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 |
|---|---|---|---|---|---|---|---|---|---|---|
29005e94-4168-4cfb-ad95-f3aeef997363 | easy | DDanlov | 2026-08-11 19:39:32.537236+00:00 | succeeded | 5b2e41ad67960d83a1b10ed1634f8322c81605e896c8ee6898f4b5175b7b0fb2 | 20,031 | 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": "29005e94-4168-4cfb-ad95-f3aeef997363",
"created_at": "2026-08-11 19:39:32.537236+00:00",
"db_md5": "b084a9617d38aedad9b688b34ca1e78c",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "7fe4d01a-7fad-4dd4-90dc-1c20838adba3",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 6.403242123691816,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013333333109815916
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.225373268127441,
"example_count": 100,... |
290cbedf-700d-4c8a-88b4-cb8dd544032a | easy | yashkant | 2026-08-11 17:23:37.022982+00:00 | succeeded | 2648c32055f395a2638e0046d761e9fb2abc2e087a3a812f7890c7a7f9fe50c5 | 34,476 | 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": "290cbedf-700d-4c8a-88b4-cb8dd544032a",
"created_at": "2026-08-11 17:23:37.022982+00:00",
"db_md5": "8bc9ff1c5fd9cc3d5e6762aa1043f6d0",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "250ccd70-7cb2-4b83-b27d-d3e64827d1fe",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 3.227717638015747,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09333333559334278
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.348257541656494,
"example_count": 300,
... |
29159501-e27c-4c53-9e12-8d628c60b72d | easy | shreyash-chonkie | 2026-08-24 06:55:18.627809+00:00 | failed | 662ad40455c610767c88241ec48333055487b74d16b98856697486919cd59838 | 10,838 | null | """Gated recurrent attention with recurrence-relative feedback."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_mo... | {
"id": "29159501-e27c-4c53-9e12-8d628c60b72d",
"created_at": "2026-08-24 06:55:18.627809+00:00",
"db_md5": "47fe39b3ed05a46394f16584a2f12fdf",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "95f5393b-a461-4557-b81a-ae6988fe617b",
"tier": "easy",
"dataset_id": "e6",
"status": "... | null
|
2918719c-0b54-4a49-b1b4-1480588e89f2 | easy | repst | 2026-08-26 18:41:44.499781+00:00 | succeeded | 8144ae1604cf4242599d6f313a674abbb9d780ec0d798a337849b3e27e3ac83b | 150,213 | null | # R749: semantic recurrent program induction with posterior particles.
#
# The model learns one shared SETUP/STEP/READOUT recurrence from complete
# answer-token likelihood. Its generated semantic vocabulary is analogous to
# an LM's custom tokenizer/vocabulary; factorized adjacent-interface energy is
# analogous to st... | {
"id": "2918719c-0b54-4a49-b1b4-1480588e89f2",
"created_at": "2026-08-26 18:41:44.499781+00:00",
"db_md5": "5872086d090313693ae9245b722b4d7a",
"submitter": "Ryan Epstein",
"github_login": "repst",
"run_id": "267a4e8d-44d5-4223-8dce-94a8d8fa6baa",
"tier": "easy",
"dataset_id": "e1",
"status": "succeed... | {
"score": {
"mean_loss": 0.019660880515105707,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.019654951989650726,
"example_count": 100,
... |
291e05b9-fc23-4784-bf3d-8de3ff7bdbb8 | easy | sapient-sapiens | 2026-08-19 17:56:06.720442+00:00 | failed | d932de560e11b5d257db118ae44a7ec095fdaed64fd1422d9b68fa19dae25efc | 15,396 | 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": "291e05b9-fc23-4784-bf3d-8de3ff7bdbb8",
"created_at": "2026-08-19 17:56:06.720442+00:00",
"db_md5": "1f0f7ed6c69840b4e2d1050ccb025455",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "26d218f4-a0ac-4e92-a7bb-4b9c771ce063",
"tier": "easy",
"dataset_id": "e5",
"status": "fa... | null
|
2921962f-6563-4709-b271-c895808ddae2 | easy | sapient-sapiens | 2026-09-01 01:10:15.667670+00:00 | succeeded | bab7a64ed720852b2980342c175e42e25cd96ad7f6c38de8fa9fabad1f049655 | 24,837 | null | """E5 transfer of the current D6 shared-character residual nested square."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,... | {
"id": "2921962f-6563-4709-b271-c895808ddae2",
"created_at": "2026-09-01 01:10:15.667670+00:00",
"db_md5": "467419dfa91036f7ede8452e93f81292",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "7a3a1acf-154d-467b-a61a-8aa0c59553b5",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.322697207696315,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0004166666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3224629481811694,
"example_count": 60... |
293557e5-ec1e-4166-a98d-6a6377443aec | easy | himalalps | 2026-08-07 11:43:28.671856+00:00 | succeeded | 9f2c798d1caffc0bbcdeeeda8f2cf13bc0833c0319708c4c6addd49beed4a01a | 18,392 | null | """Compliant learned recurrence for One Layer Deeper.
N and x are encoded independently into a 768-wide immutable context and
initial state. T is parsed only as loop control. A randomly initialized shared
transition receives current state plus fixed N context and is applied exactly
4*T times. Generic multiplicative an... | {
"id": "293557e5-ec1e-4166-a98d-6a6377443aec",
"created_at": "2026-08-07 11:43:28.671856+00:00",
"db_md5": "99a92d8ddf7930cf5c51e19eb63ca4bd",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "8fcfd50f-b7ec-4926-a8ee-538413881e58",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 2.1993039477918597,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666896392902
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2372405529022217,
"example_count": 60... |
29365c41-ff7f-4186-b802-b941a0ecaa13 | easy | sapient-sapiens | 2026-08-22 18:05:16.808411+00:00 | succeeded | 043c17c9208d5dd3658c13d0814cb276ce7b1e3ac4cfd3a7e0726225bc472c3a | 19,724 | null | """Eight-layer T-private Q-gated TRM with HybridMuon and SAM for E5.
Research hypothesis: eight width-1440 blocks provide a useful depth/throughput
midpoint for testing multiplicative attention gates on E5.
Primary experimental variable: block depth/scale, six blocks versus eight,
establishing the common Q-gated base... | {
"id": "29365c41-ff7f-4186-b802-b941a0ecaa13",
"created_at": "2026-08-22 18:05:16.808411+00:00",
"db_md5": "32df2ccfb8e90dcf7a12d028197c1ec3",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "1a7ce230-e2db-4de5-9687-369c6b4f626b",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.3270629231918063,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.002916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3311126832202946,
"example_count": 60... |
293a138c-08c8-451e-8b5b-93cf99643863 | easy | yashkant | 2026-08-12 04:07:03.280955+00:00 | succeeded | 68938a5bc3a39e8289960f6dd726536195fdf9c8d13a0e22c9e1304c8684d1dd | 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": "293a138c-08c8-451e-8b5b-93cf99643863",
"created_at": "2026-08-12 04:07:03.280955+00:00",
"db_md5": "a9eaf9c9fa3858fac12d940ff5a25d6d",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "f1d231fc-4719-48b4-99c6-15c5321017da",
"tier": "easy",
"dataset_id": "e5",
"status": "succeed... | {
"score": {
"mean_loss": 2.145134712431296,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666679084302
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1503196040611097,
"example_count": 600... |
293ef92d-86da-4067-a6d6-0331c09f4b0a | easy | TZZheng | 2026-08-06 17:34:38.050585+00:00 | succeeded | 0b4436e75af4f4a676c5df128139f2faeb767be97b6bf90b2d7197ca397d8c85 | 7,296 | null | """Self-contained standard Transformer scaling-sweep submission."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
VARIANT = {'block... | {
"id": "293ef92d-86da-4067-a6d6-0331c09f4b0a",
"created_at": "2026-08-06 17:34:38.050585+00:00",
"db_md5": "bd6ebca38a9e59e4c336966e9ce57e84",
"submitter": "TZZheng",
"github_login": "TZZheng",
"run_id": "b618e146-b400-4f18-8688-2275dd104bf9",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.4853938817977905,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.01666666753590107
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.80642032623291,
"example_count": 100,
... |
29408296-d60c-455b-90fb-0feb30e144ee | easy | ehonig | 2026-08-19 21:51:10.889128+00:00 | succeeded | 6d735b5caf3e04329ec379df05dfb21cfd33363a98fe085cf93ccec6de260875 | 19,987 | 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": "29408296-d60c-455b-90fb-0feb30e144ee",
"created_at": "2026-08-19 21:51:10.889128+00:00",
"db_md5": "e4c1dbaa61ce23d760278e5a94134a02",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "19521f90-94f8-47e4-9cb8-647204beb487",
"tier": "easy",
"dataset_id": "e3",
"status": "succ... | {
"score": {
"mean_loss": 2.364461302757263,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005000000004656613
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5037155151367188,
"example_count": 800... |
2940ee66-faab-4f61-9e39-e1bf4d863f51 | easy | karanganesan | 2026-08-06 05:53:43.800961+00:00 | succeeded | 08e24391a371e3eb49f6cba6259764b5fb5da11951ef5a35aff371a58fd2261d | 25,600 | null | """Parametric looped-transformer family (P1).
One weight-tied transformer block applied k times in latent space. Config
flags cover four P1 families with one file:
looped recall=0 gated=0 tfilm=0 plain weight-tied loop
looped-recall recall=1 re-inject the input embedding each
... | {
"id": "2940ee66-faab-4f61-9e39-e1bf4d863f51",
"created_at": "2026-08-06 05:53:43.800961+00:00",
"db_md5": "909e0c360baba0d3ed79fb6d3112c3bd",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "efc73c32-a071-4590-9918-580a021e4980",
"tier": "easy",
"dataset_id": "e3",
"status": ... | {
"score": {
"mean_loss": 7.024737358093262,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007499999832361937
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.026266574859619,
"example_count": 800,... |
294773a9-a75e-4f67-83cc-a224520f4bce | easy | kushasareen | 2026-08-10 22:43:41.255835+00:00 | succeeded | 770fe8df73454f4f9ae952ca3f5b6a825489e24637705b46ae38e204660e9a50 | 6,085 | null | """T-conditioned recurrent Transformer.
A tied-weight block is applied once per requested squaring step, so compute
scales with T and the model can extrapolate to depths never seen in training.
Untied pre-blocks encode the prompt and an untied post-block decodes the
answer digits.
"""
from __future__ import annotatio... | {
"id": "294773a9-a75e-4f67-83cc-a224520f4bce",
"created_at": "2026-08-10 22:43:41.255835+00:00",
"db_md5": "4f6635101dba30b57ab6592ef43c0245",
"submitter": "Kusha Sareen",
"github_login": "kushasareen",
"run_id": "a1b8e3a6-b903-4a0a-9cd6-ed40288ae72e",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 4.327148047853718,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006250000024835269
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.938549146523925,
"example_count": 600,... |
2948c710-76c2-4b55-a673-d5c9b7237879 | easy | DDanlov | 2026-08-31 20:12:42.284895+00:00 | succeeded | bd92f95a85627603216622fd27cff208444824f5ee00b0f45ddc414f313b39ae | 32,408 | null | """
Unified Continuous Hopfield Equilibrium Propagation Deliberation Engine
=======================================================================
Architecture:
- Deliberation: Continuous Hopfield Multi-Layer Energy System (L=4, D=768, d_ffn=2304, H=16)
- Integration: In-Place Strang-Split Conformal Symplectic Velocit... | {
"id": "2948c710-76c2-4b55-a673-d5c9b7237879",
"created_at": "2026-08-31 20:12:42.284895+00:00",
"db_md5": "7e875b7d8defb18c3dd7d548f5fbf534",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "9cce8d5d-c226-43a9-9db5-d0e484c7fca4",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.9118707742570176,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.035000000447034836
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.977464620746783,
"example_count": 100... |
29492386-96c5-4aca-a3fc-fb4049e78bac | easy | isaac0804 | 2026-08-07 22:20:49.206334+00:00 | succeeded | 00499b60ddc6ed8ed8ea6c7440681180708a5827388073d8acf03506f359f472 | 12,387 | null | """E3 candidate: optimizer-side gradient-agreement gating across modulus
subgroups, via bounded multi-pass backward (rule 4).
Every mechanism tried on E3 so far (see docs/findings.md) operates on the
model's forward pass or loss shape: routing, bottlenecks, contrastive
regularizers, curriculum. This one operates purel... | {
"id": "29492386-96c5-4aca-a3fc-fb4049e78bac",
"created_at": "2026-08-07 22:20:49.206334+00:00",
"db_md5": "a7ef578c4fba4c4cb8095a62c8f51633",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "0a7785e0-1364-44aa-84b8-2fce6460e865",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 9.827286761112116,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010000000013969838
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.255999706751366,
"example_count": 800... |
29584acb-f083-45b7-84ce-0a0d746ba738 | easy | DDanlov | 2026-08-19 21:18:33.121936+00:00 | failed | 8ccec27d49ba13776797ae2468c675e75793824b2986abcffcaaabc111923726 | 17,727 | null | """
Dynamic Submission for One Layer Deeper API Benchmark
Config: dim=640, num_heads=32, d_ff_mult=4, fixed_trec=3000
"""
from __future__ import annotations
import math
from typing import Any, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
... | {
"id": "29584acb-f083-45b7-84ce-0a0d746ba738",
"created_at": "2026-08-19 21:18:33.121936+00:00",
"db_md5": "1d0435e70f7c6e88cf7da0d65d0f84ac",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "1d37908e-6031-4401-b7bb-445e1d25bce7",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
295e0bb9-68de-4b1b-9909-c9299e5b74c5 | easy | sapient-sapiens | 2026-08-20 20:31:33.702524+00:00 | succeeded | beb38f344d33593e1aec188645805436061a171703440343602ebf4143bb294c | 38,555 | null | """Hierarchical learned digit-mixing recurrence with an answer-only objective.
The recurrent state is a bank of right-aligned digit slots. Every operator
application is followed by a soft quantization back onto the token simplex, and
the digit logits produced by that quantization are the answer logits at that
depth. T... | {
"id": "295e0bb9-68de-4b1b-9909-c9299e5b74c5",
"created_at": "2026-08-20 20:31:33.702524+00:00",
"db_md5": "f5f68bb42314937941b2c50c4fbd97df",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "5e997037-cac3-4748-9dba-a8aced5cf4cc",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.183972590633749,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333345750967
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.189042692360857,
"example_count": 600,... |
29607821-7b95-4104-bfbf-484092b2a998 | easy | Alan-Qin | 2026-08-26 14:06:03.289947+00:00 | succeeded | 850268f2eb903481d890f47540462e8b0f962c46dacdd82695ef3fbd33812677 | 24,977 | null | """v91h17 = v91 with LR-cosine horizon 1700 (Medium tiers are 600s hosted ~1.7k steps, NOTES #172; cosine must land inside the real budget).
v91 = v63lf + w-hat auxiliary head on the proven gated MLP (route B v2:
same readout, added CRT-quotient supervision; measures whether the MLP can
learn the universal quotient in... | {
"id": "29607821-7b95-4104-bfbf-484092b2a998",
"created_at": "2026-08-26 14:06:03.289947+00:00",
"db_md5": "eb439a22cbe965b79c1780dd0ee6dc34",
"submitter": "Zeyu Qin",
"github_login": "Alan-Qin",
"run_id": "3f65636d-8981-4089-9b5d-5ae043180d4d",
"tier": "easy",
"dataset_id": "e8",
"status": "succeede... | {
"score": {
"mean_loss": 1.7340237548069126e-05,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.6267418914139853e-06,
"example_count": 85,
... |
2964e061-e759-4d1a-929d-702f3a04a298 | easy | jordanrubin | 2026-08-13 00:25:47.990484+00:00 | succeeded | ab34a11198c302dd2edc317b1b463ccffb20bd9019d78abed718f3a88cc5d938 | 24,918 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "2964e061-e759-4d1a-929d-702f3a04a298",
"created_at": "2026-08-13 00:25:47.990484+00:00",
"db_md5": "19d38540224b58e35ee3887ac55c96ea",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "c71f61fe-d1f6-4f92-9a45-0550ccb72ebf",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.8571624853550985,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02874999896933635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.211991383355829,
"example_count": 600,... |
29656308-7785-42fc-be28-25e9bc3682aa | easy | pculaf | 2026-08-25 14:02:56.623075+00:00 | succeeded | 597baf57582cf77e3dc7f2c91edc50c3075aadb145afd41399e0e44b7759c3c8 | 8,287 | null | """Four-loop post-RMSNorm Transformer for One Layer Deeper."""
from __future__ import annotations
from collections.abc import Iterable
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_m... | {
"id": "29656308-7785-42fc-be28-25e9bc3682aa",
"created_at": "2026-08-25 14:02:56.623075+00:00",
"db_md5": "b018b35b15360a396a53c7a098f345e9",
"submitter": "Pavle Culafic",
"github_login": "pculaf",
"run_id": "f3fcfbc7-3c89-4a4d-9641-6d11ab63258c",
"tier": "easy",
"dataset_id": "e2",
"status": "succe... | {
"score": {
"mean_loss": 2.235713243484497,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016041667200624943
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.239830255508423,
"example_count": 300,... |
2969c952-574c-4c92-afe1-ebfbc71cfcde | easy | amirmeisami | 2026-08-17 23:21:54.183134+00:00 | succeeded | d5af187e692784669699ad415cf766f8cfd4ac8dccdf9347ce0ba3447514c482 | 12,668 | null | """T023 P1: D with learned low-base digit re-embedding."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, TokenLossBatch, assert_model_state
D_MODEL, POLY_WIDTH, INNER... | {
"id": "2969c952-574c-4c92-afe1-ebfbc71cfcde",
"created_at": "2026-08-17 23:21:54.183134+00:00",
"db_md5": "d04d668e1f08adb3af3ae13bb306bb60",
"submitter": "Amir Meisami",
"github_login": "amirmeisami",
"run_id": "93817984-7606-4bdf-ac5b-64bdd8b5206c",
"tier": "easy",
"dataset_id": "e3",
"status": "s... | {
"score": {
"mean_loss": 5.448620092817599,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.517754119637136,
"example_count": 800,
"... |
296b1403-40b4-4025-9bc1-182b9810957b | easy | sapient-sapiens | 2026-08-20 19:54:56.129957+00:00 | succeeded | 50c9a3a44f4cfa8190f5a4da34841d47e945f78848cadc4c308810f468c297a9 | 37,871 | null | """Hierarchical learned digit-mixing recurrence with an answer-only objective.
The recurrent state is a bank of right-aligned digit slots. Every operator
application is followed by a soft quantization back onto the token simplex, and
the digit logits produced by that quantization are the answer logits at that
depth. T... | {
"id": "296b1403-40b4-4025-9bc1-182b9810957b",
"created_at": "2026-08-20 19:54:56.129957+00:00",
"db_md5": "6865655718b542458d9cdb2564f8eb04",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "b6b549fe-3d03-4c7c-a9b0-6d886a2c01ef",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.1702931107194603,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666716337205
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1685718284594104,
"example_count": 60... |
297ec513-28d7-4758-870f-8b6cab6a0704 | easy | sapient-sapiens | 2026-08-09 21:32:13.599414+00:00 | succeeded | 5bb4e61f0c3687019fbff8d90d6b56ec8467719a08ba00c1fe0e69800328728e | 15,567 | null | """T-independent categorical recurrence with learned latent-depth selection."""
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,
Toke... | {
"id": "297ec513-28d7-4758-870f-8b6cab6a0704",
"created_at": "2026-08-09 21:32:13.599414+00:00",
"db_md5": "d5518602a2c831b1a0df7d2053fd26b5",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "f7a49a3c-a70a-4f14-90e1-7cc4c993861a",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.204759449947068,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333358168603
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2083328042864263,
"example_count": 600... |
2984ad67-a3f6-4d26-8e14-22c34f8c2eef | easy | chad-atexpedient | 2026-08-27 03:52:54.771411+00:00 | succeeded | d2768787efd31fbacf8ddbc449ac3678446e5f38966fc5a83b3570b4603f5e67 | 19,849 | null | """X138: multiplicative value/modulus coupling in the transition.
Every prior candidate fed the modulus to the transition only additively via
concatenation, so the network never formed a product of the value state with
itself -- the defining structure of squaring. X138 keeps X135's fixed-T-active
denoising and adds a ... | {
"id": "2984ad67-a3f6-4d26-8e14-22c34f8c2eef",
"created_at": "2026-08-27 03:52:54.771411+00:00",
"db_md5": "62b49f49b74aea3f2a58bec02d7d6999",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "68fc334f-9e3d-460d-9d99-c28313b2cea9",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.2662579451443037,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009583333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.306994140950019,
"example_count": 600... |
29944f9c-90e8-473a-8640-401c48100d04 | easy | akkolli | 2026-08-04 03:45:30.472998+00:00 | succeeded | a3237b12d4e3bef7be51a9fa79eeb0deb103fa95fde5e09f0490450263200905 | 17,821 | null | """Dual-format CE-trained bank of exact modular recurrence programs.
The model supports separate-output prompts and causal ANS/result/EOS records.
Answer-token values never enter parsing or recurrence state; causal routing uses
only public marker positions. The random selector is trained by token CE only.
"""
from _... | {
"id": "29944f9c-90e8-473a-8640-401c48100d04",
"created_at": "2026-08-04 03:45:30.472998+00:00",
"db_md5": "48c15d97f9be79619a33c6cc2aafca7a",
"submitter": "Akshay Kolli",
"github_login": "akkolli",
"run_id": "cf14fae0-7f14-454a-87c7-72457b32b769",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 0.0,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 1.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 0.0,
"example_count": 800,
"exact_accuracy": 1.0,
"... |
2994d4a3-08a9-4b82-bece-5b4c0c54eacd | easy | DDanlov | 2026-08-14 20:02:47.296768+00:00 | succeeded | dc7c6b244b3002d4a76be6c5b0fdac3ab188eeb92af808d074e46bf79deeba4a | 21,975 | null | import math
from typing import Optional
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.optim import Optimizer
try:
from benchmark import Submission, assert_model_state, OptimizerBundle
except ImportError:
try:
from client import Submission, assert_model_state, OptimizerBu... | {
"id": "2994d4a3-08a9-4b82-bece-5b4c0c54eacd",
"created_at": "2026-08-14 20:02:47.296768+00:00",
"db_md5": "a0af71418a941ca6ec8685f75dcf3be5",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "901a5122-c4b1-447e-805c-7ad461330c55",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.284205198753197,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.018125000000000002
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.2834231141338743,
"example_count": 800... |
29956e34-28c8-4812-a496-d86fbcafa514 | easy | karanganesan | 2026-08-08 03:41:59.596960+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": "29956e34-28c8-4812-a496-d86fbcafa514",
"created_at": "2026-08-08 03:41:59.596960+00:00",
"db_md5": "3e8439ba8ab561733e7dc883fab4ecb2",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "08568309-b666-4157-89df-65ae48513207",
"tier": "easy",
"dataset_id": "e3",
"status": ... | {
"score": {
"mean_loss": 9.338274478912354,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006874999729916453
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.783576965332031,
"example_count": 800,... |
2997f358-4f7b-43e3-8287-a8d66d4ff9e0 | easy | chad-atexpedient | 2026-08-14 16:17:24.500719+00:00 | succeeded | c1f2288c723e801cab4b5470c5e618bbc79c5d2d22141de43f3c02029fa0e4d6 | 17,461 | null | """X35: Cross-N delta consistency - match deltas across different N values.
Matches state deltas between examples with DIFFERENT N values when their
intermediate values are the same. This directly encourages N-invariant
representations for OOD-N transfer.
"""
from __future__ import annotations
import torch
import to... | {
"id": "2997f358-4f7b-43e3-8287-a8d66d4ff9e0",
"created_at": "2026-08-14 16:17:24.500719+00:00",
"db_md5": "df85aa74b5ee3db326965c77ac4d6f53",
"submitter": "chad-atexpedient",
"github_login": "chad-atexpedient",
"run_id": "0723b52a-0942-43f8-914d-2b07dd2b11f5",
"tier": "easy",
"dataset_id": "e5",
"st... | {
"score": {
"mean_loss": 2.263614859115634,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011666666679084302
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3021321211160566,
"example_count": 600... |
299d2db9-12d5-4f5f-95f2-f59ee42b8eb8 | easy | sapient-sapiens | 2026-08-14 11:26:54.784985+00:00 | succeeded | e4f84d6a0331b8976fcf929252223adb1406e2d20e68c1831b022df73a93b639 | 16,878 | 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": "299d2db9-12d5-4f5f-95f2-f59ee42b8eb8",
"created_at": "2026-08-14 11:26:54.784985+00:00",
"db_md5": "f74d7e8297b7507c6ce7ec38e08c7889",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "8c10f9da-ebc0-4f80-b5c5-7aafc4004eec",
"tier": "easy",
"dataset_id": "e3",
"status": "su... | {
"score": {
"mean_loss": 2.1388573861740934,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.142984462275844,
"example_count": 800,
"... |
29a049e4-b25c-488f-a567-712fd2cdbe30 | easy | shirvani-jr | 2026-08-28 11:12:08.587881+00:00 | succeeded | 6e92bc87c307530e53faf008feedb887887e14837e82bbcb80a664ed8d6dfc71 | 51,183 | 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": "29a049e4-b25c-488f-a567-712fd2cdbe30",
"created_at": "2026-08-28 11:12:08.587881+00:00",
"db_md5": "ea8cfb8774e478067705077970765c2a",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "8d1c040c-ae30-4838-8f35-2dd750022e50",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 1.9401668906211853,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016666666604578495
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.864810824394226,
"example_count": 100... |
29aa7176-48f3-42f6-a683-d8e55d3df10f | easy | khushidahi | 2026-08-06 02:53:47.729876+00:00 | succeeded | e20d757e101b7f661a3815d0033c836f8d7b3e778f7c8f88abd49a345ab5f36f | 14,389 | 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": "29aa7176-48f3-42f6-a683-d8e55d3df10f",
"created_at": "2026-08-06 02:53:47.729876+00:00",
"db_md5": "359e5de9a2f197436875a9a296ce0312",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "2aada3cc-1979-43f2-b349-2d0b6a5e8aa6",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.302255153656006,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666828095913
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.297990560531616,
"example_count": 600,... |
29b16bd0-c647-4420-abad-6d7809ad9351 | easy | shreyash-chonkie | 2026-09-01 02:40:07.521969+00:00 | succeeded | bc59fd5fa6cb169f26d5c199948c14659c555d7db751e886596eead60ee1969a | 21,415 | null | """Learned token scanner feeding a random recurrent path reservoir."""
from __future__ import annotations
from itertools import permutations
import torch
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_s... | {
"id": "29b16bd0-c647-4420-abad-6d7809ad9351",
"created_at": "2026-09-01 02:40:07.521969+00:00",
"db_md5": "23db212b90286983c6996afc148bae19",
"submitter": "Shreyash",
"github_login": "shreyash-chonkie",
"run_id": "c8197439-1aa3-464e-b716-eac23a94ad98",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 27.63102149963379,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 27.631023406982422,
"example_count": 100,
"exac... |
29b9f9f3-c592-46d2-bd73-f1cb94ceecf7 | easy | Dogacel | 2026-08-11 20:08:27.228603+00:00 | succeeded | d108c6c4afdc95f1ae53788589346e03995a5216350cc3be83d0508759642b85 | 13,249 | null | """One Layer Deeper submission."""
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,
)
D_MOD... | {
"id": "29b9f9f3-c592-46d2-bd73-f1cb94ceecf7",
"created_at": "2026-08-11 20:08:27.228603+00:00",
"db_md5": "98f6194f526e448a3e08889f096e24df",
"submitter": "Doğaç Eldenk",
"github_login": "Dogacel",
"run_id": "8f9809cb-ed88-44ad-8e2e-b39afd39dfa9",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.54499614238739,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007499999832361937
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.5654685497283936,
"example_count": 800,... |
29bc8ebc-b3ab-4b8c-afde-a7ad951ce902 | easy | ehonig | 2026-08-21 01:04:41.404244+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": "29bc8ebc-b3ab-4b8c-afde-a7ad951ce902",
"created_at": "2026-08-21 01:04:41.404244+00:00",
"db_md5": "e6d7ec5bd1fddf2d609c6fc1ca4d18f7",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "14f1ef24-346a-4957-ba54-2e534284c6a4",
"tier": "easy",
"dataset_id": "e3",
"status": "succ... | {
"score": {
"mean_loss": 2.1926138401031494,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.009374999674037099
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.201631546020508,
"example_count": 800... |
29c482e4-53ab-416a-9444-848c0fffb157 | easy | jordanrubin | 2026-08-06 19:28:57.481085+00:00 | succeeded | 2214473aabeee3e5a3252928aedd50993436405bf900fc579ecc4ef88811f4f4 | 14,788 | null | """Autonomous digit-state recurrent Transformer for repeated modular squaring.
The model is deliberately organized around one learned transition:
p_0 = right_aligned_decimal_digits(x)
p_{k+1} = F_theta(p_k, decimal_digits(N))
The same two-block Transformer cell is applied exactly T times. T is used only
as ... | {
"id": "29c482e4-53ab-416a-9444-848c0fffb157",
"created_at": "2026-08-06 19:28:57.481085+00:00",
"db_md5": "f917ba4387feee357d1f6822e4a7825f",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "2c89559a-e2a5-42dc-a15d-0c18e0896a25",
"tier": "easy",
"dataset_id": "e3",
"status": "s... | {
"score": {
"mean_loss": 2.120428019072474,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00437499990221113
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.122330397981829,
"example_count": 800,
... |
29c4c6ce-81c5-4777-84f2-166fa1951b0f | easy | DDanlov | 2026-08-31 20:28:07.685872+00:00 | succeeded | 2bfcc522c89908299c03f5c9c6350edd479aa2ec92c458cca457d035bf3462db | 32,407 | null | """
Unified Continuous Hopfield Equilibrium Propagation Deliberation Engine
=======================================================================
Architecture:
- Deliberation: Continuous Hopfield Multi-Layer Energy System (L=4, D=768, d_ffn=2304, H=16)
- Integration: In-Place Strang-Split Conformal Symplectic Velocit... | {
"id": "29c4c6ce-81c5-4777-84f2-166fa1951b0f",
"created_at": "2026-08-31 20:28:07.685872+00:00",
"db_md5": "3eb8d5933b0208081688140ca5ba6351",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "ff9afcad-c1ec-4c02-b9fa-972373eb09ae",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded"... | {
"score": {
"mean_loss": 3.2022371754099694,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00125
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.2000034033976337,
"example_count": 800,
... |
29c6d054-58f6-4b17-bb4a-7ec385bbe4b7 | easy | Yalyenea | 2026-08-13 17:29:15.551229+00:00 | succeeded | ef7f8ce9d5dadb952eda15e3730e2483d711df2a9ce1381a0a8d6e6068729bd9 | 11,892 | null | """H-oriented d384 closed decimal state with shallow-depth weighting."""
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,
TokenLossBa... | {
"id": "29c6d054-58f6-4b17-bb4a-7ec385bbe4b7",
"created_at": "2026-08-13 17:29:15.551229+00:00",
"db_md5": "5869b6a7ed67d3ecf6adbe8ac51019e4",
"submitter": "yfff",
"github_login": "Yalyenea",
"run_id": "1a6beed1-59af-4f49-ad4f-da9774ad3fd7",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",
... | {
"score": {
"mean_loss": 3.6782534962503224,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.026250000161429247
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.891453266143799,
"example_count": 600... |
29e17c21-cdd6-48a2-959f-1c7f3c14cc97 | easy | isaac0804 | 2026-08-06 14:36:15.744940+00:00 | succeeded | 066d1e125c458b544274668a2eda843bb7eed1d74ea8cc692e68ab740f4805e9 | 9,324 | null | """Diagnostic-only: isolate whether ONE of E3's own moduli can be learned to
held-out generalization at all, independent of the 26-other-moduli joint
training question.
docs/findings.md ("Session log: numerically diagnosing the E3 collapse")
established that E3's held-out failure is not step-starvation or capacity --
... | {
"id": "29e17c21-cdd6-48a2-959f-1c7f3c14cc97",
"created_at": "2026-08-06 14:36:15.744940+00:00",
"db_md5": "d099a501edc16bf978afa28971995b3d",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "e5b324e9-3f12-49aa-b8fb-157e1ac311ab",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 16.825803045246833,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0068750000093132264
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 16.59200964587849,
"example_count": 80... |
29e4e348-4509-4fcf-b308-6c4db273027b | easy | KaustubhKumar05 | 2026-08-25 16:24:55.092361+00:00 | succeeded | 0dd780da0f2643e57765b3d6e0e2a43ad3d3d1878ba500bfc9e4abda1f9875d1 | 17,167 | null | """conv_abacus: final_v1 + the CORRECT Abacus digit embedding.
final_v1 indexed positions from the end of the whole padded sequence, so the
units digit of N and the units digit of x landed on DIFFERENT embeddings --
the exact failure McLeish et al. (arXiv:2405.17399) identify. This uses a
table indexed by significance... | {
"id": "29e4e348-4509-4fcf-b308-6c4db273027b",
"created_at": "2026-08-25 16:24:55.092361+00:00",
"db_md5": "3009ed5dc9d1fec3670f0014eba79c15",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "cf7f7eef-f584-42df-939a-17eb9f0d675a",
"tier": "easy",
"dataset_id": "e8",
"status": "succee... | {
"score": {
"mean_loss": 7.0438079833984375,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.1400623954832554
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.521512985229492,
"example_count": 85,
... |
29e600a0-a356-4e9b-a694-bb0c1eb3d306 | easy | AFeinfield | 2026-08-19 23:13:56.250865+00:00 | succeeded | bf333e291b7fa2382a52de2b2dcc4dd185d0f5064ad3b957aa01a16f817e9696 | 19,260 | null | """092: INSTRUMENTED BASE — two measurement fixes, no architecture change.
(1) Step-indexed LR schedule with a fixed screening step budget. The old
schedule decayed on WALL CLOCK, so throughput jitter changed the
LR-per-step trajectory: identical files on e2 gave 1470 vs 1025 steps
and train acc 0.725 vs 0... | {
"id": "29e600a0-a356-4e9b-a694-bb0c1eb3d306",
"created_at": "2026-08-19 23:13:56.250865+00:00",
"db_md5": "2dcbfc96ca5a2d30e8fddecf166d77d7",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "2742bc21-6dd7-45e2-bcbb-3d90e54a86fa",
"tier": "easy",
"dataset_id": "e2",
"status": "succ... | {
"score": {
"mean_loss": 8.40378212928772,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.003750000149011612
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.878802299499512,
"example_count": 300,
... |
29e92c8f-cfbb-4655-ac73-7f382423dd64 | easy | DDanlov | 2026-08-28 22:19:18.570922+00:00 | succeeded | bd42cc8ea2c326d8a3ecb8c1f3a05093ebf67f7c7b0e24bf5e2a1d8242941be5 | 15,430 | null | """
Trial 42: Model D: D=512, H=8, M=8, u=2 (T=16)
"""
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from dataclasses import dataclass
from typing import Optional, Tuple, Dict, Any, List, Union
try:
from benchmark.api import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, T... | {
"id": "29e92c8f-cfbb-4655-ac73-7f382423dd64",
"created_at": "2026-08-28 22:19:18.570922+00:00",
"db_md5": "f861e8d4fb1cc15fe2cc27683fe48405",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "df838c5b-9101-49fb-8c2a-5f4015e0d499",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.2837548631399214,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0141666666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.285876351755296,
"example_count": 600,
... |
29e98dfd-095d-4aa3-a31a-5ede63481d02 | easy | AFeinfield | 2026-08-28 00:28:26.267376+00:00 | succeeded | a0ee421982bac33c0dfc402294e74848df92ef826fc996ca82c7eeba0fd2fd28 | 15,435 | null | """173: 161 + TIME-SCHEDULED X-SIZE CURRICULUM (comparator bootstrap).
Every variant plateaus at loss ~2.2: a gradient chicken-and-egg where
each operator needs the others correct to receive clean signal. The
carry op got its curriculum from the identity bias (158); this gives the
comparator one: small-x rows dominate... | {
"id": "29e98dfd-095d-4aa3-a31a-5ede63481d02",
"created_at": "2026-08-28 00:28:26.267376+00:00",
"db_md5": "d4a48cd7413c8971dfe067318ae07aa7",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "265d9e5c-53fe-4f5b-bd05-eb89344c23f1",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.2813162503170545,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.013750000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.28310619769075,
"example_count": 600,... |
29ecfbcd-eb55-4d55-ab15-0581ec50e96b | easy | karanganesan | 2026-08-05 11:14:07.700661+00:00 | succeeded | bf0e76c69f77f662a748aee1d0ab429d39874b15d8294f560ebbca6deaadcd8b | 22,081 | null | """Parametric looped-transformer family (P1).
One weight-tied transformer block applied k times in latent space. Config
flags cover four P1 families with one file:
looped recall=0 gated=0 tfilm=0 plain weight-tied loop
looped-recall recall=1 re-inject the input embedding each
... | {
"id": "29ecfbcd-eb55-4d55-ab15-0581ec50e96b",
"created_at": "2026-08-05 11:14:07.700661+00:00",
"db_md5": "02fd6bcee576dc90165fe20a25159dc2",
"submitter": "Karan Ganesan",
"github_login": "karanganesan",
"run_id": "7e9b786b-4409-434d-a1a9-9d215efd50eb",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 12.94558572769165,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.021666666492819786
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 13.566842079162598,
"example_count": 100... |
29f5931a-a027-4874-8c73-e0dcce1e79e8 | easy | sapient-sapiens | 2026-08-17 19:44:07.304595+00:00 | succeeded | 73b24d001589dea7b46cf54d4671cee005cbfcf7bd92b0eb9133263887ecc018 | 28,660 | null | """v40 lineage: T-free operator, N memory, PonderNet mix, no T in the loss.
Field-local encoder, T-free core, countable halt, answer-slot readout.
N is seeded and restored every step. T is never parsed in forward or loss.
"""
from __future__ import annotations
import math
import time
import torch
import torch.nn.fu... | {
"id": "29f5931a-a027-4874-8c73-e0dcce1e79e8",
"created_at": "2026-08-17 19:44:07.304595+00:00",
"db_md5": "ab0995ba57df7797ce934d12bf6df7e5",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "ed4171e1-d645-412c-910c-09fc7a74acd5",
"tier": "easy",
"dataset_id": "e6",
"status": "su... | {
"score": {
"mean_loss": 1.8661311987328202,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.060360362199512685
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.8443533182144165,
"example_count": 60... |
29f5f386-cacc-4e09-b3fe-62c8e5a7ae8f | easy | TheYkk | 2026-08-04 13:02:16.235548+00:00 | succeeded | fd414def27aa2e07a48b647eef13cf2a63ddb666a1817b7df87815acea83551f | 22,523 | null | """Encoder / iterated-operator / positional-decoder composition model.
Depth-selection ablation on a base that actually fits
----------------------------------------------------
Everything here -- value pathway, transition operator, Muon+AdamW optimizer and
its routing, schedule, loss, batch constants, reuse policy --... | {
"id": "29f5f386-cacc-4e09-b3fe-62c8e5a7ae8f",
"created_at": "2026-08-04 13:02:16.235548+00:00",
"db_md5": "631a1bb131860a8090194461c1642a5c",
"submitter": "Kaan Karakaya",
"github_login": "TheYkk",
"run_id": "0d590a0f-c531-49a9-a14b-c973e2d03da8",
"tier": "easy",
"dataset_id": "e1",
"status": "succe... | {
"score": {
"mean_loss": 8.939188003540039,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.024999999441206455
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.252317428588867,
"example_count": 100,... |
29f6bef8-7ed9-4278-8547-94a86462b4b5 | easy | nikhilkumar26 | 2026-08-31 02:21:31.684561+00:00 | succeeded | e219f70eefd3677b24d5f5f613fbd5efc95d779f8c051bfb2d2c7410204d8ed7 | 18,455 | null | """T-adaptive learned recurrence. T is unroll count only.
A shared step maps digit embeddings of the current integer and of N to
next-digit logits. Between unrolls, STE/argmax updates the digit state so T
composes. No problem-specific operators are hardcoded.
"""
from __future__ import annotations
import math
impor... | {
"id": "29f6bef8-7ed9-4278-8547-94a86462b4b5",
"created_at": "2026-08-31 02:21:31.684561+00:00",
"db_md5": "e8b7edda931a5da1a06cc0d3541a10ec",
"submitter": "Nikhil Kumar",
"github_login": "nikhilkumar26",
"run_id": "deaf00be-8557-46c2-9bca-bd34f24f8ae0",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.070334371105993,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0466666671137015
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.9467691249811827,
"example_count": 100,
... |
2a055a5b-858f-4131-ab8b-92610ca1fc76 | easy | ipsa-loquitur | 2026-08-26 16:29:31.271176+00:00 | failed | 2d9f4f3331ca419636cb39310881dafc828cb2a064f881d8fef0e77a7487ffa4 | 15,929 | null | """Standalone NetF-V1 submission candidate for One Layer Deeper.
The learned rule is shared across bit positions, modular-square iterations,
moduli, and tape widths. During training, exact double-and-add states provide
dense one-step supervision. During evaluation, the model receives no state
corrections and rolls out... | {
"id": "2a055a5b-858f-4131-ab8b-92610ca1fc76",
"created_at": "2026-08-26 16:29:31.271176+00:00",
"db_md5": "faf08fffbdb0cf932240bbeff7141be0",
"submitter": "ipsa-loquitur",
"github_login": "ipsa-loquitur",
"run_id": "07d3a320-9b33-4b7f-9b54-66091ef0c280",
"tier": "easy",
"dataset_id": "e3",
"status":... | null
|
2a0582ab-09ca-4d7f-99a1-04f07b476448 | easy | DDanlov | 2026-08-10 08:00:33.384477+00:00 | succeeded | 742aa178c8f34e0df18080215389c3a77fc7507be7138cf63c986ebc254c2449 | 16,672 | 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": "2a0582ab-09ca-4d7f-99a1-04f07b476448",
"created_at": "2026-08-10 08:00:33.384477+00:00",
"db_md5": "5c9396a9e78cad54a98e0b4a56c30ec9",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "b95ca0da-1d8e-44eb-934c-7695bfb09046",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 6.662388707818934,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00125
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.109464847720792,
"example_count": 600,
"e... |
2a07d3fe-17f4-47f0-9b2a-ecbcc0488388 | easy | erdavis0 | 2026-08-25 01:28:19.393330+00:00 | succeeded | 362ea16f0171e3e2ad492a460b5ba6fbdde70a64eea30dea9a6544b2214be1be | 24,120 | null | """Matched categorical-boundary panel: C4 ST oracle plus commitment.
Public delimiter syntax is used only to route opaque token identities onto
separate right-aligned tapes. One T-blind learned macro maps a categorical X
tape and a digitwise N tape to a new categorical tape through freshly reset
interaction, carry, a... | {
"id": "2a07d3fe-17f4-47f0-9b2a-ecbcc0488388",
"created_at": "2026-08-25 01:28:19.393330+00:00",
"db_md5": "19b07a0124d2461910b21f1195de2485",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "a7718d50-7432-464c-ac7a-4ba453487c8d",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.146667650257588,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333333333335
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.145254565995905,
"example_count": 600... |
2a0d1bb3-1442-4666-8c78-b504eccefdfe | easy | yashkant | 2026-08-09 02:48:39.480966+00:00 | succeeded | 9084788c57d70e9371c16efd7b28d9e381cd80dfe176a2de867c67d09ded40c6 | 22,836 | null | """Round 211 common TRM-style x/y/z shared-transition template."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
from torch import... | {
"id": "2a0d1bb3-1442-4666-8c78-b504eccefdfe",
"created_at": "2026-08-09 02:48:39.480966+00:00",
"db_md5": "3e558755dcaae2ce6d98f73727e41002",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "a46a4a37-3e74-48db-8134-b6e340d34e04",
"tier": "easy",
"dataset_id": "e4",
"status": "succeed... | {
"score": {
"mean_loss": 2.0635987726242924,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0085416666790843
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.0662513617918035,
"example_count": 1200... |
2a0dda9b-40da-475e-9a16-3de6cae3a3a2 | easy | ArkinDharawat | 2026-08-13 16:53:19.296472+00:00 | succeeded | f38157e4ec646ec2ae87f558cb9018ab67646498ffa4bbec51da9d29d60ef6ba | 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": "2a0dda9b-40da-475e-9a16-3de6cae3a3a2",
"created_at": "2026-08-13 16:53:19.296472+00:00",
"db_md5": "2df26b249b0f6d7f691138d250670eaf",
"submitter": "Arkin Dharawat",
"github_login": "ArkinDharawat",
"run_id": "22851cce-e77a-43bd-92ef-b8eca64ffb7c",
"tier": "easy",
"dataset_id": "e3",
"status"... | {
"score": {
"mean_loss": 2.274918661024201,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.012500000167638064
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.306018532952392,
"example_count": 800,... |
2a1574c8-268b-4224-9a15-d83a0d86b8b5 | easy | erdavis0 | 2026-08-27 02:53:44.482009+00:00 | succeeded | 2631fff557194c25861f5cbd2a3339df5a8a8980d811bc985ebcaec80ee4bdc0 | 31,299 | null | """Typed-prefix reversible transition B5: CE/sequence-mixture hybrid.
Public delimiter syntax routes opaque N, X, and T fields into three strict
functional roles. X initializes only an evolving auxiliary tape, N initializes
only an invariant environment lane, and T reaches only a learned monotone
prefix controller. ... | {
"id": "2a1574c8-268b-4224-9a15-d83a0d86b8b5",
"created_at": "2026-08-27 02:53:44.482009+00:00",
"db_md5": "2981fa4ef2c40f38bd0d3090c6d64a88",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "56f8f928-1306-475a-ae78-4dc44ad8f892",
"tier": "easy",
"dataset_id": "e2",
"status": "succeeded",... | {
"score": {
"mean_loss": 4.9716410636901855,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.026250001043081284
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.3914103507995605,
"example_count": 30... |
2a17a1a5-90dc-433f-b9b2-eb22495c22ea | easy | jordanrubin | 2026-08-14 01:22:02.900891+00:00 | succeeded | fc98447fc19a656845f48580dd88f53c0b2d7d00c35e32a8fe2d3dfff79a73f4 | 29,581 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "2a17a1a5-90dc-433f-b9b2-eb22495c22ea",
"created_at": "2026-08-14 01:22:02.900891+00:00",
"db_md5": "d08ba312cd6460ffc449c5d2476626a5",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "8e6cdb75-dc68-4db8-bed3-cdc5fffbd71c",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.8958707664175316,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.028749999434997637
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.251583490671064,
"example_count": 600... |
2a187c0f-e349-43ed-b5e3-c93fa5aa82e4 | easy | khushidahi | 2026-08-03 20:47:58.368404+00:00 | succeeded | 57ceae53f297e524748b11ea17f1146fc80f66389d66a191a0c93d47c2e551ef | 15,250 | 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": "2a187c0f-e349-43ed-b5e3-c93fa5aa82e4",
"created_at": "2026-08-03 20:47:58.368404+00:00",
"db_md5": "c9954932ad447559abe923bd25cf7a34",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "10916d6a-1556-4a86-9d10-170b0169f047",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.1978940963745117,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333302289248
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.193479537963867,
"example_count": 600... |
2a18b44f-3cea-4231-9b9c-913f24712dd7 | easy | amirmeisami | 2026-08-18 22:03:23.289225+00:00 | succeeded | 48be41df9cdb4008cb9dfceb31eeee6f484831081086f6e9722ad6c0f119b27d | 12,068 | null | """T021 Arm D: algebraic position-aware recurrent network."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import ModelSpec, OptimizerBundle, OptimizerSpec, Submission, TokenLossBatch, assert_model_state
D_MODEL, POLY_WIDTH, IN... | {
"id": "2a18b44f-3cea-4231-9b9c-913f24712dd7",
"created_at": "2026-08-18 22:03:23.289225+00:00",
"db_md5": "380fc4a6fcab0a79234f5bfd10c542b4",
"submitter": "Amir Meisami",
"github_login": "amirmeisami",
"run_id": "2ec820e3-bba2-4850-bc73-d87761c53509",
"tier": "easy",
"dataset_id": "e3",
"status": "s... | {
"score": {
"mean_loss": 4.942436300545635,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0075
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.958804685025341,
"example_count": 800,
"ex... |
2a1d73cc-aa3b-44d9-86be-0eea6686c82c | easy | yashkant | 2026-08-16 00:50:35.703047+00:00 | succeeded | c6c6ed6cf3b85665f7406668a44133040e70ac9c2ba9b590a435c1ece4860a21 | 36,180 | null | """Round1763 residual-conditioned shallow/deep Dykstra mixer.
Every parsed public-T cycle exposes the state after its first shared A
projection (the 1*T shallow readout) and after the full shared A-B-A cycle (the
3*T deep readout). Differentiable state-update energies plus a terminal/mean
convergence ratio drive a tin... | {
"id": "2a1d73cc-aa3b-44d9-86be-0eea6686c82c",
"created_at": "2026-08-16 00:50:35.703047+00:00",
"db_md5": "217cd04747813dc51462da7e9e80a3ff",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "fde04aad-5ef7-4bcc-88e6-ac91dcb8edc0",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 4.284561216831207,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3093750262632966
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.040772914886475,
"example_count": 300,
... |
2a1fa17e-3209-4726-8aa1-9f1e29fafa26 | easy | Ranoobaba | 2026-08-17 19:02:10.328644+00:00 | succeeded | 2d85a87161e095b725e692963c2d592bafbe4ab13614e0dd6c9278e7c57d4d9c | 5,760 | 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": "2a1fa17e-3209-4726-8aa1-9f1e29fafa26",
"created_at": "2026-08-17 19:02:10.328644+00:00",
"db_md5": "5fc4b6adc7b2105d966b958f66b07692",
"submitter": "Syed Rayyan Ali",
"github_login": "Ranoobaba",
"run_id": "516cb9d1-933b-4db7-accd-ff34f7fffd19",
"tier": "easy",
"dataset_id": "e1",
"status": "... | {
"score": {
"mean_loss": 6.894183504313899,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.056666665971279145
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.043099880218506,
"example_count": 100,... |
2a1feb2c-6e19-42da-878e-2bab677ff814 | easy | khushidahi | 2026-08-22 01:54:14.322803+00:00 | succeeded | a7977dcca57c2b7b1bd701186966c52ddcaf866a4b105d47dbbc98a37824159f | 30,755 | null | """R355: batch-256 K8 variable-modulus exposure control.
This submission tests the strongest rules-safe generic baseline suggested by
the public competition discussion. It does not implement squaring, modular
reduction, or any other task solver. Numeric prompt fields are only rearranged
onto equal-length, right-alig... | {
"id": "2a1feb2c-6e19-42da-878e-2bab677ff814",
"created_at": "2026-08-22 01:54:14.322803+00:00",
"db_md5": "833a3742ca75541f8c269a43e9d30f6a",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "0550a2aa-09b9-4951-b7c3-b1ae30110864",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.9561659965346747,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005416666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.9888701535126554,
"example_count": 60... |
2a2108b8-afd6-4e6f-8a55-72dfda4e5f45 | easy | KaustubhKumar05 | 2026-08-28 04:34:01.184857+00:00 | succeeded | 683e069bbb3ddde1e9fb919cceb56613c210f27ca986966ca1daa54852c9f623 | 14,906 | null | """cell_fix -- cell_op2 with the four ARCHITECTURE changes the arxiv sweep converged on.
Six independent search lanes kept landing on the same mechanisms, and cell_op2 had
none of them. Each is a pure forward-pass change computed from input_ids -- no data
is touched, no tokens are added, no algorithm is supplied.
1... | {
"id": "2a2108b8-afd6-4e6f-8a55-72dfda4e5f45",
"created_at": "2026-08-28 04:34:01.184857+00:00",
"db_md5": "4e2cafea9e50087cf076729e0991538d",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "e890f619-c9d0-4539-a12f-cf07b4584c48",
"tier": "easy",
"dataset_id": "e6",
"status": "succee... | {
"score": {
"mean_loss": 4.4136738777160645,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.07184685114771128
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.116267681121826,
"example_count": 60,
... |
2a21e793-5705-4b9d-942f-d7893f9d7502 | easy | jayyala | 2026-08-16 23:23:21.906057+00:00 | succeeded | 84301c5daa7cf3516f2dca8586af0e29a12d0caaf01032a2b3ca60113fc8eacd | 20,507 | null | """One learned modular-square step, applied exactly T times.
The underlying map is recovered in a modulus-equivariant form: the integer
product x*x is represented as a circular phase, which wraps automatically,
and a tied coarse-to-fine decoder inverts that phase to a residue. The same
decoder is trained by reconstru... | {
"id": "2a21e793-5705-4b9d-942f-d7893f9d7502",
"created_at": "2026-08-16 23:23:21.906057+00:00",
"db_md5": "9e92b4e9a5e58a3ab723917f67bbf1db",
"submitter": "jayyala",
"github_login": "jayyala",
"run_id": "254f220b-82ca-4803-bbad-5325ea80d12f",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded"... | {
"score": {
"mean_loss": 36.078357062261446,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.000625
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 36.21371441184346,
"example_count": 800,
... |
2a2ca12a-c95e-4e3c-9040-37f9276e22b1 | easy | 0Chris5R | 2026-08-02 16:43:09.922101+00:00 | succeeded | ba8a3c611544e9b7672a524359f3dd5c89c7513902087a73c1394ec8492a4cba | 24,605 | null | """Shared-digit Neural GPU with late full-objective EMA evaluation.
The model aligns decimal digits by significance, forms a learned low-rank
second-order feature map of the current state, and evolves a two-dimensional
workspace with the original convolutional-GRU equations. Each arithmetic
transition applies the sam... | {
"id": "2a2ca12a-c95e-4e3c-9040-37f9276e22b1",
"created_at": "2026-08-02 16:43:09.922101+00:00",
"db_md5": "b0ab50517876c2ad3db0856cbc42a75d",
"submitter": "Chris ",
"github_login": "0Chris5R",
"run_id": "8d59e6c5-be85-4263-9df7-1024f2ec9940",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 2.541117977998244,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02375000002483527
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.3860826449543904,
"example_count": 600,... |
2a2f12c0-07f3-45f7-9bce-32180859e1ca | easy | erdavis0 | 2026-08-26 04:14:12.395345+00:00 | succeeded | c67927e1f1a3eb13d9778a95b41512101bbd0baa70d6fd25c9f9b89607e3826f | 30,360 | null | """Latent-role continuous REV C2: exact length-normalized sequence mixture.
Public delimiter syntax only routes opaque token categories into distinct,
right-aligned N, X, and T tapes. Learned local cells compile separate
interaction, carry, and modulus-conditioned workspaces. N is isolated from
interaction and carry... | {
"id": "2a2f12c0-07f3-45f7-9bce-32180859e1ca",
"created_at": "2026-08-26 04:14:12.395345+00:00",
"db_md5": "b3917270504f5239445a48889c48e9ae",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "254f2fff-e506-4b0e-baa0-cd2487c340ef",
"tier": "easy",
"dataset_id": "e7",
"status": "succeeded",... | {
"score": {
"mean_loss": 4.7682945728302,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.05213903822004795
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.285816192626953,
"example_count": 85,
... |
2a359939-9a47-4cda-a0e2-0b2661f446d4 | easy | yashkant | 2026-08-12 04:15:59.180819+00:00 | succeeded | f7f9386f020f4cbd87f4243a8e51989ec2a344e2fb1b8006dcab536e7615092f | 31,750 | null | """Round1701 Dykstra width16/hidden32 architecture pair on the D416 MemGLU carrier.
The fixed Round1141 carrier first produces its ordinary pooled representation.
A compact primal state and two Dykstra correction memories are initialized
from that immutable representation. Every parsed public-T cycle applies two
cond... | {
"id": "2a359939-9a47-4cda-a0e2-0b2661f446d4",
"created_at": "2026-08-12 04:15:59.180819+00:00",
"db_md5": "6095e33e176ae791430e4fbd74b9eaa2",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "68425f98-40c2-41ae-96b7-3fe79dd07347",
"tier": "easy",
"dataset_id": "e4",
"status": "succeed... | {
"score": {
"mean_loss": 2.070310138073011,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010000000012417635
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.072859039993267,
"example_count": 1200... |
2a413cd5-3d95-47d1-b949-8767baef33a6 | easy | shirvani-jr | 2026-08-29 02:35:09.390404+00:00 | succeeded | cdcc77f3e9e85553b1d9ce04bf088d3d27a307d677cc104409f0dc753e519bad | 50,429 | 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": "2a413cd5-3d95-47d1-b949-8767baef33a6",
"created_at": "2026-08-29 02:35:09.390404+00:00",
"db_md5": "5638dd9599b8dc90204f7c26def5d9ee",
"submitter": "Ali",
"github_login": "shirvani-jr",
"run_id": "32a640ab-3e23-42ed-9a41-9a39b22d0516",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 1.869745135307312,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06666666641831398
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 1.784350037574768,
"example_count": 100,
... |
2a45f78a-3b7e-49a4-8d05-589660db37fc | easy | oupadhyay | 2026-08-15 22:41:53.594357+00:00 | succeeded | fb9b08a0b88a4d8762edcc09d2bf7972e09895650c62838396bbcc72015a803a | 4,315 | null | """Conditional structured-loss screen: hard50."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import nn
from benchmark import ModelSpec,OptimizerSpec,OptimizerBundle,Submission,TokenLossBatch,assert_model_state
W,D,H=4,64,32
class C:
def __init__(s,vocab_size,max_... | {
"id": "2a45f78a-3b7e-49a4-8d05-589660db37fc",
"created_at": "2026-08-15 22:41:53.594357+00:00",
"db_md5": "025fdb1a4ad91e3df515a880a83a263b",
"submitter": "Ojasw Upadhyay",
"github_login": "oupadhyay",
"run_id": "de96f078-b83f-430f-aa44-961a8d0b32fc",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.3091904767806066,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.182662495583162,
"example_count": 600,
"... |
2a49aa5b-c1d3-4cb2-a947-a10b87447e91 | easy | khushidahi | 2026-08-08 00:53:04.808734+00:00 | succeeded | 305274822259c716377aca107e90897443e479f351434d88bf926098edef89b5 | 13,208 | null | """R42 Fourier-number representation experiment.
All input-dependent numeric parsing stays inside differentiable GPU tensor
operations. This is a learned-model representation, not a task solver: it does
not compute modular squaring or create intermediate labels.
"""
from __future__ import annotations
import math
impor... | {
"id": "2a49aa5b-c1d3-4cb2-a947-a10b87447e91",
"created_at": "2026-08-08 00:53:04.808734+00:00",
"db_md5": "8e49302f11c16a65fa066b276d35454e",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "241cc790-f0ae-4309-8962-cd5db8d35867",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.160932183265686,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1677846908569336,
"example_count": 600... |
2a5101d8-b6db-4756-86a5-3aafe480d64a | easy | yashkant | 2026-08-09 02:28:55.681000+00:00 | succeeded | da39b642da1afd6e1de460cec068bfa4cf1d92197848ffb327a0255903cfa95d | 22,836 | null | """Round 211 common TRM-style x/y/z shared-transition template."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBatch,
assert_model_state,
)
from torch import... | {
"id": "2a5101d8-b6db-4756-86a5-3aafe480d64a",
"created_at": "2026-08-09 02:28:55.681000+00:00",
"db_md5": "d191e6ac3e82ab0f77439af3060169fc",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "5553d639-9ab1-4b8b-9c0e-e04d33a21d04",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.118572165292309,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016875000083819033
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1187308532729285,
"example_count": 800... |
2a5315c1-12c7-4a20-9463-84e266447279 | easy | ehonig | 2026-08-21 01:43:20.189313+00:00 | succeeded | a8101c7bf20f5fb52423225be6359acd59e36c1d9a8a08942c855acf40ae70ac | 31,882 | null | """A plain Transformer for repeated modular squaring, and nothing else yet.
This is a deliberate restart. The previous submission accumulated a field
parser, a carry scan, a learned reciprocal, periodic features, a digit
bottleneck, a depth selector and two cell types, most of them tuned around a
memorisation table th... | {
"id": "2a5315c1-12c7-4a20-9463-84e266447279",
"created_at": "2026-08-21 01:43:20.189313+00:00",
"db_md5": "75d8fa2efbdc026e18b4eec8692e50b8",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "4e93d390-f0fc-408e-85c9-2470c1c258a2",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.190457120739125,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00125
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.19283390045166,
"example_count": 600,
"ex... |
2a55ab45-7453-4cbf-b09a-055d025e05ee | easy | khushidahi | 2026-08-23 21:03:20.283534+00:00 | failed | dc30f96367eeccb5fe562f1ff3bdff32abfe024a824930c0ad6f787d71aa41f0 | 11,734 | null | """ART-DT recall candidate.
A learned recurrent sequence model with right-to-left Abacus positions and
global-sequence rotary attention. This variant changes only optimizer LR and
warmup schedule from its architectural parent. This branch ties the two
within-visit layer calls to one TransformerLayer object. One imm... | {
"id": "2a55ab45-7453-4cbf-b09a-055d025e05ee",
"created_at": "2026-08-23 21:03:20.283534+00:00",
"db_md5": "25f8e724d7ec7f02d7c56a4ef3401964",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "5aaa71b6-6a2a-4774-bfb6-b35533b5dfc3",
"tier": "easy",
"dataset_id": "e9",
"status": "fail... | null
|
2a6429f3-92fc-4686-bebc-68b750f3151c | easy | erdavis0 | 2026-08-27 03:28:26.488486+00:00 | succeeded | ca6f1b5c93401e93c705f4189e451ef833a9103266142696e3b482501ff8d97a | 31,305 | null | """Typed-prefix reversible transition B1: CE plus selector-oracle KL.
Public delimiter syntax routes opaque N, X, and T fields into three strict
functional roles. X initializes only an evolving auxiliary tape, N initializes
only an invariant environment lane, and T reaches only a learned monotone
prefix controller. ... | {
"id": "2a6429f3-92fc-4686-bebc-68b750f3151c",
"created_at": "2026-08-27 03:28:26.488486+00:00",
"db_md5": "3f9846560514c463a6ae282585eac958",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "2109ad72-5d00-4585-bc1b-7d9da55030aa",
"tier": "easy",
"dataset_id": "e4",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.6555096038394685,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0030208333333333333
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.6392633845453304,
"example_count": 1... |
2a663042-4520-4719-a1c3-d24830777abb | easy | gauravmishra | 2026-08-08 20:36:53.496223+00:00 | succeeded | b84d946c58e12d3d926f809e18ee7c3cfa3cba0740c97d44e43b8baf461881b5 | 4,745 | null | """B0-s25 with its learned position table indexed from the valid right edge."""
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_MODE... | {
"id": "2a663042-4520-4719-a1c3-d24830777abb",
"created_at": "2026-08-08 20:36:53.496223+00:00",
"db_md5": "5e78caad3ad92529e261f98e4b8be3ef",
"submitter": "Gaurav Mishra",
"github_login": "gauravmishra",
"run_id": "223375b7-f772-4e1f-aed8-b85be2ad8474",
"tier": "easy",
"dataset_id": "e5",
"status": ... | {
"score": {
"mean_loss": 2.3146175143395435,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666666666666
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.433651671815881,
"example_count": 600... |
2a70b0f9-f39a-4acc-95ea-aec8748aae3c | easy | DDanlov | 2026-08-11 21:12:16.427836+00:00 | failed | e8ee1d2cf0fdca5e776960e38550f18e36e60f016bfe787c0582c73f4be88ba1 | 19,621 | 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": "2a70b0f9-f39a-4acc-95ea-aec8748aae3c",
"created_at": "2026-08-11 21:12:16.427836+00:00",
"db_md5": "357d9671f37cdea78842dad68ebd5a0c",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "43030f8b-2144-4d7b-b213-648fb4aefe07",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
2a76edbe-b916-45c3-bd74-914c0d86254b | easy | KaustubhKumar05 | 2026-08-25 15:48:24.991047+00:00 | succeeded | 03d0c373e594aa5715323fd2ba901ecc5a0250f4deb9d9e32371e95dc8a36c27 | 9,417 | null | """abacus: faithful port of the Abacus recipe (McLeish et al., arXiv:2405.17399).
The load-bearing idea is the embedding. Each digit is indexed by its
SIGNIFICANCE WITHIN ITS OWN NUMBER (0 = least-significant), from a table
SHARED across N, x and T. Digit k of N and digit k of x therefore land on the
same embedding an... | {
"id": "2a76edbe-b916-45c3-bd74-914c0d86254b",
"created_at": "2026-08-25 15:48:24.991047+00:00",
"db_md5": "80b03c6db927f2e12cf645f2453373df",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "6cce5283-0c50-47a6-ad27-70640918a649",
"tier": "easy",
"dataset_id": "e8",
"status": "succee... | {
"score": {
"mean_loss": 5.874668121337891,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06599821895360947
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 5.997165203094482,
"example_count": 85,
... |
2a778551-bc70-4304-907b-0b4ce6112a11 | easy | DDanlov | 2026-08-21 02:37:31.511923+00:00 | failed | 8e4fd4bab584c83cbd19c5b14dbb23bdc439df7548ab275f16d802949d211570 | 25,133 | 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": "2a778551-bc70-4304-907b-0b4ce6112a11",
"created_at": "2026-08-21 02:37:31.511923+00:00",
"db_md5": "69157ab1aa9b83e9f69d8b5b5e90448a",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "a7ecfb71-c491-455f-a0cf-86a5f1acda86",
"tier": "easy",
"dataset_id": "e5",
"status": "failed",
... | null
|
2a7a2ce7-f8ec-423a-a36c-2510545180a9 | easy | KaustubhKumar05 | 2026-08-27 12:12:23.217884+00:00 | succeeded | 0e9f25a4cc7592f3adc331bfdb9a2f3566e06eb44239586933b609e7350a1085 | 14,701 | null | """abacus_sine — SINUSOIDAL activation instead of GELU.
§72: TWENTY candidates in this project and every single one used GELU -- the
activation was inherited from TransformerEncoderLayer and never once varied.
Salah & Yevick 2411.05353 report that grokking on modular arithmetic is
CONTROLLED by the activation function... | {
"id": "2a7a2ce7-f8ec-423a-a36c-2510545180a9",
"created_at": "2026-08-27 12:12:23.217884+00:00",
"db_md5": "445fe6eb1b47727ceb55b42fc155d11d",
"submitter": "koz",
"github_login": "KaustubhKumar05",
"run_id": "aefd9ed8-9cba-472c-8dc9-e4cc67758f1c",
"tier": "easy",
"dataset_id": "e3",
"status": "succee... | {
"score": {
"mean_loss": 3.180634360805942,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00375
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.1465976529127415,
"example_count": 800,
"... |
2a7a8d92-dbb2-4121-9c3d-af159408d47e | easy | neel04 | 2026-08-11 11:36:00.960875+00:00 | succeeded | 71aead330d7fe8432716a4bca3e75ace069321ec11f737ce9cb8f2fe9500bb74 | 7,471 | null | """Segment-size-one online training with normalized attractor dynamics."""
from __future__ import annotations
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
BatchReuseContext,
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLossBa... | {
"id": "2a7a8d92-dbb2-4121-9c3d-af159408d47e",
"created_at": "2026-08-11 11:36:00.960875+00:00",
"db_md5": "a5b75a142009b9ad64dd78365e3e9ebb",
"submitter": "Neel Gupta",
"github_login": "neel04",
"run_id": "4a7c30cd-e4c3-407f-8567-50eb6b797f63",
"tier": "easy",
"dataset_id": "e1",
"status": "succeede... | {
"score": {
"mean_loss": 2.179817795753479,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.196465253829956,
"example_count": 100,
... |
2a91dc8b-e397-4dd1-91b6-0429991f1517 | easy | sapient-sapiens | 2026-08-16 17:55:11.003370+00:00 | succeeded | 47818ed67a0a0f7f92d59d9e800f7a1d841275077f340ac0ee2fddbec507fcda | 27,080 | null | """E6 field-width padding candidate: frozen v55 plus leading-zero N/X/T pads.
The only change relative to act64_e6_w1024_dh4_L6_v55 is the prompt
representation. Each N, X, and T digit field is left-padded with digit-zero
tokens to a fixed Easy width so the same left-relative slot is always the same
decimal place. Arc... | {
"id": "2a91dc8b-e397-4dd1-91b6-0429991f1517",
"created_at": "2026-08-16 17:55:11.003370+00:00",
"db_md5": "446ec9093a467d6a5f06b29304ecddf7",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "2fd0b7ef-fbbc-4df2-8a65-adab79459746",
"tier": "easy",
"dataset_id": "e6",
"status": "su... | {
"score": {
"mean_loss": 5.6521222297459435,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.12184685529083819
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 4.752291202545166,
"example_count": 60,
... |
2a931fc1-c695-4c00-ad36-bb20b178657e | easy | jordanrubin | 2026-08-14 01:07:00.103995+00:00 | succeeded | fc98447fc19a656845f48580dd88f53c0b2d7d00c35e32a8fe2d3dfff79a73f4 | 29,581 | null | """Weight-tied MLP-cell loop over soft digit states (mlploop).
Cell = MLP over [expected digits, pairwise digit products, RNS residue
simplices (fixed differentiable mixing over Z_p), N digits]. The cell that
learns one-step modular squaring from direct pairs (52% unseen at 14k rows,
day-20 screen) inside the exact-T ... | {
"id": "2a931fc1-c695-4c00-ad36-bb20b178657e",
"created_at": "2026-08-14 01:07:00.103995+00:00",
"db_md5": "d08ba312cd6460ffc449c5d2476626a5",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "d1b10d4f-909d-4313-bc9f-5d57791ce760",
"tier": "easy",
"dataset_id": "e5",
"status": "s... | {
"score": {
"mean_loss": 2.79105395017361,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.02666666560495893
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1897040793981253,
"example_count": 600,
... |
2a9547d1-e683-496e-b92f-ec317e805015 | easy | sapient-sapiens | 2026-08-23 09:00:36.749192+00:00 | succeeded | 0726663e4f49752ee3210f0a40ae78f4ff0bc41713f2181b2b26a1386d5c9a76 | 21,190 | 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": "2a9547d1-e683-496e-b92f-ec317e805015",
"created_at": "2026-08-23 09:00:36.749192+00:00",
"db_md5": "37dce468bfe6f05f05d0eb1b30ebec43",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "e113bd75-2238-4252-92b5-2cb877313a01",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.1470103733684325,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.008333333358168602
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1499373328792672,
"example_count": 60... |
2a9d4498-ea13-436d-9436-45d809bc916c | easy | DDanlov | 2026-08-11 19:37:36.262404+00:00 | succeeded | adcb3a79c3a810a1a8935a170d03877216f9a8b22322b46acc703a61bb8d7d61 | 20,010 | 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": "2a9d4498-ea13-436d-9436-45d809bc916c",
"created_at": "2026-08-11 19:37:36.262404+00:00",
"db_md5": "1c8eff3d2f7d298c85c823356387b737",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "ed6729fc-284d-41a3-aea6-df98cd8a63bc",
"tier": "easy",
"dataset_id": "e1",
"status": "succeeded"... | {
"score": {
"mean_loss": 7.864478390710312,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333333333335
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 7.839626789093018,
"example_count": 100... |
2aa8dd68-9994-4680-bb47-f6c42d7d57e3 | easy | khushidahi | 2026-08-08 02:35:15.329198+00:00 | succeeded | 1d6425326993879de867a818e9f56b5d960798de6e55dc11f793372de0cb326a | 13,291 | null | """R42 Fourier-number representation experiment.
All input-dependent numeric parsing stays inside differentiable GPU tensor
operations. This is a learned-model representation, not a task solver: it does
not compute modular squaring or create intermediate labels.
"""
from __future__ import annotations
import math
impor... | {
"id": "2aa8dd68-9994-4680-bb47-f6c42d7d57e3",
"created_at": "2026-08-08 02:35:15.329198+00:00",
"db_md5": "3689afe74b87613f06d55f3b96bfac6b",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "12bb002c-17d4-4be6-a6b4-0f1bab5873e5",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.163823366165161,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.006666666828095913
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1710469722747803,
"example_count": 600... |
2aa9cddb-7d59-405f-972b-720dee612cec | easy | yashkant | 2026-08-12 04:29:43.899694+00:00 | succeeded | fcb6e4f4719005cd2391ce4c5fe2e3637adbe575d6494caf8c70cbf0ad8d441c | 35,038 | null | """Round1542 low-capacity periodic-fold answer/latent state machine.
A width-32 answer/latent pair is refined by one shared hidden-16 MLP once per
public-T macro. The candidate advances each field through a differentiable
triangle-wave fold, giving bounded carry-free wraparound; the equal-work
control uses a conventio... | {
"id": "2aa9cddb-7d59-405f-972b-720dee612cec",
"created_at": "2026-08-12 04:29:43.899694+00:00",
"db_md5": "e6e641a83d0464513d5fee5a23fb1f78",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "2fb55aaa-0769-4cf6-8b92-742fd46c5f12",
"tier": "easy",
"dataset_id": "e3",
"status": "succeed... | {
"score": {
"mean_loss": 2.128362542975862,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.016250000158324836
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.128176080112353,
"example_count": 800,... |
2abe9430-90af-41ac-9fdc-0a53a21ccc19 | easy | Bananafly | 2026-08-12 09:27:10.629671+00:00 | succeeded | 20593456fd79e120f88fdb5348ea113ed901c363b82f7fe9f5a505011eb163c6 | 11,652 | null | """Easy-3 T-aligned soft residue model with a 1e-3 learning rate."""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
assert_model_state,
)
D_M... | {
"id": "2abe9430-90af-41ac-9fdc-0a53a21ccc19",
"created_at": "2026-08-12 09:27:10.629671+00:00",
"db_md5": "d6bad392558cd3276cdfa989f323f97d",
"submitter": "Andre Kreidemann",
"github_login": "Bananafly",
"run_id": "2795b946-1818-4598-b9de-88df23a815c0",
"tier": "easy",
"dataset_id": "e1",
"status": ... | {
"score": {
"mean_loss": 2.502195119857788,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.497101068496704,
"example_count": 100,
"exact... |
2ac0850a-3b88-45d3-9ec7-df313d5a271f | easy | AFeinfield | 2026-08-28 00:58:34.170871+00:00 | succeeded | 9dceecd2a07ddd5d981d32edaee3dcfbe3d139911b4ddae7d9b3b60b0c8128e6 | 15,662 | null | """178: FRONTIER COMPOSITE for the long-run test — affine decode +
local-N carry context + scheduled snap (threshold 2000 calls, sized for
m-tier step counts). The 60s band is operator-precision-limited;
this is the composed best of the night given 5-10x optimization.
The frontier's test accuracy equals per-op accurac... | {
"id": "2ac0850a-3b88-45d3-9ec7-df313d5a271f",
"created_at": "2026-08-28 00:58:34.170871+00:00",
"db_md5": "28451d79d47054a02215d00222b012d8",
"submitter": "AFeinfield",
"github_login": "AFeinfield",
"run_id": "d6a8f4aa-715f-44ee-8281-ed22e8583c1f",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 2.2036129236221313,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0833333320915699
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.215914487838745,
"example_count": 100,
... |
2ac609d8-8064-4d9a-8f4c-e7613fb425d2 | easy | lpbb | 2026-08-14 11:06:27.917168+00:00 | succeeded | 0c9d08e07fd7be6ff2a64865fecd2a2f17afb6b6c580de3dcab32d695259de96 | 9,158 | null | """CHEAT L1g — stack test: grokfast (+5pts alone, L1d) AND canonical hint
min(s, N-s) (2x effective coverage, L1e). Do independent gains add?"""
from __future__ import annotations
import math
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
Optimiz... | {
"id": "2ac609d8-8064-4d9a-8f4c-e7613fb425d2",
"created_at": "2026-08-14 11:06:27.917168+00:00",
"db_md5": "4b9157872cae5c2792cb0176898bd7b7",
"submitter": "Lpbb",
"github_login": "lpbb",
"run_id": "adc6bd0e-d945-4481-b4ed-00bbd6631f9f",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",
"s... | {
"score": {
"mean_loss": 6.325162216847945,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.09812500002328306
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.487591830531639,
"example_count": 800,
... |
2acc1f2f-ec5e-4e3f-b0f1-20ef60f53e56 | easy | DDanlov | 2026-08-12 23:07:08.901257+00:00 | failed | 88cac49667d55d1dcececf8243a1a64244f82f24a9dadb35998a7baed8af1f2d | 28,966 | null | from __future__ import annotations
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as cp
from torch.optim.optimizer import Optimizer
try:
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
... | {
"id": "2acc1f2f-ec5e-4e3f-b0f1-20ef60f53e56",
"created_at": "2026-08-12 23:07:08.901257+00:00",
"db_md5": "b0cd394390a4839b67dc8fc2c7fa0cbc",
"submitter": "DDanlov",
"github_login": "DDanlov",
"run_id": "97b0e6c7-40f1-4538-8391-a08ec0ee9ba5",
"tier": "easy",
"dataset_id": "e3",
"status": "failed",
... | null
|
2ad07d0e-8b08-4aee-b6f0-800dfc2831e0 | easy | viridale | 2026-08-04 22:03:58.828838+00:00 | succeeded | 96b90a06c9282ff1501a6dd5665abb29d74e94b77f614e7055acd72f8fef8921 | 8,830 | null | """Clean autonomous recurrence with a worst-token sequence loss.
hypothesis: flat token-mean CE underweights the single weakest answer token even
though one wrong token fails the whole row and rung; a small smooth worst-token
barrier over evaluator-supplied valid labels should improve sequence exactness.
axis:... | {
"id": "2ad07d0e-8b08-4aee-b6f0-800dfc2831e0",
"created_at": "2026-08-04 22:03:58.828838+00:00",
"db_md5": "676e637a4d0f5ed8f7d057484c7b0412",
"submitter": "priormancer",
"github_login": "viridale",
"run_id": "fb793fba-1612-4aa3-ac6f-fe2c93cce6b8",
"tier": "easy",
"dataset_id": "e3",
"status": "succe... | {
"score": {
"mean_loss": 2.7754178156006493,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.00875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.8292126607433508,
"example_count": 800,
... |
2ae087a0-55e0-45f6-a34f-5e1955a45ee0 | easy | ehonig | 2026-08-14 00:42:25.229780+00:00 | succeeded | bed7b46c989217f150e8f32594973831db97cd435e186415766922012faf890c | 48,079 | null | """DigitLoop — a place-value recurrent model for repeated modular squaring.
The task is ``y = x^(2^T) mod N`` with the answer read back tail aligned, one
decimal place value per token position from the right. A stack of attention and
MLP layers has no primitive for the two operations this needs — a partial-product
co... | {
"id": "2ae087a0-55e0-45f6-a34f-5e1955a45ee0",
"created_at": "2026-08-14 00:42:25.229780+00:00",
"db_md5": "8a71b19028930edd1dd4facf7e650670",
"submitter": "Edouardo Honig",
"github_login": "ehonig",
"run_id": "58781e07-311e-44b3-a82c-6899b1190495",
"tier": "easy",
"dataset_id": "e1",
"status": "succ... | {
"score": {
"mean_loss": 5.6651225090026855,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.1783333383500576
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.05467700958252,
"example_count": 100,
... |
2ae22f4a-b49f-4f84-bc34-b1b140717fac | easy | jordanrubin | 2026-08-03 19:15:00.621965+00:00 | succeeded | df9b5a2f8d9a42f915a6d6089db0be744b5335b3b1b33342265e1113b1af3af4 | 15,389 | 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": "2ae22f4a-b49f-4f84-bc34-b1b140717fac",
"created_at": "2026-08-03 19:15:00.621965+00:00",
"db_md5": "5293d88573a9b2b00409dde3fa572f5d",
"submitter": "Jordan Rubin",
"github_login": "jordanrubin",
"run_id": "6421485f-afda-4706-9d05-14690aba953d",
"tier": "easy",
"dataset_id": "e2",
"status": "s... | {
"score": {
"mean_loss": 4.944475607931077,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.10291666522001226
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.394316359817912,
"example_count": 300,
... |
2ae3fdc3-ada6-4550-b9ab-4256e1f7b151 | easy | khushidahi | 2026-08-11 13:36:51.503951+00:00 | succeeded | 927795b0639617131b30c045972c25f87c7a60152e4cd00d2db658b1830f1c13 | 24,960 | null | """R51 learned bilinear diagonal-routing model.
The architecture uses a learned low-rank pair feature for decimal-place slots
i and j, then a parameter-free scatter-add to slot i+j. The pair features,
carry propagation, modulus conditioning, recurrent update, and output decoder
are all learned end-to-end.
No digit mu... | {
"id": "2ae3fdc3-ada6-4550-b9ab-4256e1f7b151",
"created_at": "2026-08-11 13:36:51.503951+00:00",
"db_md5": "2ef804661d8d3d43d81c0e3fb57030ae",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "e318b04c-64e8-4f2d-b072-050892842afb",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 2.1614421380382254,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.007916666666666667
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1627271838252318,
"example_count": 60... |
2ae5a457-0878-4877-aef5-f5d7c5d55ec0 | easy | yashkant | 2026-08-09 11:25:22.096736+00:00 | succeeded | 91caade4c93ef64c2a76ef25ae28e03f84a16b28730a75497fab1e1eac5ba570 | 19,621 | null | """Round 532 low-rank width513 adapter with AdamW 3.0e-3."""
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 Tenso... | {
"id": "2ae5a457-0878-4877-aef5-f5d7c5d55ec0",
"created_at": "2026-08-09 11:25:22.096736+00:00",
"db_md5": "d26dd1a446896c4b63f87764e24969aa",
"submitter": "Yash Kant",
"github_login": "yashkant",
"run_id": "f09ffc05-550c-4689-a539-959d96732520",
"tier": "easy",
"dataset_id": "e2",
"status": "succeed... | {
"score": {
"mean_loss": 4.762805700302124,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.3162500225007534
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 8.04906940460205,
"example_count": 300,
... |
2aebd620-98d6-456c-bde0-38576753b7b4 | easy | khushidahi | 2026-08-17 00:17:46.158544+00:00 | succeeded | b1755e95105002e287a6d6d8b7a66b4b174ee983f80cbed8c3c499fb36659c54 | 24,768 | null | """R180: compliant ordered/global transition with scaled annealing.
This candidate is deliberately diagnostic. It trains on evaluator-owned final
labels at several depths so that one shared transition must compose. The model
never converts N or X to native numbers and never implements multiplication or
modular reduc... | {
"id": "2aebd620-98d6-456c-bde0-38576753b7b4",
"created_at": "2026-08-17 00:17:46.158544+00:00",
"db_md5": "d8e3ca5a7c09f0411c4bca99d7b2890e",
"submitter": "khushidahi",
"github_login": "khushidahi",
"run_id": "4e261577-09d8-4e69-a115-de2b0ddb7e47",
"tier": "easy",
"dataset_id": "e5",
"status": "succ... | {
"score": {
"mean_loss": 3.189851559926587,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.048333333370586234
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 3.323679658864111,
"example_count": 600,... |
2aefaa63-d641-4bd8-a3a2-1c55b0215010 | easy | Maxwells-Demons | 2026-08-06 05:22:49.185304+00:00 | succeeded | fec53b8a0d5c47a70d8859047e85357f9cf3f30e307a846be91f3bc1baa39b66 | 13,690 | null | """Value-bottleneck recurrent model, e2 configuration (v13 champion).
Same architecture as the e1 submission (pinned digit head, leading-zero
supervision) plus SAM-style extra parameter updates inside the permitted
between_backward_passes callback (transforms parameters/optimizer state in
place; never calls optimizer.... | {
"id": "2aefaa63-d641-4bd8-a3a2-1c55b0215010",
"created_at": "2026-08-06 05:22:49.185304+00:00",
"db_md5": "42a3f1008f1320eba52b14d69f0f77a2",
"submitter": "Maxwells_Ayakashi",
"github_login": "Maxwells-Demons",
"run_id": "9735ee66-7373-4ca1-94f1-4c7b2ee6cb7a",
"tier": "easy",
"dataset_id": "e2",
"st... | {
"score": {
"mean_loss": 2.308551549911499,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.6520833373069763
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1996991634368896,
"example_count": 300,
... |
2af207e9-1de9-4074-918d-3d9cc489aaeb | easy | sapient-sapiens | 2026-08-29 21:57:07.241855+00:00 | failed | 86e7cdeaff03beb2310025e73e3a23a193fe7e9ce55eaf50cc0b663894565118 | 15,640 | null | """Three-loop one-layer Transformer for iterated modular squaring.
The learned arithmetic operator is one tied pre-norm Transformer block. Every
square makes four Horner-digit calls and every digit makes nine micro-calls.
N and the complete operand are immutable during a square; scratch resets at a
digit boundary; an... | {
"id": "2af207e9-1de9-4074-918d-3d9cc489aaeb",
"created_at": "2026-08-29 21:57:07.241855+00:00",
"db_md5": "b8320b00601055cf9694a737faf46c43",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "4d2f04a4-6761-483b-89d6-93ad9f8dd818",
"tier": "easy",
"dataset_id": "e5",
"status": "fa... | null
|
2af32780-dcba-4abd-a0aa-fdd97f14c465 | easy | sapient-sapiens | 2026-08-31 01:02:06.587065+00:00 | succeeded | a95b222d1145d60f1fab7e8e907bf68340ff032d7b025370b099297c65bda240 | 18,146 | null | """Eligible E5 direct-Horner sharply categorical boundary selector.
This preserves the conservative one-block direct-Horner recurrence: at each of
four Horner positions one tied call writes latent S and a second tied call
writes A. It ties all eight square boundaries to the corresponding categorical digit-code\nfamil... | {
"id": "2af32780-dcba-4abd-a0aa-fdd97f14c465",
"created_at": "2026-08-31 01:02:06.587065+00:00",
"db_md5": "78de764609bbe1512cc816ff64e99566",
"submitter": "Ertondy",
"github_login": "sapient-sapiens",
"run_id": "5ae500b1-b907-4795-a1bb-e4e5faa28533",
"tier": "easy",
"dataset_id": "e5",
"status": "su... | {
"score": {
"mean_loss": 2.1408511188492407,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.005833333333333334
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1415477526027527,
"example_count": 60... |
2afb1a33-ece9-4e9e-a114-216f8194ef4c | easy | erdavis0 | 2026-08-19 09:10:14.827550+00:00 | succeeded | 66f43a8197ab60870db44110e82f14f2ec7d693e1142c22c5979df7bcbfc72c9 | 8,021 | null | """Parser-free neural cellular transducer with a blank recurrent state."""
from __future__ import annotations
import math
import time
import torch
import torch.nn.functional as F
from torch import Tensor, nn
from benchmark import (
ModelSpec,
OptimizerBundle,
OptimizerSpec,
Submission,
TokenLoss... | {
"id": "2afb1a33-ece9-4e9e-a114-216f8194ef4c",
"created_at": "2026-08-19 09:10:14.827550+00:00",
"db_md5": "cc818629b0d212733078e49ff9630eae",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "b55d1e31-df1a-44e9-be48-506ecc78a3a2",
"tier": "easy",
"dataset_id": "e3",
"status": "succeeded",... | {
"score": {
"mean_loss": 2.872429646949337,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.011875
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.934359168625199,
"example_count": 800,
"... |
2afc746a-1d9d-46f4-a148-1ece55cfc586 | easy | liam-gb | 2026-08-13 01:02:21.354740+00:00 | succeeded | 6bf2cc11d65028a86e224ae6baf38a24805bf944ae234df1ce81d716e73b7ff2 | 14,798 | null | """rns_pm_gather_ct_widet_lrt10: widet with lookup tables at lr 1e-2.
The wide readout is what makes long answers reachable; the trim is what pays
for it, by not running loop iterations whose update is multiplied by zero.
Base docstring:
Single change vs rns_pm_gather_ct: the readout carries one
distance-from-units ... | {
"id": "2afc746a-1d9d-46f4-a148-1ece55cfc586",
"created_at": "2026-08-13 01:02:21.354740+00:00",
"db_md5": "7b4955e5ad1f251972eab97f3e3fa745",
"submitter": "liam-gb",
"github_login": "liam-gb",
"run_id": "0bc687ea-fa27-49d9-88b1-8e9699afda22",
"tier": "easy",
"dataset_id": "e5",
"status": "succeeded"... | {
"score": {
"mean_loss": 6.055377111711481,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.04208333356305957
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 6.343290328979492,
"example_count": 600,
... |
2afcccd8-970c-44ff-820f-0a2fe5a9366c | easy | himalalps | 2026-08-10 01:36:49.493349+00:00 | succeeded | 60cb5e04a009e2cdf36939e557b993a23f96f550ecc367912a943488d7daf7cc | 84,553 | null | """Algorithmically constrained latent recurrence for One Layer Deeper.
One masked cross-attention pass creates two independent representations:
``state_0`` from X only and an immutable context from N only. T is excluded
from both representations and is used solely to choose how many times the
learned recurrent MLP tr... | {
"id": "2afcccd8-970c-44ff-820f-0a2fe5a9366c",
"created_at": "2026-08-10 01:36:49.493349+00:00",
"db_md5": "6a4df6cf29a0759353b8741412fc97fb",
"submitter": "Haoyu Tang",
"github_login": "himalalps",
"run_id": "74e30e63-434a-4ab6-812e-694f93d81ed9",
"tier": "easy",
"dataset_id": "e5",
"status": "succe... | {
"score": {
"mean_loss": 2.175652820549838,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.010833333370586236
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 2.1828114986419678,
"example_count": 600... |
2b039a04-ca14-45ac-a05f-9abbb61d2368 | easy | isaac0804 | 2026-08-06 15:44:36.254579+00:00 | succeeded | af65d70b03b96b0d25ed3ce19dd713901f0c98d21184752b8ec88c34f42a0ae0 | 9,267 | null | """E3 candidate #7: same-modulus batch cross-attention -- an in-context /
retrieval mechanism, structurally different from every N-conditioning
attempt tried so far.
Every prior attempt this session (e3_hyper_lora, e3_hard_routing,
e3_vq_bottleneck, e3_narrow_bottleneck) conditioned each row on N using only
that row's... | {
"id": "2b039a04-ca14-45ac-a05f-9abbb61d2368",
"created_at": "2026-08-06 15:44:36.254579+00:00",
"db_md5": "1940badf501ffeaad17ff7ffb6a410bf",
"submitter": "Isaac Yong",
"github_login": "isaac0804",
"run_id": "59fb6dc9-f42b-494d-a124-baaf7ea100b0",
"tier": "easy",
"dataset_id": "e4",
"status": "succe... | {
"score": {
"mean_loss": 9.920833752914664,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.0033333333457509677
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 10.14943687037918,
"example_count": 120... |
2b056d98-1bc6-46a6-8611-ff7093e46ca4 | easy | erdavis0 | 2026-08-19 09:24:11.988948+00:00 | succeeded | e6d75591b123483c27b5a4344aa991e325184d79f5ed87c2a9fa7d7390f5f5bd | 10,552 | null | """Generic learned-slot transducer with a shared low-rank latent operator."""
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,
TokenL... | {
"id": "2b056d98-1bc6-46a6-8611-ff7093e46ca4",
"created_at": "2026-08-19 09:24:11.988948+00:00",
"db_md5": "fb39c040e0caf73a4ee6e4c2f9f4d564",
"submitter": "Ethan",
"github_login": "erdavis0",
"run_id": "fca89bcc-6d35-432c-9f1b-b1cf7d0d09ba",
"tier": "easy",
"dataset_id": "e8",
"status": "succeeded",... | {
"score": {
"mean_loss": 8.226687669754028,
"primary_metric": "mean_exact_accuracy",
"num_measurements": 2,
"mean_exact_accuracy": 0.06390374526381493
},
"seeds": [
{
"seed": 74,
"evaluation": {
"ood": {
"loss": 9.042009353637695,
"example_count": 85,
... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.