SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.10209 · 2024

Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs

arXiv · PDF · Open in the Atlas

Code that ran

13 of 18 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; 3 honoured the contract we drafted, 3 violated it and 0 have no contract verdict; 7 failures of our instrument; 5 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/2406.10209.json · badge: /api/ran/2406.10209.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
ahans30/goldfish-loss canonical (classifier label)
classifier basis: census
13 ran · 5 no recorded run
FunctionStatusWhere it lives
apply_rope Ran
ran, violated the contract we drafted
ahans30/goldfish-loss/lit_gpt/model.py
code served (permissive licence) · get_code("aa32c6faaa562b69")
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 3, dtype family float, shape (2, 3, 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 tensors for apply_rope function
# x: input tensor with shape (batch, seq_len, head_size)
# The function expects x to have a last dimension (head_size) that can be split in half
batch_size = 2
seq_len = 3
head_size = 4

x = torch.randn(batch_size, seq_len, head_size)

# cos and sin tensors should have compatible shape for broadcasting
# They should be able to broadcast with x
cos = torch.randn(batch_size, seq_len, head_size)
sin = torch.randn(batch_size, seq_len, head_size)

ARGS = {
    "x": x,
    "cos": cos,
    "sin": sin,
}
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 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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
build_mask_cache Ran
ran, violated the contract we drafted
ahans30/goldfish-loss/lit_gpt/model.py
code served (permissive licence) · get_code("7ee1c3ba71156110")
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 4, dtype family float, shape (1, 1, 8, 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

torch.manual_seed(0)

ARGS = {
    "max_seq_length": 8,
    "device": None,
}
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 4, dtype family float, shape (1, 1, 8, 8) — shape (1, 1, 8, 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
cos_sim Ran
ran, honoured the contract we drafted
ahans30/goldfish-loss/lit_gpt/multiple_negative_ranking_loss.py
code served (permissive licence) · get_code("a8052087ba311802")
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, 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
from torch import nn, Tensor

torch.manual_seed(0)

# Create small tensors for testing
# a: shape (2, 4) - 2 vectors of dimension 4
a = torch.randn(2, 4)

# b: shape (3, 4) - 3 vectors of dimension 4
b = torch.randn(3, 4)

ARGS = {
    "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 Tensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 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
count_tokenizes_and_get_metrics Ran
ran; our contract draft was wrong, not the code
ahans30/goldfish-loss/data_checks/count_token_lengths.py
code served (permissive licence) · get_code("d88b4cf25c69af3e")
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 str. 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 numpy as np
from tqdm import tqdm

class SimpleTokenizer:
    def __call__(self, text):
        # Simple tokenizer that splits on whitespace and returns input_ids
        tokens = text.split()
        return {'input_ids': tokens}

class SimpleDataset:
    def __init__(self, texts):
        self.texts = texts
    
    def __len__(self):
        return len(self.texts)
    
    def __iter__(self):
        return iter(self.texts)
    
    def __getitem__(self, idx):
        return self.texts[idx]

# Create a simple dataset with text samples
dataset = SimpleDataset([
    {'text': 'hello world this is a test'},
    {'text': 'another sample text here'},
    {'text': 'short text'},
    {'text': 'this is a longer text sample with more words to tokenize'},
    {'text': 'final sample'}
])

# Create a simple tokenizer
tokenizer = SimpleTokenizer()

ARGS = {
    'dataset': dataset,
    'tokenizer': tokenizer
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
cumulative_mean_pooling Ran
ran; our contract draft was wrong, not the code
ahans30/goldfish-loss/lit_gpt/retrieval_model.py
code served (permissive licence) · get_code("53c8bb69ab68da37")
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 3, dtype family float, shape (2, 3, 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 hidden_states tensor
# Shape: (batch_size=2, sequence_length=3, hidden_dim=4)
hidden_states = torch.randn(2, 3, 4)

ARGS = {
    "hidden_states": hidden_states,
}
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 3, dtype family float, shape (2, 3, 4) — shape (2, 3, 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
divide Ran
ran, honoured the contract we drafted
ahans30/goldfish-loss/axonn_fabric/fabric.py
code served (permissive licence) · get_code("b38ee8cd162d2d61")
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 int, ndim 0, dtype family int, 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

torch.manual_seed(0)

ARGS = {
    "a": 10,
    "b": 2,
}
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 int, ndim 0, dtype family int, 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
get_batch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ahans30/goldfish-loss/eval/eval_retrieval_anticausal.py
code served (permissive licence) · get_code("41418f480eb09bbc")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 lightning as L
from unittest.mock import Mock

torch.manual_seed(0)

# Create a mock Fabric object
fabric = Mock(spec=L.Fabric)
fabric.device = Mock()
fabric.device.type = 'cpu'
fabric.to_device = lambda x: x  # Identity function for CPU

# Create sample data with query and corpus fields
data = [
    {
        'query': [1, 2, 3],
        'corpus': [4, 5, 6, 7]
    },
    {
        'query': [8, 9],
        'corpus': [10, 11, 12]
    },
    {
        'query': [13, 14, 15, 16],
        'corpus': [17, 18]
    }
]

# Create index tensors
ix = torch.tensor([0, 1])

ARGS = {
    'fabric': fabric,
    'data': data,
    'longest_seq_ix': None,
    'ix': ix
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
lightning, 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
get_tflops Ran
ran, honoured the contract we drafted
ahans30/goldfish-loss/axonn_fabric/megatron_logging.py
code served (permissive licence) · get_code("3b42fb8763cf8824")
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 float, 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 numpy as np

class Config:
    def __init__(self):
        self.n_layer = 2
        self.block_size = 8
        self.padded_vocab_size = 16
        self.n_embd = 4
        self.intermediate_size = 12
        self.gradient_checkpointing = False

config = Config()
batch_size = 2

ARGS = {
    "config": config,
    "batch_size": batch_size
}
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 float, 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
init_normal Ran
ran; our contract draft was wrong, not the code
ahans30/goldfish-loss/lit_gpt/model_axonn.py
code served (permissive licence) · get_code("660a177a11d8a4cb")
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 math
import torch
import torch.nn as nn

torch.manual_seed(0)

# The function init_normal takes n_embd as a parameter
# It returns an initialization function that uses nn.init.trunc_normal_
# The returned init function is meant to be called on tensors

n_embd = 8

ARGS = {
    "n_embd": n_embd
}
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
process_raw_data Ran
ran; our contract draft was wrong, not the code
ahans30/goldfish-loss/eval/data_extraction.py
code served (permissive licence) · get_code("ff1deefe8ebfa94d")
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 str. 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)

# Test case 1: HuggingFaceTB/cosmopedia dataset
raw_data_cosmopedia = {
    'prompt': 'This is a prompt. ',
    'text': 'This is the text content.'
}

# Test case 2: stingning/ultrachat dataset
raw_data_ultrachat = {
    'data': ['First message', 'Second message', 'Third message']
}

# Test case 3: generic dataset with 'text' key
raw_data_generic = {
    'text': 'Some generic text content'
}

# We'll test with the generic case
ARGS = {
    'raw_data': raw_data_generic,
    'dataset': 'some_other_dataset'
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
rank_prefixed_message Ran
ran; our contract draft was wrong, not the code
ahans30/goldfish-loss/axonn_fabric/fabric.py
code served (permissive licence) · get_code("386654d652c835ee")
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 str. 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)

ARGS = {
    "msg": "Test message",
    "rank": 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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
scaled_init_normal Ran
ran; our contract draft was wrong, not the code
ahans30/goldfish-loss/lit_gpt/model_axonn.py
code served (permissive licence) · get_code("4915f91d695a9b8e")
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 torch.nn as nn
import math

torch.manual_seed(0)

ARGS = {
    "n_embd": 8,
    "layer_id": 2,
}
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
str2bool Ran
ran, violated the contract we drafted
ahans30/goldfish-loss/eval/data_extraction.py
code served (permissive licence) · get_code("1eabe88f26a677a1")
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 argparse

def str2bool(v):
    """Human friendly boolean cmdline flag parser."""
    if isinstance(v, bool):
        return v
    if v.lower() in ('yes', 'true', 't', 'y', '1'):
        return True
    elif v.lower() in ('no', 'false', 'f', 'n', '0'):
        return False
    else:
        raise argparse.ArgumentTypeError(f'Boolean value expected. Got: {str(v)}, which cannot be converted to a boolean.')

ARGS = {
    'v': 'true'
}
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 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
build_rope_cache Not yet run
unverified (harvested, no recorded run)
ahans30/goldfish-loss/lit_gpt/model.py
code served (permissive licence) · get_code("c0cabd60c81402a5")
How it ran

no invocation recorded.

generate_responses_batched Not yet run
unverified (harvested, no recorded run)
ahans30/goldfish-loss/eval/alpaca_eval_generate.py
code served (permissive licence) · get_code("55c7e661e08b2d2e")
How it ran

no invocation recorded.

get_conv_template Not yet run
unverified (harvested, no recorded run)
ahans30/goldfish-loss/eval/conversation.py
code served (permissive licence) · get_code("6db38b4d818a2c2e")
How it ran

no invocation recorded.

modified_mlp_forward Not yet run
unverified (harvested, no recorded run)
ahans30/goldfish-loss/axonn_fabric/hf_llama_tensor_parallel.py
code served (permissive licence) · get_code("beb178be96641037")
How it ran

no invocation recorded.

pretty_log Not yet run
unverified (harvested, no recorded run)
ahans30/goldfish-loss/axonn_fabric/megatron_logging.py
code served (permissive licence) · get_code("87ebed951203ba54")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 4 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 4 the same month, 4 later) and 14 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
get_conv_template 69 other papers
28 dated earlier, 8 the same month, 28 later, 4 same year, order not determinable, 1 undated
1911.03705 (2019-11)
CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning
2112.05682 (2021-12)
Self-attention Does Not Need $O(n^2)$ Memory
2204.07705 (2022-04)
SUPER-NATURALINSTRUCTIONS: Generalization via Declarative Instructions on 1600+ NLP Tasks
and 66 more not named here
build_rope_cache 6 other papers
1 dated earlier, 1 the same month, 4 later
2405.20340 (2024-05)
MotionLLM: Understanding Human Behaviors from Human Motions and Videos
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2408.16725 (2024-08)
Mini-Omni: Language Models Can Hear, Talk While Thinking in Streaming
and 3 more not named here
apply_rope 5 other papers
1 the same month, 4 later
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2408.16725 (2024-08)
Mini-Omni: Language Models Can Hear, Talk While Thinking in Streaming
2410.08102 (2024-10)
Efficient Pretraining Data Selection for Language Models via Multi-Actor Collaboration
and 2 more not named here
build_mask_cache 5 other papers
1 the same month, 4 later
2406.06046 (2024-06)
MATES : Model-Aware Data Selection for Efficient Pretraining with Data Influence Models
2408.16725 (2024-08)
Mini-Omni: Language Models Can Hear, Talk While Thinking in Streaming
2410.08102 (2024-10)
Efficient Pretraining Data Selection for Language Models via Multi-Actor Collaboration
and 2 more not named here

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

Large language models can memorize and repeat their training data, causing privacy and copyright risks. To mitigate memorization, we introduce a subtle modification to the next-token training objective that we call the goldfish loss. During training, randomly sampled subsets of tokens are excluded from the loss computation. These dropped tokens are not memorized by the model, which prevents verbatim reproduction of a complete chain of tokens from the training set. We run extensive experiments training billion-scale Llama-2 models, both pre-trained and trained from scratch, and demonstrate significant reductions in extractable memorization with little to no impact on downstream benchmarks.

For agents

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

get_harvested_code_for_paper("2406.10209")
get_code_for_paper("2406.10209")
have("2406.10209")

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

curl https://syntology.ai/api/ran/2406.10209.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.10209.svg)](https://syntology.ai/paper/2406.10209)

Connect an agent — have() is free.