SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.07207 · ICML · 2021

Straight to the Gradient: Learning to Use Novel Tokens for Neural Text Generation

Shafiq Joty, Xiang Lin, Simeng Han

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; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 4 failures of our instrument; 1 no recorded run), measured 2026-11-22; 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/2106.07207.json · badge: /api/ran/2106.07207.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
shawnlimn/ScaleGrad canonical (classifier label)
classifier basis: census
4 ran · 0 no recorded run
shawnlimn/scalegrad canonical (classifier label)
classifier basis: classifier_2025
3 ran · 1 no recorded run
FunctionStatusWhere it lives
batch_input_sequence_by_prefix_length Ran
ran on a synthesized input
shawnlimn/ScaleGrad/custom/evaluate_utils.py
code served (permissive licence) · get_code("e91bbe0f0d67f102")
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 (32, 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)

# The function expects:
# - input_sequence: a tensor with at least 2 dimensions, where dim 1 is the sequence length
# - prefix_length: an integer used to reshape and batch the sequence

# Create a small input tensor: shape (batch_size=2, seq_len=8, feature_dim=4)
input_sequence = torch.randn(2, 8, 4)

# Create a prefix_length that divides evenly into seq_len for clean batching
prefix_length = 2

ARGS = {
    "input_sequence": input_sequence,
    "prefix_length": prefix_length,
}
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 (32, 2) — shape (32, 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
get_datasets Ran
ran; our contract draft was wrong, not the code
shawnlimn/scalegrad/custom/gpt2/run_gpt2.py
code served (permissive licence) · get_code("041742ed31858d0f")
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 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 torch
import tempfile
import os

torch.manual_seed(0)

# Create temporary files with small tensors
temp_dir = tempfile.mkdtemp()

# Create train, valid, test tensors
# The function expects tensors that will be reshaped to (-1, max_len + 1)
# So we need tensors with total elements divisible by (max_len + 1)
max_len = 1536
chunk_size = max_len + 1  # 1537

# Create tensors with enough elements
# For simplicity, create tensors with 3074 elements (2 * 1537)
train_tensor = torch.randn(3074)
valid_tensor = torch.randn(3074)
test_tensor = torch.randn(3074)

train_path = os.path.join(temp_dir, 'train.pt')
valid_path = os.path.join(temp_dir, 'valid.pt')
test_path = os.path.join(temp_dir, 'test.pt')

torch.save(train_tensor, train_path)
torch.save(valid_tensor, valid_path)
torch.save(test_tensor, test_path)

ARGS = {
    'dataset_paths': {
        'train': train_path,
        'valid': valid_path,
        'test': test_path
    },
    'max_len': 1536
}
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 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
ngram_metrics Ran
ran on a synthesized input
shawnlimn/ScaleGrad/custom/metrics.py
code served (permissive licence) · get_code("d8c07afc39ab5d5e")
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 defaultdict. 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 numpy as np
from collections import defaultdict, Counter
from nltk import ngrams

torch.manual_seed(0)

# The function takes:
# - token_list: a list that can contain tokens/integers
# - pad: an integer (default 1) used to truncate the token_list

# Construct a token_list with some repeated n-grams
# Using integers as tokens for simplicity
token_list = [5, 6, 7, 5, 6, 8, 9, 5, 6, 10]

# The pad parameter is already defaulted to 1, but we can specify it
pad_value = 1

ARGS = {
    'token_list': token_list,
    'pad': pad_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
nltk, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type defaultdict — a defaultdict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a defaultdict 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
ngram_repeat_mask Ran
ran on a synthesized input
shawnlimn/ScaleGrad/custom/sequence_penalty_loss.py
code served (permissive licence) · get_code("b34473338b0463d5")
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 int, shape (2, 6) 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)

# xs: a 2D tensor of integers where each row is a sequence
# The function iterates over rows, converts to list, and checks for n-gram repeats
# n: the n-gram size to check for repeats
xs = torch.tensor([
    [1, 2, 3, 2, 3, 4],
    [5, 6, 7, 5, 6, 8]
], dtype=torch.long)

n = 2

ARGS = {
    "xs": xs,
    "n": n
}
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 int, shape (2, 6) — shape (2, 6) 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
top_k_logits Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
shawnlimn/scalegrad/custom/sequence_generator.py
code served (permissive licence) · get_code("298689919da2c5b0")
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
import math

torch.manual_seed(0)

# Create small logits tensor: batch_size=2, vocab_size=4
logits = torch.randn(2, 4)

# Set k to a valid value (less than vocab dimension)
k = 2

ARGS = {
    "logits": logits,
    "k": k,
}
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
top_k_logits Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
shawnlimn/ScaleGrad/custom/sequence_generator.py
pointer only (licence: MIT); the stored copy get_code serves, from facebookresearch/unlikelihood_training, is pointer only (licence: NOASSERTION) · get_code("ace3f464bf5b4c7a")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from facebookresearch/unlikelihood_training, and GitHub could not classify the licence of that repository (NOASSERTION). 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.

top_k_top_p_filtering Ran
ran; our contract draft was wrong, not the code
shawnlimn/scalegrad/custom/gpt2/run_gpt2.py
code served (permissive licence) · get_code("d415ae4787477f20")
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 1, dtype family float, shape (8,) 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.functional as F

torch.manual_seed(0)

# Create a small logits tensor with shape (1, vocabulary_size)
# The function expects logits of shape (vocabulary_size) after squeezing
# So we create shape (1, 8) to represent batch_size=1, vocab_size=8
logits = torch.randn(1, 8)

ARGS = {
    "logits": logits,
    "top_k": 3,
    "top_p": 0.9,
    "filter_value": -float('Inf')
}
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 1, dtype family float, shape (8,) — shape (8,) 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
sample_sequence Not yet run
unverified (harvested, no recorded run)
shawnlimn/scalegrad/custom/gpt2/run_gpt2.py
code served (permissive licence) · get_code("19a19ef6f1042ac4")
How it ran

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 7 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 6 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
top_k_logits 1 other paper
1 dated earlier
1908.04319 (2019-08)
Neural Text Generation with Unlikelihood Training

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

Advanced large-scale neural language models have led to significant success in many language generation tasks. However, the most commonly used training objective, Maximum Likelihood Estimation (MLE), has been shown problematic, where the trained model prefers using dull and repetitive phrases. In this work, we introduce ScaleGrad, a modification straight to the gradient of the loss function, to remedy the degeneration issue of the standard MLE objective. By directly maneuvering the gradient information, ScaleGrad makes the model learn to use novel tokens. Empirical results show the effectiveness of our method not only in open-ended generation, but also in directed generation tasks. With the simplicity in architecture, our method can serve as a general training objective that is applicable to most of the neural text generation tasks.

For agents

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

get_harvested_code_for_paper("2106.07207")
get_code_for_paper("2106.07207")
have("2106.07207")

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

curl https://syntology.ai/api/ran/2106.07207.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.07207.svg)](https://syntology.ai/paper/2106.07207)

Connect an agent — have() is free.