SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2007.02924 · ICLR · 2021

INT: An Inequality Benchmark for Evaluating Generalization in Theorem Proving

Jimmy Ba, Yuhuai Wu, Albert Jiang, Roger Grosse

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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 5 have no contract verdict; 0 failures of our instrument; 2 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/2007.02924.json · badge: /api/ran/2007.02924.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
albertqjiang/INT canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
execute_according_to_dictionary Ran
ran on a synthesized input
albertqjiang/INT/int_environment/algos/eval_seq.py
code served (permissive licence) · get_code("c59747934355efa5")
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 float, ndim 0, dtype family float, shape () 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 torch.nn as nn

torch.manual_seed(0)

class MockParser:
    def observation_to_source(self, observation):
        return "source_key_1"

class MockProver:
    def __init__(self):
        self.parser = MockParser()
    
    def get_observation(self):
        return {"state": "observation_data"}
    
    def apply_theorem_seq_style(self, target):
        return {
            "progress": 0.75
        }

prover = MockProver()

dictionary = {
    "source_key_1": "theorem_target_1",
    "source_key_2": "theorem_target_2"
}

ARGS = {
    "prover": prover,
    "dictionary": dictionary
}
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 float, ndim 0, dtype family float, shape () — shape () 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
flatten Ran
ran on a synthesized input
albertqjiang/INT/int_environment/algos/lib/ops.py
code served (permissive licence) · get_code("3ff091098a71d9dd")
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 list, ndim 1, dtype family int; 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 'observations' and flattens it by extending a list with each element
# observations should be an iterable of iterables (e.g., list of lists)
# Each inner element will be extended into the result list

observations = [
    [1, 2, 3],
    [4, 5, 6],
    [7, 8]
]

ARGS = {
    "observations": observations
}
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 list, ndim 1, dtype family int — a list 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
load_model Ran
ran on a synthesized input
albertqjiang/INT/int_environment/visualization/model_debug.py
code served (permissive licence) · get_code("ea3d9cdbcf7a9b4e")
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. 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
import json
import tempfile
import os

torch.manual_seed(0)

# Create a temporary directory with the required env_config.json
temp_dir = tempfile.mkdtemp()
directory = temp_dir + '/'

# Create a minimal env_config.json
env_config_data = {
    'learning_rate': 0.001,
    'batch_size': 2,
    'epochs': 1,
    'hidden_dim': 4
}

with open(directory + 'env_config.json', 'w') as f:
    json.dump(env_config_data, f)

# Create a minimal initialize function that the load_model function calls
def initialize(env_config):
    # Minimal return values matching what load_model unpacks
    data_points = [(torch.randn(2, 4), torch.randn(2, 1)) for _ in range(2)]
    batch_size = 2
    
    # Minimal neural network
    class IneqNet(nn.Module):
        def __init__(self):
            super().__init__()
            self.linear = nn.Linear(4, 1)
        
        def forward(self, x):
            return self.linear(x)
    
    ineq_net = IneqNet()
    optimizer = torch.optim.Adam(ineq_net.parameters(), lr=0.001)
    lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=1)
    
    update = 0
    training_losses = []
    lemma_acces = []
    ent_acces = []
    name_acces = []
    train_succ_rates = []
    val_succ_rates = []
    saving_directory = directory
    
    return (data_points, batch_size, ineq_net, optimizer, lr_scheduler, 
            update, training_losses, lemma_acces, ent_acces, name_acces, 
            train_succ_rates, val_succ_rates, saving_directory)

# Inject initialize into the module namespace for load_model to use
import sys
current_module = sys.modules[__name__]
current_module.initialize = initialize

ARGS = {
    'directory': directory
}
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 tuple — a tuple output has no array shape of its own; none is 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
one_hot Ran
ran on a synthesized input
albertqjiang/INT/int_environment/algos/thm_env.py
code served (permissive licence) · get_code("d760315462516b6b")
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)

# Construct arguments for one_hot function
batch = torch.tensor([0, 1, 2, 1], dtype=torch.long)  # batch of indices
depth = 4  # number of classes/categories

ARGS = {
    "batch": batch,
    "depth": depth
}
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
one_hot Ran
ran on a synthesized input
albertqjiang/INT/int_environment/algos/lib/ops.py
code served (permissive licence) · get_code("a36605330ae69c8d")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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)

device = torch.device('cpu')

# batch: a tensor of indices that will be used with index_select
# The indices should be valid indices into a tensor of size [depth]
batch = torch.tensor([0, 1, 2, 1], dtype=torch.long)

# depth: an integer representing the number of classes for one-hot encoding
depth = 4

ARGS = {
    "batch": batch,
    "depth": depth,
}
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
load_checkpoint Not yet run
unverified (harvested, no recorded run)
albertqjiang/INT/int_environment/algos/rl.py
code served (permissive licence) · get_code("95e11c5901e291dd")
How it ran

no invocation recorded.

resume_checkpoint Not yet run
unverified (harvested, no recorded run)
albertqjiang/INT/int_environment/algos/rl.py
code served (permissive licence) · get_code("720f82c493f038b5")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 7 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

In learning-assisted theorem proving, one of the most critical challenges is to generalize to theorems unlike those seen at training time. In this paper, we introduce INT, an INequality Theorem proving benchmark designed to test agents' generalization ability. INT is based on a theorem generator, which provides theoretically infinite data and allows us to measure 6 different types of generalization, each reflecting a distinct challenge, characteristic of automated theorem proving. In addition, INT provides a fast theorem proving environment with sequence-based and graph-based interfaces, conducive to performing learning-based research. We introduce baselines with architectures including transformers and graph neural networks (GNNs) for INT. Using INT, we find that transformer-based agents achieve stronger test performance for most of the generalization tasks, despite having much larger outof-distribution generalization gaps than GNNs. We further find that the addition of Monte Carlo Tree Search (MCTS) at test time helps to prove new theorems.

For agents

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

get_harvested_code_for_paper("2007.02924")
get_code_for_paper("2007.02924")
have("2007.02924")

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

curl https://syntology.ai/api/ran/2007.02924.json

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

[![Syntology run record](https://syntology.ai/api/ran/2007.02924.svg)](https://syntology.ai/paper/2007.02924)

Connect an agent — have() is free.