SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.08691 · EMNLP · 2021

The Power of Scale for Parameter-Efficient Prompt Tuning

Rami Al-Rfou, Noah Constant, Brian Lester

arXiv · PDF · Open in the Atlas

Code that ran

13 of 15 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 9 have no contract verdict; 4 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/2104.08691.json · badge: /api/ran/2104.08691.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
VE-FORBRYDERNE/mtj-softtuner community (classifier label)
classifier basis not recorded on the link
9 ran · 1 no recorded run
exelents/soft-prompt-tuning community (classifier label)
classifier basis not recorded on the link
3 ran · 1 no recorded run
RossNordby/SoftPromptsForEvaluation community (classifier label)
classifier basis not recorded on the link
1 ran · 0 no recorded run
FunctionStatusWhere it lives
core_dummy_initialized Ran
ran on a synthesized input
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/testing_utils.py
code served (permissive licence) · get_code("16ba263208ecf384")
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 function. 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 functools
from typing import Callable, TypeVar

torch.manual_seed(0)

# Define a minimal mock core module
class CoreModule:
    def __init__(self):
        self.test_mode = False
        self.initialized = False

core = CoreModule()

T = TypeVar('T', bound=Callable)

def core_test_mode(f: T) -> T:

    @functools.wraps(f)
    def decorated(*a, **k):
        test_mode_orig = core.test_mode
        core.test_mode = True
        try:
            r = f(*a, **k)
        finally:
            core.test_mode = test_mode_orig
        return r
    return decorated

def core_dummy_initialized(f: T) -> T:

    @functools.wraps(f)
    def decorated(*a, **k):
        initialized_orig = core.initialized
        core.initialized = True
        try:
            r = f(*a, **k)
        finally:
            core.initialized = initialized_orig
        return r
    return core_test_mode(decorated)

# Create a simple test function to decorate
def sample_function(x, y):
    """A simple function to be decorated"""
    return x + y

# Apply the decorator
decorated_fn = core_dummy_initialized(sample_function)

# Define ARGS with parameters for the decorated function
ARGS = {
    'f': sample_function,
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
core_test_mode Ran
ran on a synthesized input
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/testing_utils.py
code served (permissive licence) · get_code("659a594a04601c37")
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 function. 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 functools
from typing import Callable, TypeVar
import torch

T = TypeVar('T', bound=Callable)

# Mock core module to simulate the environment
class MockCore:
    test_mode = False

core = MockCore()

def core_test_mode(f: T) -> T:
    @functools.wraps(f)
    def decorated(*a, **k):
        test_mode_orig = core.test_mode
        core.test_mode = True
        try:
            r = f(*a, **k)
        finally:
            core.test_mode = test_mode_orig
        return r
    return decorated

# Create a simple test function to decorate
def sample_function(x, y):
    """A simple function that will be decorated by core_test_mode"""
    return x + y

# Setup code
torch.manual_seed(0)

# The function core_test_mode takes one parameter: a callable (function)
# We need to provide a function that can be decorated
test_func = lambda a, b: a * b

ARGS = {
    'f': test_func
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
get_config_from_root Ran
ran on a synthesized input
VE-FORBRYDERNE/mtj-softtuner/versioneer.py
code served (permissive licence) · get_code("51944f53542b7e19")
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 VersioneerConfig. 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 configparser
import os
import tempfile

class VersioneerConfig:
    """Container for Versioneer configuration parameters."""
    pass

# Create a temporary directory with a setup.cfg file
temp_dir = tempfile.mkdtemp()
setup_cfg_path = os.path.join(temp_dir, 'setup.cfg')

setup_cfg_content = """[versioneer]
VCS = git
style = pep440
versionfile_source = myproject/_version.py
versionfile_build = myproject/_version.py
tag_prefix = v
parentdir_prefix = myproject-
verbose = False
"""

with open(setup_cfg_path, 'w') as f:
    f.write(setup_cfg_content)

ARGS = {
    'root': temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type VersioneerConfig — a VersioneerConfig output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a VersioneerConfig output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
is_time_to_exit Ran
ran on a synthesized input
exelents/soft-prompt-tuning/utils.py
code served (permissive licence) · get_code("641dddbb7b534717")
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 bool, 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

class Args:
    def __init__(self):
        self.max_steps_per_epoch = 100
        self.max_steps = 1000

ARGS = {
    'args': Args(),
    'epoch_steps': 0,
    'global_steps': 0
}
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 bool, 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
label_smoothed_nll_loss Ran
ran on a synthesized input
exelents/soft-prompt-tuning/utils1.py
code served (permissive licence) · get_code("e386d3a24af4d168")
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

torch.manual_seed(0)

# Create log probabilities tensor
# Shape: (batch_size=2, vocab_size=4)
# lprobs should be log probabilities, so values between -inf and 0
lprobs = torch.log(torch.softmax(torch.randn(2, 4), dim=-1))

# Create target indices tensor
# Shape: (batch_size=2,) - indices into the vocabulary
# Values should be in range [0, vocab_size) or ignore_index
target = torch.tensor([1, 2], dtype=torch.long)

# Epsilon for label smoothing (typically small, e.g., 0.1)
epsilon = 0.1

# Ignore index for padding tokens
ignore_index = -100

ARGS = {
    "lprobs": lprobs,
    "target": target,
    "epsilon": epsilon,
    "ignore_index": ignore_index,
}
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
lmap Ran
ran; our contract draft was wrong, not the code
exelents/soft-prompt-tuning/utils1.py
code served (permissive licence) · get_code("7c43af131a6f39a6")
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
from typing import Callable, List, Iterable

torch.manual_seed(0)

# Define a simple callable function
def square(x):
    """Simple function that squares its input"""
    return x ** 2

# Define an iterable (list of numbers)
numbers = [1, 2, 3, 4, 5]

# Construct ARGS dict
ARGS = {
    "f": square,
    "x": numbers,
}
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
normalize_chess_elos Ran
ran on a synthesized input
RossNordby/SoftPromptsForEvaluation/chess_training/dataset_loading.py
code served (permissive licence) · get_code("29fe77a79233bc67")
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, 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 torch

torch.manual_seed(0)

# Test case 1: Tensor input
tensor_elos = torch.tensor([[1200, 1400], [2000, 800]], dtype=torch.float32)

# Test case 2: List of tuples input
list_elos = [(1200, 1400), (2000, 800)]

# Test case 3: Single tuple input
tuple_elos = (1200, 1400)

# We'll use the tensor input as the primary test case
ARGS = {
    "elos": tensor_elos
}
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, 2) — shape (2, 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
register_vcs_handler Ran
ran; our contract draft was wrong, not the code
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/_version.py
code served (permissive licence) · get_code("c55cafde08716f01")
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 function. 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.

# Global HANDLERS dict that the function uses
HANDLERS = {}

# Define arguments for register_vcs_handler
ARGS = {
    "vcs": "git",
    "method": "commit"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
register_vcs_handler Ran
ran on a synthesized input
VE-FORBRYDERNE/mtj-softtuner/versioneer.py
code served (permissive licence) · get_code("9dc618d67058e3e6")
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 function. 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

# Global HANDLERS dict that the function uses
HANDLERS = {}

# Build arguments for register_vcs_handler(vcs, method)
# The function takes a vcs identifier and a method name
# It returns a decorator that stores functions in HANDLERS[vcs][method]

ARGS = {
    "vcs": "git",
    "method": "clone"
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
restore_variable Ran
ran on a synthesized input
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/serialization.py
code served (permissive licence) · get_code("e4eaafa9797969b1")
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 dict. 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 os
import pickle
import tempfile
import shutil

# Create a temporary directory structure to simulate saved variables
temp_dir = tempfile.mkdtemp()
universe_num = 42
var_name = "test_variable"
folder = os.path.join(temp_dir, '.mtjs' + str(universe_num))
os.makedirs(folder, exist_ok=True)

# Create a test variable and save it
test_data = {"key": "value", "number": 123, "list": [1, 2, 3]}
path = os.path.join(folder, var_name)
with open(path, 'wb') as f:
    pickle.dump(test_data, f)

# Change to temp directory so relative paths work
original_cwd = os.getcwd()
os.chdir(temp_dir)

try:
    ARGS = {
        "universe": universe_num,
        "name": var_name,
    }
finally:
    # Note: In actual execution, the function would be called with these args
    # We keep the temp directory available during function execution
    pass
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
run_command Ran
ran; our contract draft was wrong, not the code
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/_version.py
code served (permissive licence) · get_code("dbb37355631dabcf")
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 subprocess
import sys

ARGS = {
    "commands": ["echo"],
    "args": ["hello", "world"],
    "cwd": None,
    "verbose": False,
    "hide_stderr": False,
    "env": None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
shatter Ran
ran on a synthesized input
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/core.py
code served (permissive licence) · get_code("33d302ec4e7efc18")
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 function. 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 functools
import jax
import jax.dlpack
from typing import Callable, Optional, TypeVar

# Define a simple test function to decorate
def simple_func(x, params):
    """Simple function that takes a sharded array and a broadcast parameter."""
    return x + params

# Call the shatter decorator with valid arguments
ARGS = {
    'in_axes': 'sb',
    'out_axes': 's'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
versions_from_parentdir Ran
ran; our contract draft was wrong, not the code
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/_version.py
code served (permissive licence) · get_code("5032e3b81c0fffb3")
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 dict. 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 os
import tempfile

ARGS = {
    'parentdir_prefix': 'myproject-',
    'root': '/tmp/myproject-1.2.3/src',
    'verbose': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
core_partly_initialized Not yet run
unverified (harvested, no recorded run)
VE-FORBRYDERNE/mtj-softtuner/mtj_softtuner/testing_utils.py
code served (permissive licence) · get_code("3be93347d7ab5153")
How it ran

no invocation recorded.

get_model_config_tokenizer Not yet run
unverified (harvested, no recorded run)
exelents/soft-prompt-tuning/gptj_utils.py
code served (permissive licence) · get_code("04adf1e29b3ea4b7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 7 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 3 the same month, 7 later) and 8 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
register_vcs_handler 26 other papers
11 dated earlier, 1 the same month, 14 later
1806.00550 (2018-06)
A Swiss Army Infinitesimal Jackknife
1806.03146 (2018-06)
Neural Message Passing with Edge Updates for Predicting Properties of Molecules and Materials
1806.03720 (2018-06)
Stochastic seismic waveform inversion using generative adversarial networks as a geological prior
and 23 more not named here
lmap 9 other papers
1 dated earlier, 1 the same month, 7 later
2007.08426 (2020-07)
Investigating Pretrained Language Models for Graph-to-Text Generation
2104.08400 (2021-04)
Structure-Aware Abstractive Conversation Summarization via Discourse and Action Graphs
2105.04779 (2021-05)
EL-Attention: Memory Efficient Lossless Attention for Generation
and 6 more not named here
versions_from_parentdir 9 other papers
1 dated earlier, 8 later
1806.03146 (2018-06)
Neural Message Passing with Edge Updates for Predicting Properties of Molecules and Materials
2204.04516 (2022-04)
Uncertainty-Informed Deep Learning Models Enable High-Confidence Predictions for Digital Histopathology
2205.10242 (2022-05)
EXODUS: Stable and Efficient Training of Spiking Neural Networks
and 6 more not named here
label_smoothed_nll_loss 7 other papers
1 dated earlier, 1 the same month, 5 later
2007.08426 (2020-07)
Investigating Pretrained Language Models for Graph-to-Text Generation
2104.08400 (2021-04)
Structure-Aware Abstractive Conversation Summarization via Discourse and Action Graphs
2105.04779 (2021-05)
EL-Attention: Memory Efficient Lossless Attention for Generation
and 4 more not named here
run_command 7 other papers
1 dated earlier, 6 later
1806.03146 (2018-06)
Neural Message Passing with Edge Updates for Predicting Properties of Molecules and Materials
2204.04516 (2022-04)
Uncertainty-Informed Deep Learning Models Enable High-Confidence Predictions for Digital Histopathology
2205.10242 (2022-05)
EXODUS: Stable and Efficient Training of Spiking Neural Networks
and 4 more not named here
get_config_from_root 2 other papers
1 dated earlier, 1 later
1806.03146 (2018-06)
Neural Message Passing with Edge Updates for Predicting Properties of Molecules and Materials
2207.09250 (2022-07)
Roughness of molecular property landscapes and its impact on modellability
register_vcs_handler 2 other papers
2 later
2205.10242 (2022-05)
EXODUS: Stable and Efficient Training of Spiking Neural Networks
2209.12332 (2022-09)
On the Optimal Linear Contraction Order of Tree Tensor Networks, and Beyond

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 this work, we explore "prompt tuning," a simple yet effective mechanism for learning "soft prompts" to condition frozen language models to perform specific downstream tasks. Unlike the discrete text prompts used by GPT-3, soft prompts are learned through backpropagation and can be tuned to incorporate signals from any number of labeled examples. Our end-to-end learned approach outperforms GPT-3's few-shot learning by a large margin. More remarkably, through ablations on model size using T5, we show that prompt tuning becomes more competitive with scale: as models exceed billions of parameters, our method "closes the gap" and matches the strong performance of model tuning (where all model weights are tuned). This finding is especially relevant because large models are costly to share and serve and the ability to reuse one frozen model for multiple downstream tasks can ease this burden. Our method can be seen as a simplification of the recently proposed "prefix tuning" of Li and Liang (2021) and we provide a comparison to this and other similar approaches. Finally, we show that conditioning a frozen model with soft prompts confers benefits in robustness to domain transfer and enables efficient "prompt ensembling." We release code and model checkpoints to reproduce our experiments. 1

For agents

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

get_harvested_code_for_paper("2104.08691")
get_code_for_paper("2104.08691")
have("2104.08691")

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

curl https://syntology.ai/api/ran/2104.08691.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.08691.svg)](https://syntology.ai/paper/2104.08691)

Connect an agent — have() is free.