SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.03113 · 2021

Scaling Scaling Laws with Board Games

arXiv · PDF · Open in the Atlas

Code that ran

13 of 20 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 13 have no contract verdict; 0 failures of our instrument; 7 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2104.03113.json · badge: /api/ran/2104.03113.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
andyljones/boardlaw community (classifier label)
classifier basis not recorded on the link
13 ran · 7 no recorded run
FunctionStatusWhere it lives
as_sgf Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/mohex.py
code served (permissive licence) · get_code("fc7d2bf32c4f8c81")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small observation tensor: (S, S, 2) where S=5
# Shape: (5, 5, 2) - 5x5 board with 2 channels (black and white pieces)
obs = torch.zeros(5, 5, 2, dtype=torch.float32)

# Add some pieces to make it interesting
obs[0, 0, 0] = 1  # Black piece at (0, 0)
obs[1, 2, 0] = 1  # Black piece at (1, 2)
obs[2, 3, 1] = 1  # White piece at (2, 3)
obs[3, 1, 1] = 1  # White piece at (3, 1)

# seat can be 0 or 1
seat = 0

ARGS = {
    'obs': obs,
    'seat': seat,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
combine_actions Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/analysis.py
code served (permissive licence) · get_code("a945e25cb4a786b7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (4, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

class Decision:
    def __init__(self, actions):
        self.actions = actions

# Create decisions dict with Decision objects that have .actions attribute
decisions = {
    'agent_1': Decision(torch.randn(2, 4)),
    'agent_2': Decision(torch.randn(2, 4)),
}

# Create masks dict with boolean masks for indexing
masks = {
    'agent_1': torch.tensor([True, False, True, False]),
    'agent_2': torch.tensor([False, True, False, True]),
}

ARGS = {
    'decisions': decisions,
    'masks': masks,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (4, 4) — shape (4, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
configfile Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/mohex.py
code served (permissive licence) · get_code("c5bfe30be69a2cb9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from tempfile import NamedTemporaryFile

torch.manual_seed(0)

ARGS = {
    'max_games': 50,
    'max_memory': 2.5,
    'presearch': True,
    'max_time': 60.0,
    'max_nodes': 10000,
    'solver': True,
    'extras': ['param_mohex reuse_subtree 1', 'param_mohex rave_weight_initial 1000']
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
expand Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/analysis.py
code served (permissive licence) · get_code("c1f40ed24b504f3a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Construct a small exemplar tensor
# Using float dtype with shape (batch=2, features=4)
exemplar = torch.randn(2, 4, dtype=torch.float32, device='cpu')

# n_envs is an integer specifying number of environments
n_envs = 3

ARGS = {
    'exemplar': exemplar,
    'n_envs': n_envs,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (3, 4) — shape (3, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
flops_savepoints Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/storage.py
code served (permissive licence) · get_code("6991e3c1ba7a74a3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family float, shape (21,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

ARGS = {
    "boardsize": 5,
    "n_snapshots": 21,
    "upper": None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (21,) — shape (21,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mix Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/learning.py
code served (permissive licence) · get_code("4012a2148ebeacb0")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned WorldsStandIn. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

class WorldsStandIn:
    """Mock worlds object that tracks state and provides step() and valid attribute"""
    def __init__(self, batch_size=2, num_actions=4):
        self.batch_size = batch_size
        self.num_actions = num_actions
        # valid should be a probability distribution over actions for each world
        self.valid = torch.ones(batch_size, num_actions) / num_actions
        self.state = torch.randn(batch_size, 3)
        self.step_count = 0
    
    def step(self, actions):
        """
        Takes actions and returns (new_worlds, transitions)
        actions: tensor of shape (batch_size,) with action indices
        """
        self.step_count += 1
        
        # Create new worlds state
        new_worlds = WorldsStandIn(self.batch_size, self.num_actions)
        new_worlds.state = self.state + torch.randn(self.batch_size, 3) * 0.1
        new_worlds.valid = torch.ones(self.batch_size, self.num_actions) / self.num_actions
        
        # Create transitions object (just needs to exist)
        transitions = {
            'actions': actions,
            'rewards': torch.randn(self.batch_size),
            'dones': torch.zeros(self.batch_size, dtype=torch.bool)
        }
        
        return new_worlds, transitions


# Construct valid arguments
ARGS = {
    'worlds': WorldsStandIn(batch_size=2, num_actions=4),
    'T': 5  # small number of iterations for testing
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type WorldsStandIn — a WorldsStandIn output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a WorldsStandIn output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
rel_entropy Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/learning.py
code served (permissive licence) · get_code("3f45524f680d14f5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function takes logits as input
# It uses:
# - logits > -np.inf (comparison with numpy inf)
# - torch.zeros_like(logits)
# - logits.where(valid, zeros) (tensor method)
# - logits.exp() (tensor method)
# - sum(-1) (sum over last dimension)
# - mean() (mean reduction)
# - torch.log()
# - valid.sum(-1).float()

# Create a small logits tensor (batch_size=2, vocab_size=4)
logits = torch.randn(2, 4, dtype=torch.float32)

# Make sure it has some valid and some invalid values
logits[0, 0] = -np.inf
logits[1, 2] = -np.inf

ARGS = {
    "logits": logits
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
scatter_values Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/heads.py
code served (permissive licence) · get_code("367074d89cc98d11")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (2, 4, 2) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Construct valid arguments for scatter_values
# The function expects:
# - v: a tensor of values to be scattered
# - seats: a tensor of indices (0 or 1) indicating where to place values

# Create a small tensor v with shape (2, 4)
v = torch.randn(2, 4)

# Create seats tensor with values 0 or 1, shape (2, 4)
# This indicates which position (0 or 1) each value should go to
seats = torch.randint(0, 2, (2, 4), dtype=torch.float32)

ARGS = {
    "v": v,
    "seats": seats
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 4, 2) — shape (2, 4, 2) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
solve Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/elos.py
code served (permissive licence) · get_code("36de63ed4f7ee176")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Series. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import pandas as pd
import numpy as np

torch.manual_seed(0)

# Create a small symmetric DataFrame with wins and games data
# Both must have matching index and columns (square DataFrames for rating systems)
teams = ['TeamA', 'TeamB', 'TeamC']

# wins: number of wins in head-to-head matchups
wins_data = {
    'TeamA': [0, 2, 1],
    'TeamB': [1, 0, 3],
    'TeamC': [2, 0, 0]
}
wins = pd.DataFrame(wins_data, index=teams)

# games: total number of games played
games_data = {
    'TeamA': [0, 3, 2],
    'TeamB': [3, 0, 4],
    'TeamC': [2, 4, 0]
}
games = pd.DataFrame(games_data, index=teams)

ARGS = {
    'wins': wins,
    'games': games,
    'prior': 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Series — a Series output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Series output the recorded ndim 1 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
symmetrize Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/elos.py
code served (permissive licence) · get_code("834bf0198db420c4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 3, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import pandas as pd

# Create a sample trials DataFrame with the required columns
trials = pd.DataFrame({
    'black_agent': ['agent_a', 'agent_a', 'agent_b', 'agent_b', 'agent_c'],
    'white_agent': ['agent_b', 'agent_c', 'agent_a', 'agent_c', 'agent_a'],
    'black_wins': [5, 3, 2, 4, 1],
    'white_wins': [3, 2, 6, 2, 7]
})

ARGS = {
    'trials': trials
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
time_savepoints Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/storage.py
code served (permissive licence) · get_code("2e13a5fddcf146ef")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family float, shape (21,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

ARGS = {
    "boardsize": 7,
    "n_snapshots": 21,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (21,) — shape (21,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
to_notation Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/mohex.py
code served (permissive licence) · get_code("4b902338baafdbbc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Analyzing the function:
# def to_notation(subscript):
#     row, col = subscript
#     col = chr(ord('a') + col)
#     return f'{col}{row + 1}'
#
# The function takes a single parameter 'subscript' which is unpacked into row and col.
# It expects subscript to be a tuple or list-like with exactly 2 elements.
# row should be an integer (used in row + 1)
# col should be an integer (used as ord('a') + col, so 0-25 for a-z)

torch.manual_seed(0)

# Create a valid subscript: a tuple of (row, col)
subscript = (3, 5)  # row=3, col=5 -> outputs 'f4' (chr(ord('a')+5)='f', 3+1=4)

ARGS = {
    "subscript": subscript
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
uniform_logits Ran
ran on a synthesized input
andyljones/boardlaw/boardlaw/validation.py
code served (permissive licence) · get_code("d9741f5b1322e325")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function expects 'valid' to be a tensor
# It calls: valid.float(), valid.sum(-1, keepdims=True)
# So 'valid' should be a tensor with at least 1 dimension
# Let's create a small 2D tensor (batch_size=2, num_classes=4)
valid = torch.tensor([[1, 1, 1, 1], [2, 2, 2, 2]], dtype=torch.long)

ARGS = {
    "valid": valid
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
elo_errors Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/elos.py
code served (permissive licence) · get_code("a642ea35e15e247c")
How it ran

no invocation recorded.

flops_per_sample Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/storage.py
code served (permissive licence) · get_code("e2bd1b373ebc52bb")
How it ran

no invocation recorded.

intake Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/heads.py
code served (permissive licence) · get_code("0ed1dad28421148d")
How it ran

no invocation recorded.

load Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/cuda.py
code served (permissive licence) · get_code("af40ded331a17cf3")
How it ran

no invocation recorded.

load_cpu Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/cuda.py
code served (permissive licence) · get_code("4d4db23102df95b8")
How it ran

no invocation recorded.

load_cuda Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/cuda.py
code served (permissive licence) · get_code("f665e38eed24b93d")
How it ran

no invocation recorded.

output Not yet run
unverified (harvested, no recorded run)
andyljones/boardlaw/boardlaw/heads.py
code served (permissive licence) · get_code("ffd08286e32eb008")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 20 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The largest experiments in machine learning now require resources far beyond the budget of all but a few institutions. Fortunately, it has recently been shown that the results of these huge experiments can often be extrapolated from the results of a sequence of far smaller, cheaper experiments. In this work, we show that not only can the extrapolation be done based on the size of the model, but on the size of the problem as well. By conducting a sequence of experiments using AlphaZero and Hex, we show that the performance achievable with a fixed amount of compute degrades predictably as the game gets larger and harder. Along with our main result, we further show that the test-time and train-time compute available to an agent can be traded off while maintaining performance.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2104.03113")
get_code_for_paper("2104.03113")
have("2104.03113")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2104.03113.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2104.03113.svg)](https://syntology.ai/paper/2104.03113)

Connect an agent — have() is free.