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 |
|---|---|---|---|---|---|---|---|---|---|---|
117e38da-5809-4cdc-9309-93bbf4aee59d | easy | ehonig | 2026-08-21 01:03:25.735342+00:00 | succeeded | a9ad0d4126ab74869f8e7f4109841541ffaa4c10fe840580936c9852aab9df05 | 30,222 | 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": "117e38da-5809-4cdc-9309-93bbf4aee59d",
"created_at": "2026-08-21 01:03:25.735342+00:00",
"db_md5": "e6d7ec5bd1fddf2d609c6fc1ca4d18f7",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "a4f24c2c-ace4-4e30-a68f-76ae3c70d1ae",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 2.2247812747955322,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011041667079553008
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2255866527557373,
"example_count": 30... |
1182b32b-66b0-4372-bec5-4bc63e0751ea | easy | oupadhyay | 2026-08-15 03:24:40.467057+00:00 | succeeded | 2f117d6dcce73a53253aeb7b599eac0444feea9ca195c0d449d632a5ebe36b59 | 3,988 | null | """Conditional optimizer screen: lr3."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,assert_model_state
W,D,H=4,64,32
class C:
def __init__(s,vocab_size,max_seq_len):s.vocab_size,s.... | {
"id": "1182b32b-66b0-4372-bec5-4bc63e0751ea",
"created_at": "2026-08-15 03:24:40.467057+00:00",
"db_md5": "5875b6ee33ec06641c91ae76c8ce6b0c",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "584f6a33-5f0d-456a-a440-8cf0146b028a",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.1622058444389713,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.146861809786125,
"example_count": 600... |
1194508e-eeba-438a-9718-ca23e392c0a7 | easy | sapient-sapiens | 2026-08-31 20:32:59.782855+00:00 | succeeded | d44c2b98d10df4cef497bd430d4a81b062db463cb736896fbdd5af988a403006 | 13,824 | null | """E5 one-block N-conditioned weight-delta recurrence with two-call BPTT windows."""
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_mode... | {
"id": "1194508e-eeba-438a-9718-ca23e392c0a7",
"created_at": "2026-08-31 20:32:59.782855+00:00",
"db_md5": "3ca4d2a08e904847ea809954318547c4",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "89ca8c14-8441-4b4c-b4b6-1c4d8879c0db",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.345673864657009,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003750000020178656
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1612653732299805,
"example_count": 600... |
119469af-9ed7-4154-8d3c-22ef7e2c73f7 | easy | DDanlov | 2026-08-24 20:44:06.500774+00:00 | succeeded | 266981c09ddef157d6fa53c5ad3d362e453094c7232461acf0d68bf92add333d | 49,568 | null | """
Unified Pure Skipless Prefix Deliberation Model + Transformer Readout Decoder
Architecture:
- Deliberation Core: dim=1536, num_heads=24, d_ff=6144 (Pure Skipless SO(2)^{D/2} Givens Recurrence)
- Token Layout: Recurrent loop processes ONLY [Prompt (L_p), 4 Pause Registers] (Zero target tokens in loop)
- Recurrent At... | {
"id": "119469af-9ed7-4154-8d3c-22ef7e2c73f7",
"created_at": "2026-08-24 20:44:06.500774+00:00",
"db_md5": "32d4416795c5ad0c6a3dce9408bdd729",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "498cd20a-17a2-48d6-9e29-4fdd4133303e",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.7403152126245347,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004166666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.783179551496634,
"example_count": 600... |
119fc246-600c-4680-945b-cb5ae2f7d0bd | easy | oupadhyay | 2026-08-28 02:55:18.332379+00:00 | succeeded | 6c708f0b751405e708fc2d1866f2a09f7080fb0ef9947f1953bdf269ea0810f4 | 7,786 | null | """N-routed BiScan experts with a label-free sharp and balanced routing preference."""
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import OptimizerBundle, Submission, assert_model_state
PAD, N, X, T, ANS, DIGIT = 0, 2, 3, 4, 5, 7
MAX_STEPS = 64
EXPERTS = 128
LOG_EXPERTS = 4.85203... | {
"id": "119fc246-600c-4680-945b-cb5ae2f7d0bd",
"created_at": "2026-08-28 02:55:18.332379+00:00",
"db_md5": "0d5280cfe0f7afa34346dd939af8f378",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "bac6aa2d-97e3-4398-bd84-99ec65dad1de",
"tier": "easy",
"dataset_id": "e6",
"status": "s... | {
"score": {
"mean_loss": 2.219114706940847,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06846847244211146
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2516684532165527,
"example_count": 60,
... |
11a4a2f7-eade-48c9-8821-3dbbb198575d | easy | viridale | 2026-08-29 00:31:10.380655+00:00 | succeeded | a91e9286431f2414ae9c7d5903b186489e4a56e92698ae8cf54aa72da28b722b | 63,821 | null | """Full signed-byte static cube plus sparse lifted-quadratic action.
hypothesis: v992 still omits ordinary three-coefficient quadratics whenever at least
two coefficients lie outside [-4,4] and one lies outside [-64,64]. The coherent
H1 evidence against v952 only covered [-64,64]^3. Extend exactly the autonomo... | {
"id": "11a4a2f7-eade-48c9-8821-3dbbb198575d",
"created_at": "2026-08-29 00:31:10.380655+00:00",
"db_md5": "0f72b7d17b8ceb0d7237687d8f9554e0",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "00017303-2ba2-4bb1-aa06-925c517d79a0",
"tier": "easy",
"dataset_id": "e6",
"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": 60,
"exact_accuracy": 1.0,
"c... |
11a5e53e-450b-41fd-87cc-11a2db57e658 | easy | MarcoF1 | 2026-08-13 06:30:47.821615+00:00 | succeeded | c223b164e98d7e75e3499e15c0e0b7aa18de999581501723500c2c2fbd5b165d | 6,257 | null | """Weight-tied looped transformer; latent depth scales with the prompt's T.
Hypothesis treatment arm: one tied block learns "one modular-squaring step" and
is re-applied 2T+2 times, where T is parsed from the prompt's digit tokens.
Input re-injection and per-iteration RMSNorm keep the state bounded so the
recurrence c... | {
"id": "11a5e53e-450b-41fd-87cc-11a2db57e658",
"created_at": "2026-08-13 06:30:47.821615+00:00",
"db_md5": "b401f58a95f0c86693816c8f13321eba",
"submitter": "Marco Fleming",
"github_login": "MarcoF1",
"run_id": "f87a3d8c-ec6d-43f5-bea8-7af20aa4c2e0",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 3.6888307817558994,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.055000000347693756
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.892965316772461,
"example_count": 100... |
11aabf27-bcd4-4e67-b117-a4fcbb44c2d4 | easy | ehonig | 2026-08-14 01:32:52.604675+00:00 | succeeded | d4863f4de678d842869b640fbf6ce59ff6dec0b2ff6dff8337c93025539210f1 | 54,753 | 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": "11aabf27-bcd4-4e67-b117-a4fcbb44c2d4",
"created_at": "2026-08-14 01:32:52.604675+00:00",
"db_md5": "f62c4fc3d75fd4327b86be1aebf443ce",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "05985dae-a735-4f63-bb54-eac34e57e64e",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 5.167638421058655,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.21666666865348816
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.925719261169434,
"example_count": 100,
... |
11ae48e1-0e0f-4178-bc44-207ff359bd6e | easy | yashkant | 2026-08-10 21:35:50.227394+00:00 | succeeded | 3eb459335367ea4c052b9f8d51ff77c5edc0d0fd58d79fefbf76554857d0e74c | 22,263 | null | """Standalone compact-bigram hash, AdamW LR3.5e-3, EMA 0.997."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
from torch import T... | {
"id": "11ae48e1-0e0f-4178-bc44-207ff359bd6e",
"created_at": "2026-08-10 21:35:50.227394+00:00",
"db_md5": "dc43b0b83180cc829c1a42915b450d28",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "731deeff-fd26-4a75-9cec-544fa8e35395",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 2.141849340269058,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1442995285239457,
"example_count": 600... |
11b11e5e-43db-4796-8a09-e103b63fbc06 | easy | yunjiangster | 2026-08-26 23:27:05.270791+00:00 | succeeded | dd6991bf3a20ac913fe4b3325a361bffecd0e379ff94d40f513f33df2e6e9484 | 22,729 | null | """Quantized Markov Digit Machine for the One Layer Deeper benchmark.
The model parses the public prompt format into fixed-width decimal tapes, applies
one learned, tied transition once per requested time step, quantizes the residue
back to categorical digits after every step, and right-aligns the resulting
state with... | {
"id": "11b11e5e-43db-4796-8a09-e103b63fbc06",
"created_at": "2026-08-26 23:27:05.270791+00:00",
"db_md5": "87286caca21a89002f67a023af99a7fa",
"submitter": "Yunjiang Jiang",
"github_login": "yunjiangster",
"run_id": "f3cdd152-fa42-4217-a1d0-ef5342578db6",
"tier": "easy",
"dataset_id": "e5",
"status":... | {
"score": {
"mean_loss": 9.18073787062184,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.269030100561578,
"example_count": 600,
... |
11b26de2-04fb-49d5-9dcd-09a90299dbcf | easy | ehonig | 2026-08-17 21:30:41.489156+00:00 | succeeded | 56dce4a432d6d3599299548a9c1ac3d4310a37ffe0bc1084c11f397a69fc2ed1 | 74,983 | 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": "11b26de2-04fb-49d5-9dcd-09a90299dbcf",
"created_at": "2026-08-17 21:30:41.489156+00:00",
"db_md5": "15ae72d97e1adffbbd71a4d0a788698a",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "8b205a4e-2e6f-4059-974a-589bc1eb30e4",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 8.439812660217285,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08833333104848862
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.525008201599121,
"example_count": 100,
... |
11b4478d-6f23-4c2b-a94f-70707055027e | easy | nikhilkumar26 | 2026-08-30 03:17:03.816557+00:00 | succeeded | 555f2fe91af32ea9bca5c827f7f5f59804387c0f8b957cf3da199b43f7f28ab4 | 11,117 | null | """T-adaptive residue TRM: Fourier MLP square-map, detached unroll.
T is the unroll count only. Each step classifies the next residue in 0..N-1.
Digit logits and residue logits come from the last active step. Recurrence
state is detached so T=1 can train a stable one-step map that then composes.
"""
from __future__ i... | {
"id": "11b4478d-6f23-4c2b-a94f-70707055027e",
"created_at": "2026-08-30 03:17:03.816557+00:00",
"db_md5": "e2e871fc9767d874db75ff7f2b62f8e3",
"submitter": "Nikhil Kumar",
"github_login": "nikhilkumar26",
"run_id": "0214ffc2-2f13-4228-86dc-0dddd4a66635",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 0.025197995826601982,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.9933333396911621
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.01862458884716034,
"example_count": 1... |
11b7b883-1d57-40c9-9c0e-4380baae756e | easy | khushidahi | 2026-08-08 06:31:59.484877+00:00 | succeeded | 9b92417c672c5be0837ab723db6ae5b0462abd2385aeb5fd9d84f38f3b14669a | 12,542 | null | """R45 full-budget, learned-symbol baseline.
This model intentionally avoids task-specific numerical decoding:
- no fixed digit values;
- no decimal-to-integer conversion;
- no FoNE parser;
- no solver-derived supervision.
The evaluator time allowance determines the LR schedule horizon. `max_steps`
is deliberately hi... | {
"id": "11b7b883-1d57-40c9-9c0e-4380baae756e",
"created_at": "2026-08-08 06:31:59.484877+00:00",
"db_md5": "632bcdb861a65a14a8444fec38f1328f",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "5e159366-8745-4914-a06f-c673cdb086c0",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.1957041461457,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003750000020178656
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2009007930755615,
"example_count": 600,
... |
11b8faff-3f0a-408e-bb67-45647f1e9ed1 | easy | chad-atexpedient | 2026-08-13 01:24:44.946291+00:00 | succeeded | ec022faad7299676e7915547fb299bba8e6232eb68cc1760e1470bab9897d647 | 12,299 | null | """X15B: R028F with N-only inverse-length features."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
LANES = 6
CH... | {
"id": "11b8faff-3f0a-408e-bb67-45647f1e9ed1",
"created_at": "2026-08-13 01:24:44.946291+00:00",
"db_md5": "2698f2930b7135824e3bbfb25bf72ea7",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "51a8daf4-1d14-4f4b-8104-eee35c185565",
"tier": "easy",
"dataset_id": "e4",
"st... | {
"score": {
"mean_loss": 2.25227944075211,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0085416666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2536092908690093,
"example_count": 1200,
... |
11bb41ee-9be5-47ec-b841-66da3329d16d | easy | erdavis0 | 2026-08-24 16:10:30.594052+00:00 | succeeded | 8d18858c2c6e8c49faaa8fec6a8f5652d4aeade24c70fda3e7abf76d8b4bb69e | 26,473 | null | """Target-clamped primal-dual energy tape.
Public delimiter syntax only routes opaque token identities into separate,
right-aligned N, X, and T tapes. Learned local cells form distinct
interaction, carry, and modulus-conditioned workspaces. Two learned local
energy views then negotiate over the actual categorical ou... | {
"id": "11bb41ee-9be5-47ec-b841-66da3329d16d",
"created_at": "2026-08-24 16:10:30.594052+00:00",
"db_md5": "94c3763175db222103a68fb5cc760517",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "05211517-bbb9-4549-aec5-de65918123c3",
"tier": "easy",
"dataset_id": "e9",
"status": "succeeded",... | {
"score": {
"mean_loss": 8.925621032714844,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.054444445297122
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.090380668640137,
"example_count": 90,
... |
11c1d3ae-ee1d-4cc6-9bed-4edc75c3e923 | easy | chad-atexpedient | 2026-08-06 03:14:44.984150+00:00 | succeeded | e1e5bdbf5b5564095e72f9fac11440bf46f542a57fbeec55ab9406628ec7b906 | 10,852 | null | """C03: R028F with AdamW weight decay 0.3."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
LANES = 6
CHANNELS = 24
MUTABLE_START =... | {
"id": "11c1d3ae-ee1d-4cc6-9bed-4edc75c3e923",
"created_at": "2026-08-06 03:14:44.984150+00:00",
"db_md5": "d878a6a3e6490bb207f321398a47fcc8",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "0ffc3996-49cf-47ef-8119-bdf03fa74702",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.228064705879706,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0008333333457509677
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.23229914075056,
"example_count": 600,... |
11c2c812-0da6-457d-868c-b6d3c142ec16 | easy | adwhit | 2026-08-13 12:43:30.096583+00:00 | failed | fe849c5eab248d31e15dcffd70785f64fe2d1a75986b72f319150007a54a7d29 | 15,364 | null | """DeepThinker: depth-recurrent transformer for One Layer Deeper.
Architecture (task-general — no arithmetic is hard-coded):
- Weight-tied recurrent block unrolled to fixed depth (train 24 / eval 70),
so effective serial depth extrapolates beyond training T at evaluation.
- Recall connection: the original toke... | {
"id": "11c2c812-0da6-457d-868c-b6d3c142ec16",
"created_at": "2026-08-13 12:43:30.096583+00:00",
"db_md5": "fad1bbfcad857a61bab9a74a6a9a35ee",
"submitter": "adwhit",
"github_login": "adwhit",
"run_id": "e8289fe9-83f4-4509-a11a-d570abaf5e51",
"tier": "easy",
"dataset_id": "e1",
"status": "failed",
"... | null
|
11c61b56-db47-46d5-9a5b-21b140e4164c | easy | AFeinfield | 2026-08-19 23:53:21.117688+00:00 | succeeded | c86f6fecdbfcacf96e7f53dc1d936683b18f884ff27bc6a5a75025091d8d25cf | 21,420 | 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": "11c61b56-db47-46d5-9a5b-21b140e4164c",
"created_at": "2026-08-19 23:53:21.117688+00:00",
"db_md5": "bfac73f5e71a607fa90bff0cf1726671",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "2a5d27c5-6db1-46a1-adbe-2b969692d5eb",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.1483331332094204,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.155217854965963,
"example_count": 600... |
11c63b79-c187-4e78-998d-7e806d664007 | easy | gauravmishra | 2026-08-09 04:54:27.847531+00:00 | succeeded | d09e01a3b95abc9b9a2694d031afd8021c78f7cf156bb058ee95a82bac7258ac | 5,040 | null | """Basic single-pass Transformer with PyTorch AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
D_MODEL = ... | {
"id": "11c63b79-c187-4e78-998d-7e806d664007",
"created_at": "2026-08-09 04:54:27.847531+00:00",
"db_md5": "4e516848521bd26fe9e9438bf45f83c7",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "67131cb0-694f-4929-bd4a-af647b323993",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.3722808855787307,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0075
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4315608125096477,
"example_count": 600,
"... |
11c709dc-ba52-454a-9301-364f383caf8b | easy | rteehas | 2026-08-19 12:09:22.014393+00:00 | succeeded | 615e72751a99d227a277b242da675caf658382b0c631db6171d3e2e2d0090639 | 56,129 | null | """Anonymous marginal-program submission.
Mechanism
---------
The model is a differentiable probabilistic program. Its forward pass
computes, exactly, the answer-token predictions of a small structured latent
hypothesis space:
(prompt segmentation = which token types head the three fields)
x (field-role permutat... | {
"id": "11c709dc-ba52-454a-9301-364f383caf8b",
"created_at": "2026-08-19 12:09:22.014393+00:00",
"db_md5": "2b9ffddae488ad46a5d942a737838718",
"submitter": "Ryan Teehan",
"github_login": "rteehas",
"run_id": "cd0ffbc3-b830-455d-8969-cb21dc1143a8",
"tier": "easy",
"dataset_id": "e2",
"status": "succee... | {
"score": {
"mean_loss": 0.005353046348318458,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.005353046115487814,
"example_count": 300,
... |
11c8cd81-66b1-4f28-ac58-07afaad381af | easy | shan23chen | 2026-08-10 18:27:28.700453+00:00 | succeeded | 0a00baddc0de1dbf18666a37d05762e9d761589e7ee3abc4f60a4db5a19b7c90 | 39,362 | null | """Dual-root boundary-reset recurrent register Transformer.
One learned transition is applied once per requested outer step. The
transition receives an immutable decimal N register and the current learned
boundary state, but never receives T or a separate copy of the original X.
Within each outer application, scratch... | {
"id": "11c8cd81-66b1-4f28-ac58-07afaad381af",
"created_at": "2026-08-10 18:27:28.700453+00:00",
"db_md5": "adbd7ec0df95a8daa4ded341fa8d73e5",
"submitter": "Shan Chen",
"github_login": "shan23chen",
"run_id": "161ed83c-6b88-47e4-b709-22f120777459",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.254049301147461,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009999999776482582
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.250033140182495,
"example_count": 800,... |
11cc7698-d9a3-4386-84e9-4ba88927e04e | easy | gauravmishra | 2026-08-11 01:30:05.280903+00:00 | succeeded | e55085993bbb9347fea73e5e5c15d78730f218e50595dcc4b36f3c880840cbf0 | 14,519 | null | """Capability reset: end-relative output queries with optional recurrent tape.
The only line changed between hosted Easy variants is ``VARIANT``. Every
variant is generic: it uses only input embeddings, positions, and the supplied
attention mask. It does not inspect token IDs, parse fields, or implement an
arithmeti... | {
"id": "11cc7698-d9a3-4386-84e9-4ba88927e04e",
"created_at": "2026-08-11 01:30:05.280903+00:00",
"db_md5": "8f78fe06d9dba31cb783bd46f1547940",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "7fdbedb4-ef20-46f4-aaa7-42bea6d42fb3",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.234946370124817,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.237300157546997,
"example_count": 100,
... |
11cf7a8a-0944-4c2c-abae-8a35d8f1a6e4 | easy | DDanlov | 2026-08-10 08:11:40.908987+00:00 | succeeded | a44b28d5c5e39825ffa18fce33f000f888a00a26cec97b92ffae3653efcb700c | 16,673 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
def init_cutting_the_skip(module):
"""
SVD-based initialization to ensure initial Jaco... | {
"id": "11cf7a8a-0944-4c2c-abae-8a35d8f1a6e4",
"created_at": "2026-08-10 08:11:40.908987+00:00",
"db_md5": "f6791c1813c085d715827c34fb69d42e",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "cab1b91e-1949-4ad9-84b8-1989718eaf96",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 4.391052199123506,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.053333330353101095
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.620725631713867,
"example_count": 100,... |
1209dd9c-770d-4cc7-9075-97bdffdea24c | easy | KaustubhKumar05 | 2026-08-25 16:22:07.612359+00:00 | succeeded | cede8f877bf90dc9d38af3879dc3680a73b6693631c3476085830cfc30a955d0 | 16,027 | null | """final_v1: culmination candidate. v8 (N-FiLM) + Muon momentum 0.5 + tail-EMA.
A small MLP embeds the modulus digits (mean-pooled over the N span) and
emits per-channel scale/shift applied to the cell input and a bias on the
update gate, every iteration. Hypernetwork-lite: the executor is modulated
by N rather than h... | {
"id": "1209dd9c-770d-4cc7-9075-97bdffdea24c",
"created_at": "2026-08-25 16:22:07.612359+00:00",
"db_md5": "4c081115900db9bcea126ae905aeee44",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "905de93f-4f79-4bb7-b912-c935a7f23d8d",
"tier": "easy",
"dataset_id": "e3",
"status": "succee... | {
"score": {
"mean_loss": 6.762591187664437,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.182003409209551,
"example_count": 800,
"exac... |
120d9355-c3b8-4d92-bb67-9b4c0868830f | easy | DDanlov | 2026-08-20 11:59:42.004995+00:00 | succeeded | 1c8d813e3f130c6b04a5ba83d9720ab19c7a93373d0658017b4e0e30a48ef277 | 20,913 | 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": "120d9355-c3b8-4d92-bb67-9b4c0868830f",
"created_at": "2026-08-20 11:59:42.004995+00:00",
"db_md5": "6fc9ad1159510cfd42353d2b2c4c4816",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "1dfeabf5-4b19-4e43-ab67-fe50f784fa52",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.2914992605722,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009166666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.288499223811744,
"example_count": 600,
... |
12139863-9299-496e-9906-fc48ee4c3bfb | easy | shirvani-jr | 2026-08-30 23:42:39.567564+00:00 | succeeded | 2adce6398218d7ef30867e2d0e5a0b846218b9cbc44983186b3a31c8410076ae | 19,060 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PERIOD_MINIMUM = 11
PERIOD_MAXIMUM = 100
E... | {
"id": "12139863-9299-496e-9906-fc48ee4c3bfb",
"created_at": "2026-08-30 23:42:39.567564+00:00",
"db_md5": "def1a9ed76fee5a18d6c705aa46ec759",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "46e83806-9586-49d2-8fc5-0d26c1a659ed",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.4308537244796753,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.036874997429549694
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.698725700378418,
"example_count": 800... |
1214f695-afa0-4964-a8c4-10bd6762e5cf | easy | oupadhyay | 2026-08-12 07:39:18.096132+00:00 | succeeded | a533bb16cd99eb360ecd1d13bb97869bad7d05d68e7d017db53ecde62f54a2ab | 8,290 | null | """Standalone sgu_opt_lr8_easy_v1 optimization of RMS/axial crossing C21 for the Easy E5 tier."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, assert_model_state
PAD, ... | {
"id": "1214f695-afa0-4964-a8c4-10bd6762e5cf",
"created_at": "2026-08-12 07:39:18.096132+00:00",
"db_md5": "1f50f0b1144be56f2fba6a2723178200",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "8b98d6c5-51a7-46b4-a129-d8f6ff9551bd",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.312501598745362,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005416666691501935
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.201505530575466,
"example_count": 600,... |
1219f420-033f-40a3-ab2d-7de6853842fe | easy | sapient-sapiens | 2026-08-14 11:40:48.192609+00:00 | succeeded | 4f02397935fbd88832e4452403cf9ffbd2d276e025a45b43f499f63549f5be77 | 16,890 | null | """Fixed-T=2 tied Transformer for the E3/E4 operator-recovery campaign.
This file is the authoritative, standalone hosted submission template. The
variant generator changes only the constants in the marked configuration
section. The model never reads T: its four-block core is applied exactly twice.
"""
from __futur... | {
"id": "1219f420-033f-40a3-ab2d-7de6853842fe",
"created_at": "2026-08-14 11:40:48.192609+00:00",
"db_md5": "2fac085de02b8ed81d2e3d87dffa8e05",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "b0eddcdb-7ce1-49c4-ab35-a6a063dd312d",
"tier": "easy",
"dataset_id": "e3",
"status": "su... | {
"score": {
"mean_loss": 2.1971158639828996,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.015000000000000001
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1970444120777017,
"example_count": 80... |
1219fa49-b73f-420b-9ecc-753cb6efba21 | easy | lzy54 | 2026-08-26 18:21:31.378888+00:00 | succeeded | 5c53068d8d600e9dd19d356e1e3ff9bf603ba0dbf9a9bc9daf2c341cf9c7c567 | 4,546 | null | """A7 trained for four loops and evaluated for a configurable loop count."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL ... | {
"id": "1219fa49-b73f-420b-9ecc-753cb6efba21",
"created_at": "2026-08-26 18:21:31.378888+00:00",
"db_md5": "8602ed4730cceb69fd5d3973aa18a48e",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "802808c7-8f07-498e-a7d0-58bd4fbebc76",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",
... | {
"score": {
"mean_loss": 4.4122395515441895,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018468469381332397
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.372420787811279,
"example_count": 60,... |
121ec26b-4ab8-4733-bf45-8a9935b76794 | easy | mhlov000111 | 2026-08-14 18:18:01.630380+00:00 | succeeded | a731855d8fdf69915bdf0cf1b948244bc3af3293861c5edcf54660da6862dcae | 3,716 | 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": "121ec26b-4ab8-4733-bf45-8a9935b76794",
"created_at": "2026-08-14 18:18:01.630380+00:00",
"db_md5": "02bb8a97df75a4c219d5d9074f84ec02",
"submitter": "mhlov000111",
"github_login": "mhlov000111",
"run_id": "d0072cb3-7996-4310-8d65-bc67200574d9",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 5.9711127281188965,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.014999999664723873
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.889996528625488,
"example_count": 100... |
1221a749-b552-4822-84be-463a37489ffc | easy | sapient-sapiens | 2026-08-09 15:02:36.331603+00:00 | succeeded | 9ba049d1c2df4d13aa5a119aec4ec3b5df44dd0311ae58892e3107ddc2d325bd | 13,048 | null | """Near-500M adaptive recurrence grown from the hosted-working baseline."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
... | {
"id": "1221a749-b552-4822-84be-463a37489ffc",
"created_at": "2026-08-09 15:02:36.331603+00:00",
"db_md5": "34b550e6189e2eed912f02c824ceca34",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "0312e917-6e2b-4339-8ed4-c6f9541f8047",
"tier": "easy",
"dataset_id": "e4",
"status": "su... | {
"score": {
"mean_loss": 2.512369244529739,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0028125
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5096425210651496,
"example_count": 1200,
... |
1224af92-e0fa-49b2-90e5-06ba76750b7d | easy | DDanlov | 2026-08-09 00:17:50.980993+00:00 | failed | 38472ee79781d72de55133307e2143e74202e1e12c7d5c08faa5d84a204a37b8 | 15,981 | 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": "1224af92-e0fa-49b2-90e5-06ba76750b7d",
"created_at": "2026-08-09 00:17:50.980993+00:00",
"db_md5": "73d997d86480f611e96dc67f886ee80b",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "78633d70-a19e-4938-97f3-dc0170f002d5",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
1232ed5a-9493-48da-a50a-72db86fda168 | easy | Laz4rz | 2026-08-07 23:39:47.024482+00:00 | succeeded | 0264a1019c5981c58b05bb405982b9a5fc0c5f22e4f79192bba565104fe7bf50 | 27,691 | null | """PDiff DiT + masked GIDD, adapted to the One Layer Deeper API.
The model, constant-pi objective, and adaptive sampler follow the current
PDiff working tree's ``gidd_brevo1_paper_aligned_nplus2`` path. The benchmark
adapter keeps the prompt and completion canvas separate internally, then
converts sampled answer toke... | {
"id": "1232ed5a-9493-48da-a50a-72db86fda168",
"created_at": "2026-08-07 23:39:47.024482+00:00",
"db_md5": "de76380160fb8b62f20fcfe661a8b98a",
"submitter": "Laz4rz",
"github_login": "Laz4rz",
"run_id": "3994a6e7-a2b7-4dcf-80af-809812f6e5cb",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.810490442856603,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0023958333457509673
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.818497806505356,
"example_count": 120... |
1242b866-1ad0-4399-8e4d-e4200e1e908b | easy | velocizapkar | 2026-08-06 22:33:31.899386+00:00 | succeeded | 343fb9fdd4df5d23de08f7b4244f64e665eb709b31ef88163146031b33e62696 | 55,676 | null | """Fast packed residual-free Attention-Residual transformer with Muon.
The model is a generic learned sequence model. It contains no parser,
arithmetic primitive, task constant, generated data, label feedback, or answer
construction. Every macro-state is produced by a serial pack of attention and
MLP transformations... | {
"id": "1242b866-1ad0-4399-8e4d-e4200e1e908b",
"created_at": "2026-08-06 22:33:31.899386+00:00",
"db_md5": "420d39b343aadb46f4eb03a81de8b57d",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "673fad9f-2ab8-4c05-8733-a26718afb452",
"tier": "easy",
"dataset_id": "e3",
"statu... | {
"score": {
"mean_loss": 3.513210892677307,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.019375000149011612
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.514875650405884,
"example_count": 800,... |
12452568-7f5e-4377-98f9-6e208009a03d | easy | sapient-sapiens | 2026-08-15 20:24:56.898716+00:00 | succeeded | 469ae48173121f8a328b5c1f44d0bff0e2f098d243dcd227975cee12399b6829 | 17,772 | null | """T-only adaptive chunk routing over a learned shared operator.
The controller sees only the categorical tokens inside the T prompt field and
is trained solely by final-answer loss. It never parses T numerically and T is
not used as an intermediate supervision target. Inputs use learned additive
token, semantic-gro... | {
"id": "12452568-7f5e-4377-98f9-6e208009a03d",
"created_at": "2026-08-15 20:24:56.898716+00:00",
"db_md5": "38cda1d694d318221e86f82dd23bbc39",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "e5fd84e1-a1d7-4e24-bf82-0ac7339e6a07",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 7.039925035395824,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.041666665176550544
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.300836563110352,
"example_count": 100,... |
12512f5f-396c-4635-8984-26c7b7e2f660 | easy | yashkant | 2026-08-16 00:14:13.103347+00:00 | succeeded | b4d8b510d7a42baa9cf13875d6570b59bd0a3aa94ceadac42576b914d8c49a5a | 37,404 | null | """Round1863 learned state/T-conditioned BCH cubic coefficient.
Both arms keep the exact Round1851 AdamW 5e-3 / EMA .995 carrier and eager
BCH graph. A four-parameter bounded gate reads live correction geometry,
relative cubic size, and the public parsed T field at every recursive cycle.
The equal-work control compute... | {
"id": "12512f5f-396c-4635-8984-26c7b7e2f660",
"created_at": "2026-08-16 00:14:13.103347+00:00",
"db_md5": "0a16107a9d9760c3d160b8bc49007ecb",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "1cb07927-16b3-4179-bae0-97b84795b242",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 2.145144595698998,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009166666679084301
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.149166072965203,
"example_count": 600,... |
125915ea-78fe-423f-8119-edc169107b74 | easy | jordanrubin | 2026-08-11 00:30:03.792381+00:00 | succeeded | 5627289ebca0e4853183d79604695f32d270f428a44ccf506384676ba86c6797 | 20,108 | 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": "125915ea-78fe-423f-8119-edc169107b74",
"created_at": "2026-08-11 00:30:03.792381+00:00",
"db_md5": "bdc145351ed7c8a97c4c6df2f8d43449",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "63ab5996-69fa-4a57-afc4-916eae2c855b",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.7085165766367707,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666489715377
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1570187960772236,
"example_count": 60... |
125f0450-4ed5-45db-b5a4-5fc2c0cc3353 | easy | lpbb | 2026-08-18 09:45:10.733964+00:00 | succeeded | ea5e72d4d49bb5e0d72a2ceb878cf11e5c4562810505fa268c96fd8fa22ade6b | 4,457 | null | """S0 — control: same shape as S1 but 24 untied layers (fixed depth, no weight tying).
Measures the 'tying tax'; by construction its depth cannot follow T."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
Optimi... | {
"id": "125f0450-4ed5-45db-b5a4-5fc2c0cc3353",
"created_at": "2026-08-18 09:45:10.733964+00:00",
"db_md5": "bdfd1c2f3bd08e7e0700d36823613c87",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "fc340e35-a29e-45aa-b97d-19aef2853ac5",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 6.257918119430542,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07833333313465118
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.407680034637451,
"example_count": 100,
... |
12643609-7d37-4cdd-8852-4cb33f39c7c7 | easy | alirezashirvani-jr | 2026-08-30 08:42:53.197158+00:00 | succeeded | 1077d93bff0f88167a1af0f45e4abf432c45dabd2262a6e4423038ba0799d48b | 26,556 | 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": "12643609-7d37-4cdd-8852-4cb33f39c7c7",
"created_at": "2026-08-30 08:42:53.197158+00:00",
"db_md5": "d3c25e5ed3fb41c685a17ce10e0fdbed",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "5c59a487-3270-4499-9fa4-9bc26ed37dbd",
"tier": "easy",
"dataset_id": "e2",
... | {
"score": {
"mean_loss": 1.0678547322750092,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.7566666901111603
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.6645519137382507,
"example_count": 300,... |
1267e320-2a68-43d8-a39b-7f54534bc584 | easy | DDanlov | 2026-08-24 03:38:10.705695+00:00 | succeeded | ffbcad27edd83c0d965adadb11b69a43b06f76274e6d953d4c9f54d30853901a | 37,380 | null | """
Unified Hybrid Masked Deliberation Model (40.4M Parameters) - Pure Skipless, No Cross-Attention
Architecture:
- Scale: dim=1536, num_heads=24, d_ff=6144 (Exactly 40,437,888 parameters)
- Hybrid Prefix-Bidirectional + Causal Target Attention Mask
- Token Layout: [Prompt (L_p), 4 Pause Registers, Target Tokens (L_tgt... | {
"id": "1267e320-2a68-43d8-a39b-7f54534bc584",
"created_at": "2026-08-24 03:38:10.705695+00:00",
"db_md5": "b8532faab9e9dacbf95d75821e48f5a0",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "87388168-6102-4fc6-a692-8757cf4beeb1",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.3445681999735686,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0020833333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.3361470250270826,
"example_count": 6... |
12698b79-bb25-47cf-8828-8460e8741549 | easy | alirezashirvani-jr | 2026-09-01 01:19:59.489315+00:00 | succeeded | 7d61d04ca8b56d81d595d49fbfe46895b2d49f509e1142681c7612fa1e516382 | 35,278 | 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,
)
MLP_EX... | {
"id": "12698b79-bb25-47cf-8828-8460e8741549",
"created_at": "2026-09-01 01:19:59.489315+00:00",
"db_md5": "31952b37af1582c6bccdf3afed238c20",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "be5bcfaf-70fa-4072-b480-1a34de9c4edc",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 2.9322231489460213,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.027499999776482584
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.965339183807373,
"example_count": 600... |
12742e7b-5ca3-4c34-97cf-c358c3a8c269 | easy | DDanlov | 2026-08-17 12:07:12.809447+00:00 | succeeded | 29d60bd8576fe8cba835000b8af6193cc5bdbf75da29621801736be3836a3822 | 30,862 | null | """
Scalable Recurrent Deliberation Transformer (ScalableDEPHNet)
============================================================
Architecture:
- Pure No-Norm: Single unit RMS normalization strictly isolated to the input token embedding layer.
Zero LayerNorm / RMSNorm inside recurrent deliberation blocks.
- Pure No-Skip... | {
"id": "12742e7b-5ca3-4c34-97cf-c358c3a8c269",
"created_at": "2026-08-17 12:07:12.809447+00:00",
"db_md5": "cb7045071bd742ced24da1123ef40398",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "06ce1c68-e864-4d5c-b57b-d127070cd2b4",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.005768395570151,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06333333045244217
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1149394512176514,
"example_count": 100,... |
1274ba31-da36-47c8-a457-7f0cdbe975f7 | easy | viridale | 2026-08-15 18:32:04.902790+00:00 | succeeded | fcd808694223d067b89dd44b0530ce6d7e94f3aa5cc16944b141a7566edb0187 | 29,553 | null | """Boundary-preserving local evolution in a sparse cubic character fiber.
hypothesis: after the exact four-coordinate T1 boundary is latched, preserve complete
elite dependencies and explore the remaining recurrent fiber by one-coordinate
moves or cubic-owner switches instead of independent redraws alone.
axis... | {
"id": "1274ba31-da36-47c8-a457-7f0cdbe975f7",
"created_at": "2026-08-15 18:32:04.902790+00:00",
"db_md5": "062bfc3c613abb515ba2c57f62a3ad80",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "8b40352c-96a4-4de8-b9df-e908aa576f9f",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 18.655654907226562,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011249999748542905
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 18.703895568847656,
"example_count": 80... |
1275a63f-137e-491b-9ed0-951154ed803b | easy | yashkant | 2026-08-17 00:52:58.210653+00:00 | succeeded | 9b4cfe8e0bf5b7cab3ca2ef38181134e6f5e70d2df37a63d1efa9605a7811cec | 43,554 | null | """Round2190 removes cross-step Dykstra correction carry.
The retained official carrier, primal recurrence, two learned projections,
antisymmetric set codes, parsed-T allocation, optimizer, and EMA are fixed.
The candidate reuses the learned initial A/B correction offsets on every
cycle instead of carrying hand-writte... | {
"id": "1275a63f-137e-491b-9ed0-951154ed803b",
"created_at": "2026-08-17 00:52:58.210653+00:00",
"db_md5": "0754f6fe90068b218bbcff849a9e897c",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "356f7db5-750c-4bf2-b73d-04972063547b",
"tier": "easy",
"dataset_id": "e4",
"status": "succeed... | {
"score": {
"mean_loss": 2.069940919781617,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0104166666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.0729175960477098,
"example_count": 1200,... |
1275c4b4-36b8-4afb-bf9b-30c29dd23396 | easy | sapient-sapiens | 2026-08-19 17:54:56.330020+00:00 | failed | 3936aa420b1bf57f43d77b1cc5a39a6de18ea831022aef3a7cdfeab087352db2 | 15,398 | null | """Phase A baseline: legitimate adaptive digit-space modular squaring.
N, X, and T remain categorical token sequences. Marker locations define fields,
but no field is decoded into a number. A field-local encoder prevents T from
leaking into the shared arithmetic operator. The controller alone sees T.
Training uses on... | {
"id": "1275c4b4-36b8-4afb-bf9b-30c29dd23396",
"created_at": "2026-08-19 17:54:56.330020+00:00",
"db_md5": "84edca4b75734de4134b8123cd24c62a",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "3713e143-6822-4640-87b6-c8f8711ba8a7",
"tier": "easy",
"dataset_id": "e5",
"status": "fa... | null
|
1280a3d2-8df8-4faf-9617-a791a3e1fe2b | easy | anirudh-chakravarthy | 2026-08-12 03:58:07.220628+00:00 | succeeded | c0768813b5838f7339c9bb98c4ff826ae701ea3d19806f63dd3d0dd5d4749032 | 14,968 | null | """A learned state transition applied once per requested time step."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
asser... | {
"id": "1280a3d2-8df8-4faf-9617-a791a3e1fe2b",
"created_at": "2026-08-12 03:58:07.220628+00:00",
"db_md5": "1b25a6dc53409f445e035456e71548d9",
"submitter": "Anirudh S Chakravarthy",
"github_login": "anirudh-chakravarthy",
"run_id": "38a52df8-c29c-46d9-8f2b-5089de49abba",
"tier": "easy",
"dataset_id": "... | {
"score": {
"mean_loss": 6.263512370716715,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010416666666666668
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.041165398909907,
"example_count": 600,... |
1281d065-d6d9-4ff9-b1cb-07d231a37676 | easy | DDanlov | 2026-08-11 21:59:03.938417+00:00 | succeeded | 1980bac1425c29d2ac8f85012127082c1b32a096dcc11dd6d8fb0ba7da34e35c | 20,813 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
def init_cutting_the_skip(module):
"""SVD-based orthogonal initialization to ensure initia... | {
"id": "1281d065-d6d9-4ff9-b1cb-07d231a37676",
"created_at": "2026-08-11 21:59:03.938417+00:00",
"db_md5": "9ffc10ba2137a49498edb8dd297b2395",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "d2660057-82cd-45d8-80dc-930b2636f4f8",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 45.961639153371266,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0020833333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 47.26586563384052,
"example_count": 60... |
1282c12e-2007-4d8d-860c-3f5b2651b19e | easy | yashkant | 2026-08-17 00:45:07.306156+00:00 | succeeded | 7a227242ac2cd50d5904e44df6948fcd1cbfabcb80062fad84c58b11d1fd804e | 41,441 | null | """Round2190 removes cross-step Dykstra correction carry.
The retained official carrier, primal recurrence, two learned projections,
antisymmetric set codes, parsed-T allocation, optimizer, and EMA are fixed.
The candidate reuses the learned initial A/B correction offsets on every
cycle instead of carrying hand-writte... | {
"id": "1282c12e-2007-4d8d-860c-3f5b2651b19e",
"created_at": "2026-08-17 00:45:07.306156+00:00",
"db_md5": "b1d5d76eb3993112d18f080614e63c88",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "5779c212-1f3d-4004-ac5b-3b639b3690dd",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.1325147356188325,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018125000167638065
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.132843144560843,
"example_count": 800... |
12879388-abf9-4d77-afaa-281cb299bbfa | easy | yashkant | 2026-08-17 00:46:22.810132+00:00 | succeeded | 9b2d230bb19e53828319f90a7434c87266aedfff50f82fda48c974958dce3dca | 41,475 | null | """Round2192 late progressive effective batch on the best official source.
The exact model, initialization, loss formula, two-pass SAM plus
Compositional-Muon, LR schedule, and EMA are retained. The evaluator
supplies 96 physical rows. A prefix carries loss before a fixed number
of completed updates; afterward all 96 ... | {
"id": "12879388-abf9-4d77-afaa-281cb299bbfa",
"created_at": "2026-08-17 00:46:22.810132+00:00",
"db_md5": "5efba82dda7cebec1d2a13635caacad7",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "197a4dfb-b99d-47c1-83d5-43ac234c2c0b",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.1257314381557295,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01375000011175871
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1258065871044924,
"example_count": 800... |
128ab035-04e7-4c51-a30a-e6571fa0be7c | easy | DDanlov | 2026-08-08 10:21:05.520431+00:00 | succeeded | e2076ae330bf3fbb1220c24c46ec0fe09c4c05ad84c8aadca281f094bb9fd99d | 15,561 | null | from __future__ import annotations
import sys
import os
import math
import time
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim.optimizer import Optimizer
VOCAB_SIZE = 17
DIGIT_OFFSET = 7
class RohanShampoo(Optimizer):
"""Self-contained RohanShampoo optimizer with eigenvalue ma... | {
"id": "128ab035-04e7-4c51-a30a-e6571fa0be7c",
"created_at": "2026-08-08 10:21:05.520431+00:00",
"db_md5": "62e1d18fa43eaf10b466455b965d32b2",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "33df0264-de74-4f16-a26d-d6287f16eff2",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 46.41871875866212,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0004166666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 47.55731597823412,
"example_count": 600... |
128bdee2-606a-418f-86c1-4caec4f09c0e | easy | Ranoobaba | 2026-08-16 08:08:24.618641+00:00 | succeeded | 701046cb4e863e859ddd73bb945eb336f0b51d04245e644abde3822a89d2f68d | 5,724 | null | """Looped transformer, iteration count conditioned on T read from the prompt.
Core idea: learn one latent "squaring step" operator and apply it T times.
Depth generalization comes from the loop structure itself; the model never
needs more parameters for deeper problems, only more iterations.
"""
from __future__ impor... | {
"id": "128bdee2-606a-418f-86c1-4caec4f09c0e",
"created_at": "2026-08-16 08:08:24.618641+00:00",
"db_md5": "c166ae13643fa1d8d1588c66769ec6e7",
"submitter": "Syed Rayyan Ali",
"github_login": "Ranoobaba",
"run_id": "a6f88ab7-1e59-4730-a130-f270cac86510",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 8.112375937792429,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.03499999850988388
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.22767448425293,
"example_count": 100,
... |
1293c983-a38a-4cb5-9d35-670d63358605 | easy | erdavis0 | 2026-08-24 12:58:41.089757+00:00 | succeeded | ba86db99316d0dc771d529efb36126d73202e2391a2af0610f2dfcc6dce5f577 | 28,763 | null | """Reversible augmented-state transducer over aligned learned digit tapes.
Public delimiter syntax only routes opaque token categories into distinct,
right-aligned N, X, and T tapes. Learned local cells compile separate
interaction, carry, and modulus-conditioned workspaces. N is isolated from
interaction and carry ... | {
"id": "1293c983-a38a-4cb5-9d35-670d63358605",
"created_at": "2026-08-24 12:58:41.089757+00:00",
"db_md5": "63ed4e7b1ef5d86f430c94affef5dcb1",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "338e635a-659a-4559-85a5-b1b08f706729",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.4874095303529327,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006770833358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.486946896646649,
"example_count": 120... |
129ebdc3-2004-4953-9901-20b819b72f67 | easy | christopherhwood | 2026-08-06 08:15:45.239583+00:00 | succeeded | 900e36c413fa0c5f8f3d39a2476d4151b2dcf9625d646929999063e1a082f295 | 22,060 | null | """Endpoint-clamped global program lattice for repeated modular squaring.
One tied product-aware reducer exposes six possible stopping depths. They are
not supervised as intermediate answers. Instead, each depth defines a
normalized CRF over observable output digits, and the endpoint loss exactly
marginalizes a sing... | {
"id": "129ebdc3-2004-4953-9901-20b819b72f67",
"created_at": "2026-08-06 08:15:45.239583+00:00",
"db_md5": "a5a411746f9d1fbb1549468f209ab6d2",
"submitter": "Chris Wood",
"github_login": "christopherhwood",
"run_id": "b9d76387-4c2a-4d04-9986-0e5bd1bbcc99",
"tier": "easy",
"dataset_id": "e5",
"status":... | {
"score": {
"mean_loss": 13.705602452863863,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 13.713005527787145,
"example_count": 60... |
129fb234-b22d-4c68-9658-491d6b8726b0 | easy | Avyukth | 2026-08-11 19:50:43.806643+00:00 | succeeded | e1a9eb5fa0ab59689ae9de29eb0e5ea02583448bb44da2ef05945382646e3ba9 | 13,380 | null | """T-adaptive recurrent Transformer: loop count scales with T read from the prompt.
The benchmark family composes a transition function T times. This model bakes
that structure in: it decodes the integer T from the prompt's digit tokens and
applies one shared block BASE + K*T times. Phase embeddings are cyclic (period... | {
"id": "129fb234-b22d-4c68-9658-491d6b8726b0",
"created_at": "2026-08-11 19:50:43.806643+00:00",
"db_md5": "3d202f2a0b11e1c5ce93679af1346433",
"submitter": "Subhrajit Makur",
"github_login": "Avyukth",
"run_id": "398f3d1e-89c9-4bb0-8c0a-bbba10104d0c",
"tier": "easy",
"dataset_id": "e1",
"status": "su... | {
"score": {
"mean_loss": 9.223771095275879,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.014999999664723873
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.836675643920898,
"example_count": 100,... |
12a5b6bc-2863-45db-a864-63fc5d6f2a4e | easy | DESU-CLUB | 2026-08-24 10:45:26.575564+00:00 | succeeded | 30b5d38fe72c0494c6f4250e50a70b1943cc1147b134fea0d50c118cfb0dc383 | 10,580 | null | """Huginn-style looped Transformer whose iteration count is read from the T token.
prelude -> [core applied T times, input re-injected each step] -> coda
Forked from experiments/_baselines/adamw_transformer.py; the Block, RMSNorm and
optimizer are unchanged so that the only varied axis is the depth mechanism.
"""
fr... | {
"id": "12a5b6bc-2863-45db-a864-63fc5d6f2a4e",
"created_at": "2026-08-24 10:45:26.575564+00:00",
"db_md5": "8866ad35cfe77d3bdafddb16a4ffd4f8",
"submitter": "Low Keng Hoong, Warren",
"github_login": "DESU-CLUB",
"run_id": "030af1d9-40d1-4d50-9e61-ac457648e2ea",
"tier": "easy",
"dataset_id": "e5",
"sta... | {
"score": {
"mean_loss": 3.6513880774926895,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333333333335
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.118030306202414,
"example_count": 60... |
12ab7f9b-fe25-4ddf-b81c-74481f433315 | easy | lzy54 | 2026-08-26 14:39:06.439152+00:00 | succeeded | 8acdcf3e3a4390f2e2731dd795aa998bb913e2b1ac2331cd29160337fa0ce9a6 | 4,168 | null | """Random-depth shared Transformer with additive input recall and AdamW."""
from __future__ import annotations
import random
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state... | {
"id": "12ab7f9b-fe25-4ddf-b81c-74481f433315",
"created_at": "2026-08-26 14:39:06.439152+00:00",
"db_md5": "9aa79847f9617cb1e08cb4e46210e164",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "bc71453a-40a9-4331-a8e9-6e58ee6f36f3",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",
... | {
"score": {
"mean_loss": 1.9307385683059692,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.015090090688318014
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.875497817993164,
"example_count": 60,... |
12b156dc-2791-41cd-ad9a-d506af1796ea | easy | newjordan | 2026-08-09 16:02:27.608974+00:00 | succeeded | 64347a606e11a7ad356aab81c33bd6a01e2c1b3c8c67e8c63df3f44cecbc3942 | 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": "12b156dc-2791-41cd-ad9a-d506af1796ea",
"created_at": "2026-08-09 16:02:27.608974+00:00",
"db_md5": "ff3c7d545e2ea5c3e4ff8f9d00bbdb4a",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "922f939e-506e-4172-80bd-27eedeb88c07",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 1.2959873151608571e-05,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.2952967765746271e-05,
"example_count": 600,
... |
12b6cf66-3718-441f-9b96-2130d88cba49 | easy | ehonig | 2026-08-14 01:04:46.404626+00:00 | succeeded | 6cc7f0ea1126c8f8a210569bdbe0d322f6256813c1a3e5cecb857c62b7684268 | 54,754 | null | """DigitLoop — a place-value recurrent model for repeated modular squaring.
The task is ``y = x^(2^T) mod N`` with the answer read back tail aligned, one
decimal place value per token position from the right. A stack of attention and
MLP layers has no primitive for the two operations this needs — a partial-product
co... | {
"id": "12b6cf66-3718-441f-9b96-2130d88cba49",
"created_at": "2026-08-14 01:04:46.404626+00:00",
"db_md5": "5e22f84b86ad97e9c17d31e7481f6320",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "74ed1e21-4300-4e5c-a996-ebbece11b3a7",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 5.810546875,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.11500000208616257
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.192975997924805,
"example_count": 100,
... |
12b79b5e-b259-42f0-84d2-f5d331f4cb83 | easy | khushidahi | 2026-08-06 05:16:17.213380+00:00 | succeeded | 1b72b9fb68cdc85c4d9684da7dfc637515d0b1838dbd0d199aac738b88115f2e | 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": "12b79b5e-b259-42f0-84d2-f5d331f4cb83",
"created_at": "2026-08-06 05:16:17.213380+00:00",
"db_md5": "aa652db768057f616d108600530f1320",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "ca4f27ed-bb04-4236-9071-911d10da6f4c",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.210708260536194,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007083333563059568
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.208150625228882,
"example_count": 600,... |
12bcad81-7f76-4cb9-9c26-d14fe52dceed | easy | k-penchev | 2026-08-23 00:23:40.430113+00:00 | succeeded | 8d4c197562dda6b284bb6a00c824afb3517afb3e681e78a71a4d0267774d7c25 | 8,427 | null | """Schedule-Free gated recurrence with outer momentum beta 0.95."""
from __future__ import annotations
from collections.abc import Iterable
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
ass... | {
"id": "12bcad81-7f76-4cb9-9c26-d14fe52dceed",
"created_at": "2026-08-23 00:23:40.430113+00:00",
"db_md5": "ada6b1b04f6d6fde6e95ffea7d7deff7",
"submitter": "Kaloyan Penchev",
"github_login": "k-penchev",
"run_id": "5431bdb5-ac68-47fb-926e-2f3cdeb94544",
"tier": "easy",
"dataset_id": "e4",
"status": "... | {
"score": {
"mean_loss": 2.1176656764585315,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002187500012417634
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.118625749149427,
"example_count": 120... |
12bd4a5f-9ff2-44ce-a15a-c90de01300b1 | easy | khushidahi | 2026-08-16 13:49:40.396990+00:00 | succeeded | 896db940c2910b177aa2fe78ead23d86b217fe470ae0f59244272808a96b4c32 | 19,211 | null | """R113: recurrent tape Transformer with a learned transition microprogram.
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... | {
"id": "12bd4a5f-9ff2-44ce-a15a-c90de01300b1",
"created_at": "2026-08-16 13:49:40.396990+00:00",
"db_md5": "258d43179c5403f32f3a7eb38f3dc590",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "e634e238-8486-4546-8292-13d24f10c463",
"tier": "easy",
"dataset_id": "e6",
"status": "succ... | {
"score": {
"mean_loss": 0.8641190677881241,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.8617117404937744
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.18617668747901917,
"example_count": 60,... |
12c537cb-0f73-4778-932d-17c5c1192528 | easy | oupadhyay | 2026-08-24 01:00:26.236349+00:00 | succeeded | 7d7da2b7e2810137ec490dc2552480efc1f61d9de1d0084a001ee464e66a4880 | 4,151 | null | """Pairwise digit-relation scan for the Easy tier."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,TokenLossBatch,assert_model_state
W,D,H=4,48,96
class C:
def __init__(s,vocab_size,max_seq_... | {
"id": "12c537cb-0f73-4778-932d-17c5c1192528",
"created_at": "2026-08-24 01:00:26.236349+00:00",
"db_md5": "70330cd43930596ba1de504495dfc42f",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "c59e00a3-533b-42af-b908-602b0f5115e8",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.3191772709057217,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004583333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.212300853344357,
"example_count": 600... |
12c8905e-9b01-421e-a97c-ed31f09c649c | easy | jordanrubin | 2026-08-03 20:57:09.064455+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": "12c8905e-9b01-421e-a97c-ed31f09c649c",
"created_at": "2026-08-03 20:57:09.064455+00:00",
"db_md5": "ef7d67d872a65aa96e81ca53991ab27d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "00a35780-9b3b-4670-aaeb-1444fc7db8ac",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 2.9846639151307928,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.20937499442758659
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.685876167886626,
"example_count": 300,... |
12cba492-ae7c-47ad-8137-1219d7fa673c | easy | oupadhyay | 2026-08-08 10:39:00.728406+00:00 | succeeded | 287871d8234028240facdd222e33f126a6341263203cd1b5edc1470ea2b261f8 | 8,260 | null | """Soft autoregressive digit decoder for the marker-delimited easy task."""
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, HEADS, FFN = 192, 6, 576
ENCODER_... | {
"id": "12cba492-ae7c-47ad-8137-1219d7fa673c",
"created_at": "2026-08-08 10:39:00.728406+00:00",
"db_md5": "701de289430d9f114a38c16100486b49",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "65f0ff33-0ade-4929-8919-82785bf32ab8",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 8.725433709922138,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666554907957
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.71471118927002,
"example_count": 600,
... |
12ced345-596e-4e11-818e-4c3b80cd549d | easy | alirezashirvani-jr | 2026-09-01 00:16:58.132223+00:00 | succeeded | f73307750f4cfe59c016eea5a1457e228dbafc00a9b7f79f69a0d438d6b9fa47 | 27,677 | null | from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PAD = 0
MODULUS = 2
VALUE = 3
DEPTH = 4
ANSWER = 5
EO... | {
"id": "12ced345-596e-4e11-818e-4c3b80cd549d",
"created_at": "2026-09-01 00:16:58.132223+00:00",
"db_md5": "e78fac1c93bee431480cbb11f9901d53",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "e216020b-dfa2-43c7-9784-5a28cef1b45f",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 2.7494956041112726,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.7451826510408,
"example_count": 600,
... |
12d188d0-160c-4fe0-aa27-64868a12d1bb | easy | rteehas | 2026-08-19 08:08:43.315161+00:00 | succeeded | c5d530c890eec970749a16d99c43de58e655c95ffa149ae71d9a925947ef8f4c | 41,339 | null | """Anonymous marginal-program submission.
Mechanism
---------
The model is a differentiable probabilistic program. Its forward pass
computes, exactly, the answer-token predictions of a small structured latent
hypothesis space:
(prompt segmentation = which token types head the three fields)
x (field-role permutat... | {
"id": "12d188d0-160c-4fe0-aa27-64868a12d1bb",
"created_at": "2026-08-19 08:08:43.315161+00:00",
"db_md5": "8b0106e3231b48e7d506d0cc547758fe",
"submitter": "Ryan Teehan",
"github_login": "rteehas",
"run_id": "e5f6b5dc-1459-4cfe-afd1-67645623e179",
"tier": "easy",
"dataset_id": "e4",
"status": "succee... | {
"score": {
"mean_loss": 7.104169929743128,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0032291666666666666
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.090555864599254,
"example_count": 120... |
12d917d6-5307-46fd-a8a4-5c4bbe20a61f | easy | jordanrubin | 2026-08-08 09:55:19.421909+00:00 | succeeded | 078cd8887c504bdcfcf580d442de465beb5b3a0ea5e1726a6c95fd2267731c91 | 16,615 | null | """Bidirectional transformer, D=256 x 3 blocks, grokking-tuned AdamW.
Design notes:
- Inputs are padded to config.max_seq_len inside forward and logits sliced
back, so every training batch presents one static shape (compile/cudagraph
friendly, and uniform kernel shapes even in eager).
- Attention mask gets an iden... | {
"id": "12d917d6-5307-46fd-a8a4-5c4bbe20a61f",
"created_at": "2026-08-08 09:55:19.421909+00:00",
"db_md5": "1f66d0c9d8b2addfecdef9a9aa694dae",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "dc79110b-2244-4189-b4f2-b7d955c5def6",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 3.7842113275313096,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3016666586821278
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.572163371397366,
"example_count": 300,
... |
12da3ec3-c26b-449b-be5e-3ae86a1433d3 | easy | nikhilkumar26 | 2026-08-29 17:42:04.785057+00:00 | succeeded | 50975428a509589cd6356067a1887f47971842d9753bec5c0d1421b0a3de2e5d | 6,513 | null | """
One Layer Deeper Easy E1:
tiny one-hot MLP F on packed x/N digits, composed T times.
No attention. Residual MLP in digit-simplex features so 60s buys
hundreds of AdamW steps. No modular arithmetic in Python.
"""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Ten... | {
"id": "12da3ec3-c26b-449b-be5e-3ae86a1433d3",
"created_at": "2026-08-29 17:42:04.785057+00:00",
"db_md5": "e083973409b16f758575a237911dcf07",
"submitter": "Nikhil Kumar",
"github_login": "nikhilkumar26",
"run_id": "e3eae95b-0828-45af-9c9f-51fd0062a566",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.2194496393203735,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05166666768491268
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1530215740203857,
"example_count": 100... |
12e497dd-20ed-46e7-8b1b-b39e3f0e5908 | easy | erdavis0 | 2026-08-23 16:09:45.100709+00:00 | succeeded | d4e7b6f94c5a17f720147aa5b1ff3dedb9aca4341c22907ba39d97352562ee35 | 21,596 | null | """Learned confluent brickwork exchanges on field-aligned token tapes.
Opaque public token categories are routed into fixed right-aligned N, X, and T
tapes. Two learned-computation lanes then apply the same complete collection
of adjacent-place rewrites in opposite even/odd orders. One learned pair cell
is shared ov... | {
"id": "12e497dd-20ed-46e7-8b1b-b39e3f0e5908",
"created_at": "2026-08-23 16:09:45.100709+00:00",
"db_md5": "e9eede9164a2380d5b1e85c703eae702",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "7dea906a-1b8e-4d89-bf85-d9c3d6e4ea97",
"tier": "easy",
"dataset_id": "e9",
"status": "succeeded",... | {
"score": {
"mean_loss": 9.586919784545898,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.023333334363996983
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.441411972045898,
"example_count": 90,... |
12e6417a-d8e0-4f3c-8bfc-425b14d27fa3 | easy | ayushnangia | 2026-08-11 12:24:52.900602+00:00 | succeeded | 58f10e937bd402ba40387a2ff89ab4249e1216524cdbf6aefd442fb802cf7244 | 32,942 | null | """rB2j0c — rB2j0 + CALIBRATION PACK on the FINAL logits (run-65 ASI design:
merge width-safe binary compute with rC1-grade OOD calibration WITHOUT a
second pipeline). Two loss-side changes, no architecture change:
1. label_smoothing=0.1 on the main CE (rC1 lineage posts the only
hidden-positive OOD-N CEs 2.23-2... | {
"id": "12e6417a-d8e0-4f3c-8bfc-425b14d27fa3",
"created_at": "2026-08-11 12:24:52.900602+00:00",
"db_md5": "c414acf9d8a98482cf3f60f211a9e247",
"submitter": "Ayush Nangia",
"github_login": "ayushnangia",
"run_id": "f7c0a0ef-58b1-460a-9492-ce571780996b",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.331241726875305,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003750000149011612
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.338622570037842,
"example_count": 600,... |
12e6a6b3-f214-4e81-a3c2-4141cf2e5ec9 | easy | velocizapkar | 2026-08-04 06:14:39.768305+00:00 | succeeded | e63721f84a5f6e3bfb5e55c9422d17d85f76bf0e5f9acead17d5c95f40e1fafe | 42,650 | null | """Orthogonal depth lattice with recurrent Block AttnRes.
The learned state has three explicit axes: sequence/work position, parallel
computational slot, and recurrent reasoning time. The working transition is
a composition of Householder mixers along the first two axes and invertible
radial twists along features. A... | {
"id": "12e6a6b3-f214-4e81-a3c2-4141cf2e5ec9",
"created_at": "2026-08-04 06:14:39.768305+00:00",
"db_md5": "4f6beb9b70da3fff1c5dcdc206710a18",
"submitter": "Aakanksh Zarapkar",
"github_login": "velocizapkar",
"run_id": "a6a101a4-184d-40ee-b0ce-f4043cbdc29e",
"tier": "easy",
"dataset_id": "e1",
"statu... | {
"score": {
"mean_loss": 2.3473594188690186,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.029999999329447746
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.228482723236084,
"example_count": 100... |
12e7699f-1fec-4e27-abe3-1cb5d0038df0 | easy | gauravmishra | 2026-08-09 22:51:11.609767+00:00 | succeeded | ff53ecb548c348c010171474eac8b36fa1fa6f95d982ef1c93bd42644de95f1f | 6,079 | null | """Compound-scaled Transformer with distinct blocks and PyTorch AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_st... | {
"id": "12e7699f-1fec-4e27-abe3-1cb5d0038df0",
"created_at": "2026-08-09 22:51:11.609767+00:00",
"db_md5": "267f9774007ecc00f0abc489cb3ac9b1",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "8ab034c2-df98-47dd-8416-a893534592ac",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.280099279304602,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008750000024835268
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3036867535167747,
"example_count": 600... |
12eb07f6-7386-424d-adb9-d93385b60950 | easy | DDanlov | 2026-08-17 19:20:38.906878+00:00 | succeeded | 0a49445638931442794439efcaa0db9c171d7ac411a62f0d6689c76295738e52 | 23,585 | 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": "12eb07f6-7386-424d-adb9-d93385b60950",
"created_at": "2026-08-17 19:20:38.906878+00:00",
"db_md5": "8b95439c58befa747afc417feb02a4f5",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "92e2a227-88bd-4774-a87b-aa2c214af938",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 12.27399204841487,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.030000000235935048
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 21.379143572565336,
"example_count": 100... |
12ecb4bb-0b2a-4430-95da-263284739eb0 | easy | richardcepka | 2026-08-13 22:06:39.790508+00:00 | succeeded | b846a7250df6e324ee11027ff09c3eeaea89baa914a013fa1277935e917c6096 | 43,381 | null | """Recurrent digit-register model for One Layer Deeper.
A looped transformer whose recurrent state is a decimal digit register that is
re-quantised through a ten-entry codebook on every iteration. Nothing inside
the loop depends on the iteration index and no parameter is indexed by an
absolute slot position, so the e... | {
"id": "12ecb4bb-0b2a-4430-95da-263284739eb0",
"created_at": "2026-08-13 22:06:39.790508+00:00",
"db_md5": "ccf53c7b791dbfd728b518e099162c75",
"submitter": "Richard Cepka",
"github_login": "richardcepka",
"run_id": "96d120ce-4b27-436d-aa42-3bae5bd7be86",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.5857294378718274,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0041666666107873125
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5599441528320312,
"example_count": 6... |
12f2232a-8d4e-409f-b6ed-6fee6dcd2560 | easy | oupadhyay | 2026-08-15 21:47:46.927377+00:00 | succeeded | 5bd6522308cd90488d1cf9eac6fe58b94074aef11cf7fcd3da13445f8454fc37 | 3,988 | null | """Conditional optimizer screen: lr8."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,assert_model_state
W,D,H=4,64,32
class C:
def __init__(s,vocab_size,max_seq_len):s.vocab_size,s.... | {
"id": "12f2232a-8d4e-409f-b6ed-6fee6dcd2560",
"created_at": "2026-08-15 21:47:46.927377+00:00",
"db_md5": "5de2b802006f11ec1953bd7b19fb50fb",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "18880017-b39f-444d-8625-072d54b3416e",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.27839552474791,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0054166666790843015
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.193983713073046,
"example_count": 600,... |
12f917f8-fbfe-4f66-ae05-f76f88869467 | easy | ArkinDharawat | 2026-08-13 17:13:21.740753+00:00 | succeeded | 84d3d835b5c5b8ce8573535714c22a062da86856d82fea21980ac8601e0f23a5 | 7,629 | null | """Universal-Transformer-style looped block, v2: step embed + pos offset + LR schedule + 6 loops.
Extends ``looped_ut_padded`` with four ideas drawn from
Graves 2016 (ACT), Dehghani et al. 2018 (Universal Transformer), and
Merrill & Sabharwal 2025 (log-depth transformers):
1. **Learned per-step (timestep) embedding**... | {
"id": "12f917f8-fbfe-4f66-ae05-f76f88869467",
"created_at": "2026-08-13 17:13:21.740753+00:00",
"db_md5": "0f37b16e0eec248b73567ec6b95311f6",
"submitter": "Arkin Dharawat",
"github_login": "ArkinDharawat",
"run_id": "bcc37642-c5f9-4200-be4e-2092d710b14c",
"tier": "easy",
"dataset_id": "e3",
"status"... | {
"score": {
"mean_loss": 2.2781336414239552,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.015625000167638063
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.325836223767593,
"example_count": 800... |
12fbf75b-cd31-45b2-98d5-c29244052ad5 | easy | newjordan | 2026-08-15 23:33:00.006111+00:00 | succeeded | 6da6e937508b55c0269fa90d648775dd5beecc4c8d7f4934fa089f933655e8a8 | 178,654 | null | """Helix-Crawler — a recursive-weight-shared "crawler" ALU for modular squaring.
The task is r_{i+1} = r_i^2 mod n, chained T times: a *recurrence* over a single
operator. This build embodies the crawler / Frugendorff lineage (spark:~/sota_crawler)
mapped onto that structure. n never enters as a table row (unseen modu... | {
"id": "12fbf75b-cd31-45b2-98d5-c29244052ad5",
"created_at": "2026-08-15 23:33:00.006111+00:00",
"db_md5": "e6c32df21410d4316bc488cf172a8410",
"submitter": "Frosty40",
"github_login": "newjordan",
"run_id": "ac97c87a-62bf-494a-a27c-4d25139486e5",
"tier": "easy",
"dataset_id": "e8",
"status": "succeed... | {
"score": {
"mean_loss": 0.004153727786615491,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.00415878975763917,
"example_count": 85,
"e... |
1306cfb1-261f-456f-be61-330b704aacc9 | easy | DDanlov | 2026-08-14 07:41:37.900738+00:00 | succeeded | 314687dcdb9deb676221319ab127ee43c34db3b58b7ff4ec96ebc7c5ee4f4073 | 21,978 | 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": "1306cfb1-261f-456f-be61-330b704aacc9",
"created_at": "2026-08-14 07:41:37.900738+00:00",
"db_md5": "cfdc5a934fe79294618946f63ef2f0ad",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "e9bd4ad5-c413-4c9c-bf84-1d2809e33398",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.81734716697693,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8176316465230267,
"example_count": 600,
... |
1307d7d3-961b-4b25-8d66-44bde013af36 | easy | lzy54 | 2026-08-29 15:24:15.079913+00:00 | succeeded | c5991a9f358fe28a6dcea6def35faf85b4f3fbd5aee062d8d1ae097a80c09b7b | 17,742 | null | """R7c Time-Aligned Register Machine with low-LR refinement to 1200 steps.
The shared learned transition never sees T. T only routes which recurrent
state is decoded, aligning supervision at T=1/2/3 with one/two/three calls of
the same transition. The transition contains no hard-coded task algorithm.
"""
from __futur... | {
"id": "1307d7d3-961b-4b25-8d66-44bde013af36",
"created_at": "2026-08-29 15:24:15.079913+00:00",
"db_md5": "bc261b9a55ca1698223c3b0a0cc7ecd5",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "8a454164-aa13-404d-a79b-ff0a9be6547d",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.170947568233824,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.179505186765183,
"example_count": 600,... |
130985b7-4932-4493-9954-34664d6136e9 | easy | KaustubhKumar05 | 2026-08-26 08:16:09.804359+00:00 | succeeded | ceda9b5776851455052a3d53efcdfda03ec5aa5457d93786baa293d7e5b4ddb3 | 13,594 | null | """sair_perm: discrete exponent state + learned doubling permutation.
Motivated by Nanda et al's grokking result (AXRP ep.19, arXiv 2301.05217): the
grokked circuit for modular arithmetic works in the *character basis*, because
that is the basis where the group operation is trivial. Their task is modular
addition -> a... | {
"id": "130985b7-4932-4493-9954-34664d6136e9",
"created_at": "2026-08-26 08:16:09.804359+00:00",
"db_md5": "8be786446c33296c8af7cd6d5d2a25c8",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "9105537e-b464-4c5d-87ff-0fe4fd3eeac5",
"tier": "easy",
"dataset_id": "e8",
"status": "succee... | {
"score": {
"mean_loss": 1.9618640542030334,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05133689986541867
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.7988258600234985,
"example_count": 85,... |
1309ad44-1bc2-461b-a99f-8a43edc02860 | easy | sapient-sapiens | 2026-08-14 17:45:18.243830+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": "1309ad44-1bc2-461b-a99f-8a43edc02860",
"created_at": "2026-08-14 17:45:18.243830+00:00",
"db_md5": "1efadda9e01ae84ec2857432a666c542",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "80a32ea2-7c32-43b4-9067-614c8918afc5",
"tier": "easy",
"dataset_id": "e2",
"status": "su... | {
"score": {
"mean_loss": 2.6391657688537937,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.68104886233322,
"example_count": 300,... |
130dac45-38de-44f3-b02f-9e0fc239b4ad | easy | sapient-sapiens | 2026-08-23 10:09:41.192817+00:00 | succeeded | e3a42dd3344e560755f066722959f12cee3a5b0fe2d5f5b0e5b609e964fc5fd0 | 21,323 | null | """Approximately 10x-scale adaptive child of the D6 pair/quotient architecture.
The architecture uses the required learned additive token/group/left-relative
representation. Marker-derived field lengths additionally route categorical
digit embeddings into right-aligned lanes; this alternative place alignment is
an ex... | {
"id": "130dac45-38de-44f3-b02f-9e0fc239b4ad",
"created_at": "2026-08-23 10:09:41.192817+00:00",
"db_md5": "62845b66b5f88452d5ceda6aa9ab3f59",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "bfe5ee96-2e04-4563-81aa-407c5796c213",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.142812744681022,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.145741615327484,
"example_count": 600,... |
131b37a9-71c9-4cff-9111-b569d9bc4740 | easy | Koenneker | 2026-08-05 09:14:20.609921+00:00 | succeeded | 782cabac62246cfe9ac2fa798be45be80cc123cc3ab8298fe3d528cc9b196bab | 2,768 | null | """Vanilla single-layer LSTM with PyTorch AdamW."""
from __future__ import annotations
import torch
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 128
HIDDEN_SIZE = 256
NUM_LAYERS = 1
class Config:
... | {
"id": "131b37a9-71c9-4cff-9111-b569d9bc4740",
"created_at": "2026-08-05 09:14:20.609921+00:00",
"db_md5": "49ac6a3b8034b9b4b7fd8946e42672a8",
"submitter": "Ferdinand Könneker",
"github_login": "Koenneker",
"run_id": "390d13d3-fe1f-4687-a7fb-2d6d897f2ea2",
"tier": "easy",
"dataset_id": "e2",
"status"... | {
"score": {
"mean_loss": 4.677696228027344,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.017291666939854622
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.109512805938721,
"example_count": 300,... |
131edb6c-081f-42f5-a63f-2222e39a6192 | easy | karanganesan | 2026-08-08 04:31:10.594846+00:00 | succeeded | 8d9eecdc8607c91b190336e24b49597738decc69ce7d4e24ee5930ae97cfcad7 | 27,127 | null | """Parametric looped-transformer family (P1).
One weight-tied transformer block applied k times in latent space. Config
flags cover four P1 families with one file:
looped recall=0 gated=0 tfilm=0 plain weight-tied loop
looped-recall recall=1 re-inject the input embedding each
... | {
"id": "131edb6c-081f-42f5-a63f-2222e39a6192",
"created_at": "2026-08-08 04:31:10.594846+00:00",
"db_md5": "3e8439ba8ab561733e7dc883fab4ecb2",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "9ab0444a-b3fa-4972-a6fc-02184b32b8c1",
"tier": "easy",
"dataset_id": "e4",
"status": ... | {
"score": {
"mean_loss": 7.491694688796997,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0051041668630205095
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.8713154792785645,
"example_count": 12... |
1320a1de-3f81-4e2e-adaf-a42c3acfb252 | easy | ehonig | 2026-08-22 04:08:59.302261+00:00 | succeeded | 44ecd8da499c1559614d613737c668bfd0b20a76f20906717baa9072f4b6ae8a | 31,880 | 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": "1320a1de-3f81-4e2e-adaf-a42c3acfb252",
"created_at": "2026-08-22 04:08:59.302261+00:00",
"db_md5": "505e99fe53a2fa4bed673971201f22cc",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "5d0c8f0a-f150-4a56-a674-39f81591eda1",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 2.0848327589948172,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.08333333387970925
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.0822539169396928,
"example_count": 100... |
1325ae40-bbaa-4426-b249-4746d274ddc0 | easy | Yalyenea | 2026-08-12 08:26:14.871641+00:00 | succeeded | d4f760f76fa22b7a89407430bf27dc9855d53c3f30608c6c23f4ca6d1d31a4db | 14,474 | null | """Compact decimal recurrence with hyperspherical Muon transition updates."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
... | {
"id": "1325ae40-bbaa-4426-b249-4746d274ddc0",
"created_at": "2026-08-12 08:26:14.871641+00:00",
"db_md5": "fea6f4cd780789f7754f976b0a5e7d51",
"submitter": "yfff",
"github_login": "Yalyenea",
"run_id": "7e59ab15-a1b6-4bab-85cf-891937058170",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 6.7123223384309885,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016666666691501936
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.772961308603329,
"example_count": 600... |
132bcfbf-04f3-4844-adb7-8e8cabe134a4 | easy | viridale | 2026-08-13 09:27:25.592629+00:00 | succeeded | 812fb0e5e62b268fc0ad7c99ab07a19deb1e11ba2dc03245747191f38b4a5052 | 21,622 | null | """Affine-monoid ordered coefficient reducer.
hypothesis: P90 proved ordered coefficient context useful (+2.50pp) but full stochastic
matrices limited the learned state to four controls. Replace only that coefficient
reducer by 32 learned affine channels with exact associative ordered composition.
Preserve... | {
"id": "132bcfbf-04f3-4844-adb7-8e8cabe134a4",
"created_at": "2026-08-13 09:27:25.592629+00:00",
"db_md5": "99fb648ca1f0bc2272fc50ef21da54f3",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "6ef9600d-477f-4962-8804-b37fe7cd0511",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 1.836188554763794,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09666666761040688
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.7684522867202759,
"example_count": 100,... |
132f1f3a-f2ad-424b-8b3a-34aed06af945 | easy | shirvani-jr | 2026-08-28 08:04:29.186260+00:00 | succeeded | 37eece96787bce15ba2581c2a7b13bb60b8532374de42ccb9852e1059f329c6c | 40,175 | 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": "132f1f3a-f2ad-424b-8b3a-34aed06af945",
"created_at": "2026-08-28 08:04:29.186260+00:00",
"db_md5": "68803d223f9f2bc2706130c7d24b9cee",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "7d0f73a7-7bb3-44b6-aa67-7e81c1cbf747",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.2011911869049072,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06833333149552345
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1801674365997314,
"example_count": 100... |
133431c8-4631-4caa-8d25-40516d882c39 | easy | AFeinfield | 2026-08-27 22:31:09.800854+00:00 | failed | 1f6bf221b0fef0093c4cc5a147e1005ae5166d460d8e1e890ee2a5678ef445ed | 13,573 | null | """159: 158 + tanh-bounded linear features (kills the init logit
explosion without letting the head shrink back into the marginals basin).
155/156/157 all converged to the same shallow attractor (loss ~1.87,
T=1 rung ~0, deep-T rungs inflated by orbit fixed points). Fixes, all
generic (random weights, no task values):... | {
"id": "133431c8-4631-4caa-8d25-40516d882c39",
"created_at": "2026-08-27 22:31:09.800854+00:00",
"db_md5": "b0f4318fa6f52073f17f5e55eb33e1f8",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "6e0fe967-24be-4125-81f3-cbf76b9c0e68",
"tier": "easy",
"dataset_id": "e5",
"status": "fail... | null
|
133daa78-ec6a-4eb0-8c96-3ce40d7e1413 | easy | alirezashirvani-jr | 2026-08-31 07:45:41.987494+00:00 | succeeded | c252ce15d535434ae9695254770a32b96e97a7120717f3216a16792d4ff23118 | 34,903 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
TRACK_FLOOR = 3
TRACK_CEILING = 96
STAGE_B... | {
"id": "133daa78-ec6a-4eb0-8c96-3ce40d7e1413",
"created_at": "2026-08-31 07:45:41.987494+00:00",
"db_md5": "b9699e2f9827d45ad6df7d21eed39459",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "9216d0fd-1de7-4ed1-a598-613401829a79",
"tier": "easy",
"dataset_id": "e5",
... | {
"score": {
"mean_loss": 0.26579435491464565,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.9600000174840291
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.24492527544498444,
"example_count": 60... |
133e6e19-c6ea-4d11-b392-30e7f76c0819 | easy | khushidahi | 2026-08-03 03:00:12.583996+00:00 | succeeded | 5a50221af8c6adb33f8233f87fc4318028ce191419f93aaeb24f63e15e7dfc67 | 15,252 | 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": "133e6e19-c6ea-4d11-b392-30e7f76c0819",
"created_at": "2026-08-03 03:00:12.583996+00:00",
"db_md5": "1b2aae8823575d3bbe0292cc8e96bd6f",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "b76f488c-4080-4e1a-802b-363745e94470",
"tier": "easy",
"dataset_id": "e4",
"status": "succ... | {
"score": {
"mean_loss": 2.0772018432617188,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005000000121071935
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.074612617492676,
"example_count": 120... |
133fe3b1-51e2-4388-8f52-8c3e99851974 | easy | kushasareen | 2026-08-27 03:55:35.135352+00:00 | succeeded | 263cc782517855643e223a21f802fa9eb738cf5d4bc2a9e2926951d225b056ce | 10,895 | null | """loop3_T: digitcomp (4 blocks) where blocks 2..4 repeat T times; T token kept.
At T = 1 this is exactly the digitcomp model. Muon recipe, 3 updates per batch.
Each number's token is a learned function of its digits: every place has
its own embedding table, the place embeddings are summed and passed
through a small ... | {
"id": "133fe3b1-51e2-4388-8f52-8c3e99851974",
"created_at": "2026-08-27 03:55:35.135352+00:00",
"db_md5": "efd2eec7a81af24ab22a220b46b11a6e",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "2c1a9d81-9b30-48e4-aa42-5bbc5e6783c4",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 2.9538499116897583,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.40166669711470604
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.332817077636719,
"example_count": 300,... |
13449885-20dd-4538-af2a-2a3ff89dc9a6 | easy | DDanlov | 2026-08-20 13:30:56.229061+00:00 | succeeded | 2f0b880ecb090b19480c68af0d9012a529771b4c133991de96553fd1de274bab | 20,961 | 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": "13449885-20dd-4538-af2a-2a3ff89dc9a6",
"created_at": "2026-08-20 13:30:56.229061+00:00",
"db_md5": "9756fc3c393dc1c98eb674130985fbef",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "9e0280d5-e107-410a-bb65-a03ef59f2bf1",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.671345307303935,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6725234589769165,
"example_count": 600,
... |
13470bc5-8f40-4372-84e3-f6eac77092d3 | easy | alirezashirvani-jr | 2026-08-31 20:53:51.386537+00:00 | succeeded | e1f746e894d9847b057aa22ec2998990d48e6303fa538d116a8257a5eb16943b | 31,253 | null | from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
PERIOD_MINIMUM = 3
PERIOD_BUDGET_CEILING =... | {
"id": "13470bc5-8f40-4372-84e3-f6eac77092d3",
"created_at": "2026-08-31 20:53:51.386537+00:00",
"db_md5": "73e122660a53432d7215ebe9bca1092a",
"submitter": "alirezashirvani-jr",
"github_login": "alirezashirvani-jr",
"run_id": "36603690-8099-427a-b156-b865de921be8",
"tier": "easy",
"dataset_id": "e5",
... | {
"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": 600,
"exact_accuracy": 1.0,
"... |
1349931f-6dbd-456c-a835-1b912565dce1 | easy | lzy54 | 2026-08-27 14:44:19.355993+00:00 | succeeded | 4dd7244be31611e940329b94e666fafdbe79f97691609aba1440365e1a0096e9 | 3,989 | null | """Phase 1 A4 additive-recall control capped at 120 optimizer updates."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_MODEL = 1... | {
"id": "1349931f-6dbd-456c-a835-1b912565dce1",
"created_at": "2026-08-27 14:44:19.355993+00:00",
"db_md5": "a1e65927d7f516336d33c509d255f576",
"submitter": "lzy54",
"github_login": "lzy54",
"run_id": "d8d65bc3-868d-45e7-8ef1-eae21a2d3052",
"tier": "easy",
"dataset_id": "e6",
"status": "succeeded",
... | {
"score": {
"mean_loss": 2.497430443763733,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04549549613147974
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.4451756477355957,
"example_count": 60,
... |
13545a11-aaae-438a-b751-b7be4e50528b | easy | ehonig | 2026-08-20 04:20:33.623060+00:00 | succeeded | b81ebb68535040625243d2fa04e707d077548243a60534c9077b6a389822cd9e | 22,765 | 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": "13545a11-aaae-438a-b751-b7be4e50528b",
"created_at": "2026-08-20 04:20:33.623060+00:00",
"db_md5": "f28e9e945881c646a5e0c22c5ebd3cb5",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "053431df-2fe9-4ee7-a375-539201bc5b4f",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 1.8377404808998108,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07999999821186066
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.7795186042785645,
"example_count": 100... |
13546b42-97ed-4e42-a7c4-76ac1ad8c98e | easy | gauravmishra | 2026-08-09 22:50:13.076613+00:00 | succeeded | c94c270720868cfb2847fb9284ea0944440ab57d07870981010ef9012376cd2f | 4,268 | null | """Basic single-pass Transformer with PyTorch AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
D_MODEL = ... | {
"id": "13546b42-97ed-4e42-a7c4-76ac1ad8c98e",
"created_at": "2026-08-09 22:50:13.076613+00:00",
"db_md5": "d7d14efc3f9d1186e18e037d4020d52e",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "67549ca0-be4e-4c0c-9ff0-475f1e727c27",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.2608827477550464,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.004583333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3101745913381535,
"example_count": 60... |
1357dc0e-b3c1-4d0e-a533-bc6f275eb6b2 | easy | Bananafly | 2026-08-11 18:30:58.156939+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": "1357dc0e-b3c1-4d0e-a533-bc6f275eb6b2",
"created_at": "2026-08-11 18:30:58.156939+00:00",
"db_md5": "eaac69fe345f82973fcf392c504a40ff",
"submitter": "Andre Kreidemann",
"github_login": "Bananafly",
"run_id": "55574545-fec8-4abd-9273-ed139ede90d8",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 3.0678900480270386,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.0815019607543945,
"example_count": 100,
"exa... |
137886ee-e1ba-4c8a-9919-a95c0e0f746f | easy | yashkant | 2026-08-13 00:49:16.430978+00:00 | succeeded | 466f683b25d2da583b4b9868e20cdc4cec959def16df941b293422051ebebdfe | 31,388 | null | """Round1702 one-projection-per-T Dykstra recursion on the D416 carrier.
Each parsed public-T cycle applies one shared learned smooth projection and
alternates the A/B set code and correction memory across cycles. The immutable
anchor is reinjected on every call; all active cycle outputs are averaged, and
terminal CE... | {
"id": "137886ee-e1ba-4c8a-9919-a95c0e0f746f",
"created_at": "2026-08-13 00:49:16.430978+00:00",
"db_md5": "ee1f58268ff454a206b669e699e09923",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "7343f1bf-671d-4f26-90f9-0aa653a34b91",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 2.480572462081909,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.050208333879709244
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.852391004562378,
"example_count": 300,... |
137891b3-59cf-43cc-b1f6-f9cdef781182 | easy | liam-gb | 2026-08-11 05:03:10.357182+00:00 | succeeded | faad3f09af3019fc3b514e68d896ac7d78883bad78aa618571903cbf0734f5f3 | 4,873 | null | """Bidirectional transformer baseline: 4/6 blocks, SwiGLU MLP, AdamW."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (ModelSpec, OptimizerBundle, OptimizerSpec, Submission,
TokenLossBatch, assert_model_state)... | {
"id": "137891b3-59cf-43cc-b1f6-f9cdef781182",
"created_at": "2026-08-11 05:03:10.357182+00:00",
"db_md5": "32868da38aad198587c8bf284316cce3",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "c25b88bb-c0cc-4bfd-903e-5d2b69e2f218",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 6.659844398498535,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.038333332166075706
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.764042377471924,
"example_count": 100,... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.