SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.07402 · ACL · 2022

Revisiting the Compositional Generalization Abilities of Neural Sequence Models

Phil Blunsom, Navin Goyal, Satwik Bhattamishra, Arkil Patel

arXiv · PDF · Open in the Atlas

Code that ran

2 of 2 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 2 have no contract verdict; 0 failures of our instrument; 0 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/2203.07402.json · badge: /api/ran/2203.07402.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
arkilpatel/compositional-generalization-seq2seq canonical (classifier label)
classifier basis not recorded on the link
2 ran · 0 no recorded run
FunctionStatusWhere it lives
bert_sim Ran
ran on a synthesized input
arkilpatel/compositional-generalization-seq2seq/code/lstm/src/confidence_estimation.py
code served (permissive licence) · get_code("e377a9696d9833a7")
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 2, dtype family float, shape (2, 3) 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
import numpy as np

torch.manual_seed(0)

class MockBertModel(nn.Module):
    """Minimal BERT-like model that returns [CLS] token representation"""
    def __init__(self, hidden_dim=4):
        super().__init__()
        self.hidden_dim = hidden_dim
        self.embedding = nn.Embedding(1000, hidden_dim)
        self.linear = nn.Linear(hidden_dim, hidden_dim)
    
    def forward(self, input_ids):
        """
        Args:
            input_ids: tensor of shape [batch_size, seq_len]
        Returns:
            tuple where first element is tensor of shape [batch_size, seq_len, hidden_dim]
        """
        # Embed input tokens
        embedded = self.embedding(input_ids)  # [batch_size, seq_len, hidden_dim]
        # Apply linear transformation
        output = self.linear(embedded)  # [batch_size, seq_len, hidden_dim]
        return (output,)

class MockBertSeq2Exp(nn.Module):
    """Mock bert_seq2exp model with a bert attribute"""
    def __init__(self):
        super().__init__()
        self.bert = MockBertModel(hidden_dim=4)

# Create mock model
model = MockBertSeq2Exp()

# Create mock query batch: [batch_size=2, seq_len=3]
queries = torch.randint(0, 100, (2, 3))

# Create mock key batch: [batch_size=3, seq_len=3]
keys = torch.randint(0, 100, (3, 3))

ARGS = {
    "queries": queries,
    "keys": keys,
    "model": model
}
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 ndarray, ndim 2, dtype family float, shape (2, 3) — shape (2, 3) 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
posterior_based_conf Ran
ran on a synthesized input
arkilpatel/compositional-generalization-seq2seq/code/lstm/src/confidence_estimation.py
code served (permissive licence) · get_code("7f20db9fa1afeca3")
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 numpy as np

torch.manual_seed(0)

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def greedy_decode(self, test_ques, return_probs=False):
        # Simulate decoding: return decoded words and log probabilities
        batch_size = 2
        seq_length = 3
        
        # decoded_words: list of lists of word strings
        decoded_words = [
            ["hello", "world", "end"],
            ["foo", "bar", "stop"]
        ]
        
        # decoded_log_probs: list of lists of log probabilities
        # Each inner list corresponds to log probs for one sequence
        decoded_log_probs = [
            [-0.5, -0.3, -0.2],
            [-0.4, -0.6, -0.1]
        ]
        
        if return_probs:
            return decoded_words, decoded_log_probs
        else:
            return decoded_words

# Create a mock test input (can be any format since it's just passed through)
test_ques = torch.randn(2, 4)

# Create the model
model = MockModel()

ARGS = {
    "test_ques": test_ques,
    "model": model
}
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 — 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

Also present in other papers' code

Of 2 samples lifted from this paper's linked repositories, 2 carry a function identity and 0 do not. Of the 2 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 2 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

Compositional generalization is a fundamental trait in humans, allowing us to effortlessly combine known phrases to form novel sentences. Recent works have claimed that standard seqto-seq models severely lack the ability to compositionally generalize. In this paper, we focus on one-shot primitive generalization as introduced by the popular SCAN benchmark. We demonstrate that modifying the training distribution in simple and intuitive ways enables standard seq-to-seq models to achieve nearperfect generalization performance, thereby showing that their compositional generalization abilities were previously underestimated. We perform detailed empirical analysis of this phenomenon. Our results indicate that the generalization performance of models is highly sensitive to the characteristics of the training data which should be carefully considered while designing such benchmarks in future.

For agents

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

get_harvested_code_for_paper("2203.07402")
get_code_for_paper("2203.07402")
have("2203.07402")

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

curl https://syntology.ai/api/ran/2203.07402.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.07402.svg)](https://syntology.ai/paper/2203.07402)

Connect an agent — have() is free.