SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2411.01142 · 2024

NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference

arXiv · PDF · Open in the Atlas

Code that ran

3 of 5 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; 0 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/2411.01142.json · badge: /api/ran/2411.01142.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
MachineLearningSystem/25MLSYS-NEO community (classifier label)
classifier basis not recorded on the link
3 ran · 2 no recorded run
FunctionStatusWhere it lives
get_lat_avg Ran
ran on a synthesized input
MachineLearningSystem/25MLSYS-NEO/evaluation/illustrator.py
code served (permissive licence) · get_code("f1eae6f23294c7bb")
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 json
import tempfile
import os

# Create a temporary JSON file with sample data
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.json')

# Create sample data with the expected structure
sample_data = [
    {'start': 0.0, 'end': 0.1, 'output_len': 10},
    {'start': 0.1, 'end': 0.2, 'output_len': 10},
    {'start': 0.2, 'end': 0.3, 'output_len': 10},
    {'start': 0.3, 'end': 0.4, 'output_len': 10},
    {'start': 0.4, 'end': 0.5, 'output_len': 10},
    {'start': 0.5, 'end': 0.6, 'output_len': 10},
    {'start': 0.6, 'end': 0.7, 'output_len': 10},
    {'start': 0.7, 'end': 0.8, 'output_len': 10},
]

json.dump(sample_data, temp_file)
temp_file.close()

ARGS = {
    'file': temp_file.name
}
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 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
get_tp Ran
ran on a synthesized input
MachineLearningSystem/25MLSYS-NEO/evaluation/illustrator.py
code served (permissive licence) · get_code("4affe90a85f8702c")
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 json
import numpy as np
import tempfile
import os

# Create temporary JSON files with the required structure
temp_dir = tempfile.mkdtemp()

# Create first test file
file1_path = os.path.join(temp_dir, "file1.json")
data1 = [
    {"end": 1.0},
    {"end": 1.5},
    {"end": 2.2},
    {"end": 3.1},
    {"end": 4.0},
    {"end": 5.5},
    {"end": 6.7},
    {"end": 7.9},
    {"end": 9.0},
    {"end": 10.5},
]
with open(file1_path, 'w') as f:
    json.dump(data1, f)

# Create second test file
file2_path = os.path.join(temp_dir, "file2.json")
data2 = [
    {"end": 0.5},
    {"end": 1.2},
    {"end": 2.0},
    {"end": 2.9},
    {"end": 3.8},
    {"end": 4.5},
    {"end": 5.3},
    {"end": 6.1},
    {"end": 7.0},
    {"end": 8.2},
]
with open(file2_path, 'w') as f:
    json.dump(data2, f)

ARGS = {
    "filenames": [file1_path, file2_path],
    "interv": (0.1, 0.9)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
prepare_mock_test Ran
ran on a synthesized input
MachineLearningSystem/25MLSYS-NEO/evaluation/benchmark.py
code served (permissive licence) · get_code("678f874464900355")
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 os
import random

# Set seed for reproducibility
random.seed(0)

# Create a minimal config dict
config = {
    'model_path': '/path/to/model'
}

# Define arguments for prepare_mock_test
ARGS = {
    'nreqs': 5,
    'input_len': 128,
    'output_len': 64,
    'server_name': 'test_server',
    'config': config
}
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 — 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
draw_one_rl_diagram Not yet run
unverified (harvested, no recorded run)
MachineLearningSystem/25MLSYS-NEO/evaluation/illustrator.py
code served (permissive licence) · get_code("a4f31d440ecccee6")
How it ran

no invocation recorded.

prepare_real_test Not yet run
unverified (harvested, no recorded run)
MachineLearningSystem/25MLSYS-NEO/evaluation/benchmark.py
code served (permissive licence) · get_code("75de68bb7abb836e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 5 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.

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

Online LLM inference powers many exciting applications such as intelligent chatbots and autonomous agents. Modern LLM inference engines widely rely on request batching to improve inference throughput, aiming to make it cost-efficient when running on expensive GPU accelerators. However, the limited GPU memory has largely limited the batch size achieved in practice, leaving significant GPU compute resources wasted. We present NEO, an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU, effectively increasing the GPU batch size and thus inference throughput. To this end, NEO proposes asymmetric GPU-CPU pipelining and load-aware scheduling to balance GPU and CPU loads and fully utilize their compute and memory resources. We evaluate NEO on a wide range of workloads (i.e., code generation, text summarization), GPUs (i.e., T4, A10G, H100), and LLM models (i.e., 7B, 8B, 70B). NEO achieves up to 7.5$\times$, 26%, and 14% higher throughput compared to GPU-only approach on T4, A10G, and H100 GPUs, respectively, while maintaining the same latency; with more powerful CPUs, NEO achieves up to 79.3% throughput gain on A10G GPU.

For agents

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

get_harvested_code_for_paper("2411.01142")
get_code_for_paper("2411.01142")
have("2411.01142")

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

curl https://syntology.ai/api/ran/2411.01142.json

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

[![Syntology run record](https://syntology.ai/api/ran/2411.01142.svg)](https://syntology.ai/paper/2411.01142)

Connect an agent — have() is free.