SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2005.13111 · ACL · 2020

Rationalizing Text Matching: Learning Sparse Alignments via Optimal Transport

Kyle Swanson, Tao Lei, Lili Yu

arXiv · PDF · Open in the Atlas

Code that ran

4 of 4 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 0 have no contract verdict; 4 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/2005.13111.json · badge: /api/ran/2005.13111.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
asappresearch/rationale-alignment marked official in Papers with Code
Papers with Code archive
4 ran · 0 no recorded run
FunctionStatusWhere it lives
bmv Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
asappresearch/rationale-alignment/sinkhorn/sinkhorn.py
code served (permissive licence) · get_code("0ce59eaf7b896853")
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)

# Construct arguments for bmv function
# m: (num_batches x n1 x n2) - 3D tensor
# v: (num_batches x n2) - 2D tensor

num_batches = 2
n1 = 4
n2 = 3

m = torch.randn(num_batches, n1, n2)
v = torch.randn(num_batches, n2)

ARGS = {
    "m": m,
    "v": v,
}
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
compute_masks Ran
ran; our contract draft was wrong, not the code
asappresearch/rationale-alignment/sinkhorn/sinkhorn.py
code served (permissive licence) · get_code("fcbefbb7e4928ae5")
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

torch.manual_seed(0)

# C is a cost matrix: can be (n x m) or (num_batches x n x m)
# Using unbatched form: (n x m) where n=4, m=3
C = torch.randn(4, 3, dtype=torch.float32)

# a is row marginals: can be (n,) or (num_batches x n)
# Using unbatched form: (n,) where n=4
# Including some zeros and some non-zeros
a = torch.tensor([1.0, 2.0, 0.0, 1.5], dtype=torch.float32)

# b is column marginals: can be (m,) or (num_batches x m)
# Using unbatched form: (m,) where m=3
# Including some zeros and some non-zeros
b = torch.tensor([0.5, 0.0, 1.0], dtype=torch.float32)

ARGS = {
    "C": C,
    "a": a,
    "b": b,
}
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
mask_log Ran
ran; our contract draft was wrong, not the code
asappresearch/rationale-alignment/sinkhorn/sinkhorn.py
code served (permissive licence) · get_code("981f258607ac4ba8")
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)

# Create a small FloatTensor for the input x
# Shape: (2, 4) - batch size 2, feature dimension 4
x = torch.randn(2, 4).abs() + 0.1  # Add 0.1 to ensure all values are positive for log

# Create a mask tensor with 1s for content and 0s for padding
# Same shape as x: (2, 4)
mask = torch.tensor([
    [1.0, 1.0, 1.0, 0.0],
    [1.0, 1.0, 0.0, 0.0]
])

ARGS = {
    "x": x,
    "mask": mask
}
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
sinkhorn Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
asappresearch/rationale-alignment/sinkhorn/sinkhorn.py
code served (permissive licence) · get_code("8df4c2374ea066b4")
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) and returned Tensor, ndim 3, dtype family float, shape (2, 4, 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

torch.manual_seed(0)

# Create small tensors for testing
# C: cost matrix (num_batches x n x m) = (2 x 4 x 3)
C = torch.randn(2, 4, 3, dtype=torch.float32)

# a: row marginals (num_batches x n) = (2 x 4)
a = torch.ones(2, 4, dtype=torch.float32) / 4

# b: column marginals (num_batches x m) = (2 x 3)
b = torch.ones(2, 3, dtype=torch.float32) / 3

# alpha: initial log scaling for rows (num_batches x n) = (2 x 4)
alpha = torch.zeros(2, 4, dtype=torch.float32)

# beta: initial log scaling for columns (num_batches x m) = (2 x 3)
beta = torch.zeros(2, 3, dtype=torch.float32)

# epsilon: entropy regularization weight
epsilon = 0.01

# num_iter_max: maximum iterations
num_iter_max = 50

# unbalanced_lambda: weighting factor for unbalanced OT (None for balanced)
unbalanced_lambda = None

# error_threshold: convergence threshold
error_threshold = 1e-5

# error_check_frequency: how often to check error
error_check_frequency = 10

# return_log: whether to return logging info
return_log = False

ARGS = {
    'C': C,
    'a': a,
    'b': b,
    'alpha': alpha,
    'beta': beta,
    'epsilon': epsilon,
    'num_iter_max': num_iter_max,
    'unbalanced_lambda': unbalanced_lambda,
    'error_threshold': error_threshold,
    'error_check_frequency': error_check_frequency,
    'return_log': return_log,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 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

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 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 4 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

Selecting input features of top relevance has become a popular method for building selfexplaining models. In this work, we extend this selective rationalization approach to text matching, where the goal is to jointly select and align text pieces, such as tokens or sentences, as a justification for the downstream prediction. Our approach employs optimal transport (OT) to find a minimal cost alignment between the inputs. However, directly applying OT often produces dense and therefore uninterpretable alignments. To overcome this limitation, we introduce novel constrained variants of the OT problem that result in highly sparse alignments with controllable sparsity. Our model is end-to-end differentiable using the Sinkhorn algorithm for OT and can be trained without any alignment annotations. We evaluate our model on the Stack-Exchange, MultiNews, e-SNLI, and MultiRC datasets. Our model achieves very sparse rationale selections with high fidelity while preserving prediction accuracy compared to strong attention baseline models. †

For agents

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

get_harvested_code_for_paper("2005.13111")
get_code_for_paper("2005.13111")
have("2005.13111")

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

curl https://syntology.ai/api/ran/2005.13111.json

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

[![Syntology run record](https://syntology.ai/api/ran/2005.13111.svg)](https://syntology.ai/paper/2005.13111)

Connect an agent — have() is free.