SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.06694 · ICML · 2026

NANOQUANT: Efficient Sub-1-Bit Quantization of Large Language Models

Dongkyu Kim, Minseop Choi, Hyochan Chong, Changdong Kim

arXiv · PDF · Open in the Atlas

Code that ran

4 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 0 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/2602.06694.json · badge: /api/ran/2602.06694.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
SamsungLabs/NanoQuant community
Papers with Code archive
4 ran · 1 no recorded run
FunctionStatusWhere it lives
_admm_solve_step Ran
ran; our contract draft was wrong, not the code
SamsungLabs/NanoQuant/src/nanoquant/core/admm_nq.py
code served (permissive licence) · get_code("03a29319d960aeb3")
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 (4, 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

torch.manual_seed(0)

# Construct tensors with distinct sizes for different dimensions
# X: (n_samples, n_features) - shape (2, 4)
# Y: (n_samples, n_targets) - shape (2, 3)
# Z: (n_features, n_targets) - shape (4, 3)
# U: (n_features, n_targets) - shape (4, 3)

X = torch.randn(2, 4, dtype=torch.float32)
Y = torch.randn(2, 3, dtype=torch.float32)
Z = torch.randn(4, 3, dtype=torch.float32)
U = torch.randn(4, 3, dtype=torch.float32)

rho = 1.0
reg = 0.01

ARGS = {
    "X": X,
    "Y": Y,
    "Z": Z,
    "U": U,
    "rho": rho,
    "reg": reg,
    "eps": 1e-12
}
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 (4, 3) — shape (4, 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
power_iteration Ran
ran; our contract draft was wrong, not the code
SamsungLabs/NanoQuant/src/nanoquant/core/admm_nq.py
code served (permissive licence) · get_code("fbf2e7de4ef108c7")
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

torch.manual_seed(0)

# Construct a small matrix A for power iteration
# Shape (4, 3) - 4 rows, 3 columns
# This will work with power_iteration which expects A.shape[1] to give column dimension
A = torch.randn(4, 3, dtype=torch.float32)

ARGS = {
    "A": A,
    "num_iters": 5
}
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
rank1_approx Ran
ran; our contract draft was wrong, not the code
SamsungLabs/NanoQuant/src/nanoquant/core/admm_nq.py
code served (permissive licence) · get_code("3764bd907e747fab")
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 (4, 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

torch.manual_seed(0)

# Create a small test matrix W
# The function rank1_approx calls svid, which calls power_iteration
# power_iteration expects a 2D tensor and performs matrix operations
# We need a small matrix with distinct dimensions for testing
W = torch.randn(4, 3)

ARGS = {
    "W": W,
    "inner_iters": 5,
    "eps": 1e-12
}
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 (4, 3) — shape (4, 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
svid Ran
ran; our contract draft was wrong, not the code
SamsungLabs/NanoQuant/src/nanoquant/core/admm_nq.py
code served (permissive licence) · get_code("7c5ad3e8a613b779")
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

torch.manual_seed(0)

# Create a small test matrix for W
# The function calls W.sign(), W.abs(), and passes W to power_iteration
# power_iteration expects A to have shape (m, n) and uses A.shape[1], A.mT, torch.mv(A, v)
# So W should be a 2D tensor
W = torch.randn(4, 3)

ARGS = {
    "W": W,
    "inner_iters": 5,
    "eps": 1e-12,
}
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
factorize_admm_nanoquant Not yet run
unverified (harvested, no recorded run)
SamsungLabs/NanoQuant/src/nanoquant/core/admm_nq.py
code served (permissive licence) · get_code("5cbf859dd3d691a0")
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

Weight-only quantization has become a standard approach for efficiently serving large language models (LLMs). However, existing methods fail to efficiently compress models to binary (1-bit) levels, as they either require large amounts of data and compute or incur additional storage. In this work, we propose NANOQUANT, a post-training quantization (PTQ) method to compress LLMs to both binary and sub-1-bit levels. NANOQUANT formulates quantization as a low-rank binary factorization problem, and compresses full-precision weights to low-rank binary matrices and scales. Specifically, it utilizes an efficient alternating direction method of multipliers (ADMM) solver to precisely initialize latent binary matrices and scales, and then tunes the initialized parameters through a block and model reconstruction process. Consequently, NANOQUANT establishes a new Pareto frontier in low-memory post-training quantization, and enables sub-1-bit compression. NANOQUANT makes large-scale deployment feasible on consumer hardware. For example, it compresses Llama-2-70B by 24× in just 13 hours on a single H100, enabling a 70B model to operate on a consumer 8 GB GPU. Code is available at github.com/SamsungLabs/NanoQuant.

For agents

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

get_harvested_code_for_paper("2602.06694")
get_code_for_paper("2602.06694")
have("2602.06694")

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

curl https://syntology.ai/api/ran/2602.06694.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.06694.svg)](https://syntology.ai/paper/2602.06694)

Connect an agent — have() is free.