.988551+00:00 | succeeded | 14560d3bd3c01f3f77573c831a8cdaf33344da418ab92298b6f39afcee8b93e2 | 14,801 | 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": "32ce0e71-bbf3-41b7-917e-ec53436fd1af",
"created_at": "2026-08-06 05:24:39.988551+00:00",
"db_md5": "5bab557d781de25bb9a7e3430739dddd",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "354ebcef-21fd-4591-8c96-d2fa2164b014",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 2.231923460960388,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010208333376795053
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.233626365661621,
"example_count": 300,... |
32cef475-78a3-4444-b6eb-458a5d0e3fc2 | easy | khushidahi | 2026-08-04 01:25:00.141531+00:00 | succeeded | 1d48f2c13ace5326a591d5150f1b73308a01381a9b42081e04faac85b052bda3 | 14,965 | 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": "32cef475-78a3-4444-b6eb-458a5d0e3fc2",
"created_at": "2026-08-04 01:25:00.141531+00:00",
"db_md5": "dbc48e5d83d927bdf19d39d90f6efb9f",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "a65708c3-d46e-45b0-afaf-5154c4ce7367",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.2096070051193237,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00750000006519258
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.20685076713562,
"example_count": 600,
... |
32d1f6f0-59c1-4a29-9e3e-18e5702c046f | easy | Bananafly | 2026-08-12 09:59:24.476406+00:00 | succeeded | 1346df934cbf63ae1a767e71889a9f8c8d470ef1103dc522b6a9ae7db9476732 | 13,593 | null | """Easy-3 residue model with learned decimal digit-pair memory."""
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_MOD... | {
"id": "32d1f6f0-59c1-4a29-9e3e-18e5702c046f",
"created_at": "2026-08-12 09:59:24.476406+00:00",
"db_md5": "bab4d5acbd9cbde59d24639d2c756ea6",
"submitter": "Andre Kreidemann",
"github_login": "Bananafly",
"run_id": "91bbbd5e-2904-4f69-9346-c8b3f0f2a646",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.1495964761051507,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.014166666977107525
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1515212059020996,
"example_count": 60... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.