SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.11867 · ACL · 2020

Improving Massively Multilingual Neural Machine Translation and Zero-Shot Translation

Biao Zhang, Ivan Titov, Rico Sennrich, Philip Williams

arXiv · PDF · Open in the Atlas

Code that ran

2 of 13 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; 2 failures of our instrument; 11 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/2004.11867.json · badge: /api/ran/2004.11867.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
bzhangGo/zero marked official in Papers with Code
Papers with Code archive
2 ran · 11 no recorded run
FunctionStatusWhere it lives
combine_heads Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("9faf4852d7f74a91")
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 EagerTensor, ndim 3, dtype family float, shape (2, 4, 15) 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 tensorflow as tf

# Create a small tensor with shape [batch, heads, length, channels]
# Using distinct sizes: batch=2, heads=3, length=4, channels=5
inputs = tf.constant([
    [[[1.0, 2.0, 3.0, 4.0, 5.0],
      [6.0, 7.0, 8.0, 9.0, 10.0],
      [11.0, 12.0, 13.0, 14.0, 15.0],
      [16.0, 17.0, 18.0, 19.0, 20.0]],
     [[21.0, 22.0, 23.0, 24.0, 25.0],
      [26.0, 27.0, 28.0, 29.0, 30.0],
      [31.0, 32.0, 33.0, 34.0, 35.0],
      [36.0, 37.0, 38.0, 39.0, 40.0]],
     [[41.0, 42.0, 43.0, 44.0, 45.0],
      [46.0, 47.0, 48.0, 49.0, 50.0],
      [51.0, 52.0, 53.0, 54.0, 55.0],
      [56.0, 57.0, 58.0, 59.0, 60.0]]],
    [[[61.0, 62.0, 63.0, 64.0, 65.0],
      [66.0, 67.0, 68.0, 69.0, 70.0],
      [71.0, 72.0, 73.0, 74.0, 75.0],
      [76.0, 77.0, 78.0, 79.0, 80.0]],
     [[81.0, 82.0, 83.0, 84.0, 85.0],
      [86.0, 87.0, 88.0, 89.0, 90.0],
      [91.0, 92.0, 93.0, 94.0, 95.0],
      [96.0, 97.0, 98.0, 99.0, 100.0]],
     [[101.0, 102.0, 103.0, 104.0, 105.0],
      [106.0, 107.0, 108.0, 109.0, 110.0],
      [111.0, 112.0, 113.0, 114.0, 115.0],
      [116.0, 117.0, 118.0, 119.0, 120.0]]]
], dtype=tf.float32)

ARGS = {
    'inputs': inputs,
    'name': 'test_combine_heads'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 3, dtype family float, shape (2, 4, 15) — shape (2, 4, 15) 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
split_heads Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("688a6272dc140cd0")
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 EagerTensor, ndim 4, dtype family float, shape (2, 2, 3, 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 tensorflow as tf
import numpy as np

# Create a small input tensor with shape [batch, length, channels]
# batch=2, length=3, channels=4
batch_size = 2
length = 3
channels = 4
num_heads = 2

inputs = tf.constant(np.random.randn(batch_size, length, channels).astype(np.float32))

ARGS = {
    "inputs": inputs,
    "num_heads": num_heads,
    "name": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 4, dtype family float, shape (2, 2, 3, 2) — shape (2, 2, 3, 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
add_timing_signal Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("4b52e600cf72c348")
How it ran

no invocation recorded.

attention_bias Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("b88b2db9eb46e73d")
How it ran

no invocation recorded.

decoder Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("c01bf5a6fe99ddfb")
How it ran

no invocation recorded.

dot_attention Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("b04d0b5aeab65a69")
How it ran

no invocation recorded.

encoder Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("dd6b86752ec524cd")
How it ran

no invocation recorded.

extract_encodes Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("5b2126720aede56f")
How it ran

no invocation recorded.

ffn_layer Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("36e7645b57365d0d")
How it ran

no invocation recorded.

layer_norm Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("9dcd5100e4896669")
How it ran

no invocation recorded.

linear Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("c9710c0dc3de38fe")
How it ran

no invocation recorded.

residual_fn Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("a58e7c7393bd5449")
How it ran

no invocation recorded.

train_fn Not yet run
unverified (harvested, no recorded run)
bzhangGo/zero/models/transformer_l0drop.py
code served (permissive licence) · get_code("81602c4d87412b16")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 10 are also present in other papers whose code we hold (9 with at least one paper dated earlier, 0 the same month, 10 later) and 3 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.

FunctionAlso present inEarliest-datedOthers named
combine_heads
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later · 4 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2104.07012 (2021-04)
Sparse Attention with Linear Units
and 1 more not named here
layer_norm
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later · 4 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2104.07012 (2021-04)
Sparse Attention with Linear Units
and 1 more not named here
linear
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later · 4 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2104.07012 (2021-04)
Sparse Attention with Linear Units
and 1 more not named here
split_heads
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later · 4 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2104.07012 (2021-04)
Sparse Attention with Linear Units
and 1 more not named here
add_timing_signal
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later · 3 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2305.01778 (2023-05)
SLTUNET: A Simple Unified Model for Sign Language Translation
attention_bias
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later · 3 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2305.01778 (2023-05)
SLTUNET: A Simple Unified Model for Sign Language Translation
ffn_layer
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later · 3 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2305.01778 (2023-05)
SLTUNET: A Simple Unified Model for Sign Language Translation
residual_fn
brick of this paper's assembly: helper
3 other papers
1 dated earlier, 2 later · 3 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
2305.01778 (2023-05)
SLTUNET: A Simple Unified Model for Sign Language Translation
encoder
brick of this paper's assembly: model
2 other papers
1 dated earlier, 1 later · 1 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation
dot_attention
brick of this paper's assembly: model
1 other paper
1 later · 1 only as an assembly brick
2010.08518 (2020-10)
Adaptive Feature Selection for End-to-End Speech Translation

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

Massively multilingual models for neural machine translation (NMT) are theoretically attractive, but often underperform bilingual models and deliver poor zero-shot translations. In this paper, we explore ways to improve them. We argue that multilingual NMT requires stronger modeling capacity to support language pairs with varying typological characteristics, and overcome this bottleneck via language-specific components and deepening NMT architectures. We identify the off-target translation issue (i.e. translating into a wrong target language) as the major source of the inferior zero-shot performance, and propose random online backtranslation to enforce the translation of unseen training language pairs. Experiments on OPUS-100 (a novel multilingual dataset with 100 languages) show that our approach substantially narrows the performance gap with bilingual models in both oneto-many and many-to-many settings, and improves zero-shot performance by ∼10 BLEU, approaching conventional pivot-based methods. 1

For agents

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

get_harvested_code_for_paper("2004.11867")
get_code_for_paper("2004.11867")
have("2004.11867")

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

curl https://syntology.ai/api/ran/2004.11867.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.11867.svg)](https://syntology.ai/paper/2004.11867)

Connect an agent — have() is free.