SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.12585 · 2024

Breaking the Ceiling of the LLM Community by Treating Token Generation as a Classification for Ensembling

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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 0 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/2406.12585.json · badge: /api/ran/2406.12585.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
yaoching0/gac canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
check_priorities Ran
ran, honoured the contract we drafted
yaoching0/gac/utils/gac_gen_call.py
code served (permissive licence) · get_code("250e3d0aab21f5dc")
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
import numpy as np

# Construct valid arguments for check_priorities function
# The function expects a list of dictionaries, each with a 'priority' key

dict_list = [
    {'priority': 'supportive', 'name': 'item1'},
    {'priority': 'primary', 'name': 'item2'},
    {'priority': 'supportive', 'name': 'item3'},
]

ARGS = {
    'dict_list': dict_list
}
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 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
setup_custom_logger Ran
ran; our contract draft was wrong, not the code
yaoching0/gac/utils/logger.py
pointer only (licence: MIT); the stored copy get_code serves, from yoon6503/SAFE, is pointer only (licence: NONE) · get_code("e008364719672578")
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 yoon6503/SAFE, and GitHub detected no licence in that repository when it was checked. 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.

synchronize_unfinished_sequences Ran
ran; our contract draft was wrong, not the code
yaoching0/gac/utils/gac_gen_utils.py
code served (permissive licence) · get_code("3b4f685120e04a9c")
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, 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 list of unfinished_sequences tensors
# Each tensor represents which sequences in a batch are still unfinished
# Shape: (batch_size, sequence_length) where values are 0 or 1
unfinished_sequences_1 = torch.tensor([
    [1, 1, 1, 1],
    [1, 0, 1, 1]
], dtype=torch.long)

unfinished_sequences_2 = torch.tensor([
    [1, 1, 0, 1],
    [1, 1, 1, 1]
], dtype=torch.long)

unfinished_sequences_3 = torch.tensor([
    [1, 1, 1, 1],
    [1, 1, 1, 0]
], dtype=torch.long)

unfinished_sequences_list = [
    unfinished_sequences_1,
    unfinished_sequences_2,
    unfinished_sequences_3
]

ARGS = {
    "unfinished_sequences_list": unfinished_sequences_list
}
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, 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
update_input_ids_and_model_kwargs Ran
ran; our contract draft was wrong, not the code
yaoching0/gac/utils/gac_gen_utils.py
code served (permissive licence) · get_code("a1d6799b4d0d673e")
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. 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

torch.manual_seed(0)

class MockConfig:
    def __init__(self):
        self.is_encoder_decoder = False

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.config = MockConfig()
    
    def _update_model_kwargs_for_generation(self, outputs, model_kwargs, is_encoder_decoder):
        return model_kwargs

model = MockModel()

batch_size = 2
seq_len = 3
vocab_size = 100
pad_token_id = 0
eos_token_id = 2

input_ids = torch.randint(1, vocab_size, (batch_size, seq_len), dtype=torch.long)
input_ids[0, 0] = 5  # ensure non-pad token

outputs = {
    'logits': torch.randn(batch_size, 1, vocab_size)
}

next_tokens_list = [
    [torch.tensor(15, dtype=torch.long)],
    [torch.tensor(25, dtype=torch.long)]
]

attention_mask = torch.ones(batch_size, seq_len, dtype=torch.long)

model_kwargs = {
    'attention_mask': attention_mask,
    'past_key_values': None
}

unfinished_sequences = [True, True]

eos_token_id_tensor = torch.tensor(eos_token_id, dtype=torch.long)

state = {
    'outputs': outputs,
    'input_ids': input_ids,
    'next_tokens_list': next_tokens_list,
    'model_kwargs': model_kwargs,
    'unfinished_sequences': unfinished_sequences,
    'pad_token_id': pad_token_id,
    'eos_token_id_tensor': eos_token_id_tensor
}

ARGS = {
    'model': model,
    'state': state
}
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 — 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
validate_and_update_quantization Ran
ran; our contract draft was wrong, not the code
yaoching0/gac/utils/gac_gen_call.py
code served (permissive licence) · get_code("4fda25a3a31731c5")
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 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 numpy as np

# Setup for validate_and_update_quantization function
# The function takes a list of dictionaries representing model configurations

ARGS = {
    'model_config': [
        {'name': 'model1', 'layers': '12'},
        {'name': 'model2', 'quantization': '8bit'},
        {'name': 'model3', 'quantization': 'none'},
        {'name': 'model4'},
    ]
}
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 float — 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
get_remote_model_generator_class Not yet run
unverified (harvested, no recorded run)
yaoching0/gac/utils/ray_actor.py
code served (permissive licence) · get_code("5640e58ac0f3f8c3")
How it ran

no invocation recorded.

setup_model_actors_and_data Not yet run
unverified (harvested, no recorded run)
yaoching0/gac/utils/gac_gen_call.py
code served (permissive licence) · get_code("03e59517a81a3729")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 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 (0 with at least one paper dated earlier, 0 the same month, 1 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
setup_custom_logger 1 other paper
1 later
2510.15346 (2025-10)
When to Ensemble: Identifying Token-Level Points for Stable and Fast LLM Ensembling WHEN TO ENSEMBLE: IDENTIFYING TOKEN-LEVEL POINTS FOR STABLE AND FAST LLM ENSEMBLING

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

Ensembling multiple models has always been an effective approach to push the limits of existing performance and is widely used in classification tasks by simply averaging the classification probability vectors from multiple classifiers to achieve better accuracy. However, in the thriving open-source Large Language Model (LLM) community, ensembling methods are rare and typically limited to ensembling the full-text outputs of LLMs, such as selecting the best output using a ranker, which leads to underutilization of token-level probability information. In this paper, we treat the Generation of each token by LLMs as a Classification (GaC) for ensembling. This approach fully exploits the probability information at each generation step and better prevents LLMs from producing early incorrect tokens that lead to snowballing errors. In experiments, we ensemble state-of-the-art LLMs on several benchmarks, including exams, mathematics and reasoning, and observe that our method breaks the existing community performance ceiling. Furthermore, we observed that most of the tokens in the answer are simple and do not affect the correctness of the final answer. Therefore, we also experimented with ensembling only key tokens, and the results showed better performance with lower latency across benchmarks.

For agents

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

get_harvested_code_for_paper("2406.12585")
get_code_for_paper("2406.12585")
have("2406.12585")

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

curl https://syntology.ai/api/ran/2406.12585.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.12585.svg)](https://syntology.ai/paper/2406.12585)

Connect an agent — have() is free.