SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2212.04231 · 2022

Harnessing the Power of Multi-Task Pretraining for Ground-Truth Level Natural Language Explanations

arXiv · PDF · Open in the Atlas

Code that ran

7 of 8 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; 1 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 1 failure of our instrument; 1 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/2212.04231.json · badge: /api/ran/2212.04231.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
ofa-x/ofa-x canonical (classifier label)
classifier basis not recorded on the link
7 ran · 1 no recorded run
FunctionStatusWhere it lives
apply_half Ran
ran, honoured the contract we drafted
ofa-x/ofa-x/OFA/evaluate.py
code served (permissive licence) · get_code("357d3a4f0be679d9")
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 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 apply_half takes a single parameter 't' which should be a tensor
# It checks if t.dtype is torch.float32 and converts to torch.half if so
# Otherwise returns t unchanged

# Create a small float32 tensor to test the conversion
t = torch.randn(2, 4, dtype=torch.float32)

ARGS = {
    "t": t
}
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 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
construct_rdrop_sample Ran
ran on a synthesized input
ofa-x/ofa-x/OFA/criterions/label_smoothed_encouraging_loss.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from OFA-Sys/OFA, is pointer only (licence: Apache-2.0) · get_code("e1397e09cb4c249d")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from OFA-Sys/OFA, and Apache-2.0, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

custom_to_pil Ran
ran; our contract draft was wrong, not the code
ofa-x/ofa-x/OFA/criterions/clip_scst_loss.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from OFA-Sys/OFA, is pointer only (licence: Apache-2.0) · get_code("01478032e19f9e68")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from OFA-Sys/OFA, and Apache-2.0, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

kl_loss Ran
ran on a synthesized input
ofa-x/ofa-x/OFA/criterions/label_smoothed_encouraging_loss.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from OFA-Sys/OFA, is pointer only (licence: Apache-2.0) · get_code("e7939f8552fd8729")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from OFA-Sys/OFA, and Apache-2.0, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

label_smoothed_nll_loss Ran
ran on a synthesized input
ofa-x/ofa-x/OFA/criterions/label_smoothed_encouraging_loss.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from OFA-Sys/OFA, is pointer only (licence: Apache-2.0) · get_code("cbd8615e3eba731b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from OFA-Sys/OFA, and Apache-2.0, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

label_smoothed_nll_loss Ran
ran on a synthesized input
ofa-x/ofa-x/OFA/criterions/label_smoothed_cross_entropy.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from taokz/biomedgpt, is pointer only (licence: Apache-2.0) · get_code("cdbc222d0057fb31")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from taokz/biomedgpt, and Apache-2.0, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

label_smoothed_nll_loss Ran
ran on a synthesized input
ofa-x/ofa-x/OFA/criterions/label_smoothed_cross_entropy_expl.py
code served (permissive licence) · get_code("3a3448cb08438214")
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 torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for the function
batch_size = 2
vocab_size = 4

# lprobs: log probabilities, shape (batch_size, vocab_size)
lprobs = F.log_softmax(torch.randn(batch_size, vocab_size), dim=-1)

# target: token indices, shape (batch_size,)
target = torch.randint(0, vocab_size, (batch_size,))

# epsilon: label smoothing parameter
epsilon = 0.1

# update_num: training step counter
update_num = 100

# reduce: whether to reduce loss
reduce = True

# drop_worst_ratio: ratio of worst examples to drop
drop_worst_ratio = 0.0

# drop_worst_after: only drop after this many updates
drop_worst_after = 0

# use_rdrop: whether to use R-Drop regularization
use_rdrop = False

# reg_alpha: regularization weight
reg_alpha = 1.0

# constraint_masks: optional masks for constraining which tokens can be selected
constraint_masks = None

# constraint_start and constraint_end: range for constraint
constraint_start = None
constraint_end = None

ARGS = {
    'lprobs': lprobs,
    'target': target,
    'epsilon': epsilon,
    'update_num': update_num,
    'reduce': reduce,
    'drop_worst_ratio': drop_worst_ratio,
    'drop_worst_after': drop_worst_after,
    'use_rdrop': use_rdrop,
    'reg_alpha': reg_alpha,
    'constraint_masks': constraint_masks,
    'constraint_start': constraint_start,
    'constraint_end': constraint_end,
}
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, 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
scst_loss Not yet run
unverified (harvested, no recorded run)
ofa-x/ofa-x/OFA/criterions/clip_scst_loss.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from OFA-Sys/OFA, is pointer only (licence: Apache-2.0) · get_code("b0d5d75f27b8ce4f")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from OFA-Sys/OFA, and Apache-2.0, as GitHub detects it on that repository, is in the permissive set, but a licence decision keeps the body a pointer: its licence file names an outside party, so it may cover vendored code rather than the repository owner's files (recorded as pointer:licence_names_third_party_corporation). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 7 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 7 later) and 1 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
custom_to_pil 8 other papers
8 later
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks
2306.02115 (2023-06)
Table and Image Generation for Investigating Knowledge of Entities in Pre-trained Vision and Language Models
and 5 more not named here
apply_half 6 other papers
2 dated earlier, 4 later
2202.03052 (2022-02)
OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
2208.02532 (2022-08)
Prompt Tuning for Generative Multimodal Pretrained Models
2305.11172 (2023-05)
ONE-PEACE: Exploring One General Representation Model Toward Unlimited Modalities
and 3 more not named here
construct_rdrop_sample 2 other papers
2 later
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks
2306.02115 (2023-06)
Table and Image Generation for Investigating Knowledge of Entities in Pre-trained Vision and Language Models
kl_loss 2 other papers
2 later
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks
2306.02115 (2023-06)
Table and Image Generation for Investigating Knowledge of Entities in Pre-trained Vision and Language Models
label_smoothed_nll_loss 2 other papers
2 later
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks
2306.02115 (2023-06)
Table and Image Generation for Investigating Knowledge of Entities in Pre-trained Vision and Language Models
scst_loss 2 other papers
2 later
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks
2306.02115 (2023-06)
Table and Image Generation for Investigating Knowledge of Entities in Pre-trained Vision and Language Models
label_smoothed_nll_loss 1 other paper
1 later
2305.17100 (2023-05)
BiomedGPT: A Generalist Vision-Language Foundation Model for Diverse Biomedical Tasks

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

Natural language explanations promise to offer intuitively understandable explanations of a neural network's decision process in complex vision-language tasks, as pursued in recent VL-NLE models. While current models offer impressive performance on task accuracy and explanation plausibility, they suffer from a range of issues: Some models feature a modular design where the explanation generation module is poorly integrated with a separate module for task-answer prediction, employ backbone models trained on limited sets of tasks, or incorporate ad hoc solutions to increase performance on single datasets. We propose to evade these limitations by applying recent advances in large-scale multi-task pretraining of generative Transformer models to the problem of VL-NLE tasks. Our approach outperforms recent models by a large margin, with human annotators preferring the generated explanations over the ground truth in two out of three evaluated datasets. As a novel challenge in VL-NLE research, we propose the problem of multi-task VL-NLE and show that jointly training on multiple tasks can increase the explanation quality. We discuss the ethical implications of high-quality NLE generation and other issues in recent VL-NLE research.

For agents

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

get_harvested_code_for_paper("2212.04231")
get_code_for_paper("2212.04231")
have("2212.04231")

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

curl https://syntology.ai/api/ran/2212.04231.json

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

[![Syntology run record](https://syntology.ai/api/ran/2212.04231.svg)](https://syntology.ai/paper/2212.04231)

Connect an agent — have() is free.