SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.07548 · 2024

Image and Video Tokenization with Binary Spherical Quantization

arXiv · PDF · Open in the Atlas

Code that ran

12 of 14 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; 2 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 9 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/2406.07548.json · badge: /api/ran/2406.07548.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
zhaoyue-zephyrus/bsq-vit canonical (classifier label)
classifier basis: census
12 ran · 2 no recorded run
FunctionStatusWhere it lives
codebook_entropy Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/models/quantizer/bsq.py
code served (permissive licence) · get_code("82547443813db80a")
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 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 torch.nn as nn

torch.manual_seed(0)

# Analyze the function to understand what each parameter needs to be:
# - zq: tensor that gets normalized and multiplied by basis, then indexed
# - basis: tensor that zq is multiplied with, should have shape compatible with zq
# - K: integer used to determine size of scatter_reduce output (2**K)
# - eps: float used in probability calculation

# From the forward pass:
# zb = (zq + 1) / 2  -> zq is a tensor
# zi = (zb * basis).sum(-1).to(torch.int64)  -> basis has same or compatible shape with zq
# cnt = torch.scatter_reduce(..., 2**K, ...)  -> K determines output size
# The scatter_reduce needs zi.flatten() to be valid indices into range [0, 2**K)

# zq shape: (batch, dim) or (batch, seq, dim)
# basis shape: needs to match zq's last dimension to allow broadcasting
# After (zq + 1) / 2, values in [0, 1]
# After * basis and sum(-1), we get integers in range [0, 2**K)

# Let's use batch=2, dim=4, K=2 so 2**K=4
batch_size = 2
dim = 4
K = 2
num_codes = 2 ** K  # 4

# Create zq: shape (batch, dim) with values in roughly [-1, 1] so after normalization they're in [0, 1]
zq = torch.randn(batch_size, dim) * 0.5

# Create basis: shape (dim,) with values that when multiplied and summed give valid indices
# After (zq + 1) / 2, zq is in [0, 1] roughly
# We need (zb * basis).sum(-1) to be in range [0, 2**K)
# Simple approach: basis values that scale appropriately
basis = torch.arange(dim, dtype=torch.float32) + 1  # [1, 2, 3, 4]
# Scale basis so the sum stays in valid range
basis = basis / basis.sum() * (num_codes - 0.5)

# eps for numerical stability
eps = 0.0001

ARGS = {
    'zq': zq,
    'basis': basis,
    'K': K,
    'eps': eps,
}
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 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
compute_fvd_score Ran
ran, honoured the contract we drafted
zhaoyue-zephyrus/bsq-vit/transcoder/evaluations/fvd/fvd_score.py
code served (permissive licence) · get_code("476d0c33c781d232")
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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create fake features: shape (batch_size, feature_dim)
# Using distinct dimensions: 2 samples, 4 features
feats_fake = np.random.randn(2, 4).astype(np.float32)

# Create real features: same shape as fake features
feats_real = np.random.randn(2, 4).astype(np.float32)

ARGS = {
    "feats_fake": feats_fake,
    "feats_real": feats_real,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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
compute_stats Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/evaluations/fvd/fvd_score.py
code served (permissive licence) · get_code("4d85e9c25e578ed4")
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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create a small feature matrix: shape (batch_size=2, num_features=4)
# This will be used to compute mean and covariance
feats = np.random.randn(2, 4).astype(np.float32)

ARGS = {
    "feats": feats,
}
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 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_attention_mask Ran
ran, violated the contract we drafted
zhaoyue-zephyrus/bsq-vit/transcoder/models/attention_mask.py
code served (permissive licence) · get_code("d8ca8331fdf571f8")
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, 32) 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)

# Test case 1: block-causal mask with default block_size
ARGS = {
    'sequence_length': 32,
    'device': 'cpu',
    'mask_type': 'block-causal',
    'block_size': 16
}
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, 32) — shape (32, 32) 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_ckpt_path Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/losses/lpips.py
code served (permissive licence) · get_code("726780cba749fec2")
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 os
import tempfile
import hashlib

# Create a temporary directory for testing
temp_dir = tempfile.mkdtemp()

# Create a dummy checkpoint file with known content
dummy_ckpt_dir = os.path.join(temp_dir, 'checkpoints')
os.makedirs(dummy_ckpt_dir, exist_ok=True)

dummy_ckpt_path = os.path.join(dummy_ckpt_dir, 'vgg.pth')

# Create a file with content that matches the expected MD5
# We'll create a simple binary file
dummy_content = b'vgg_lpips_model_data'
with open(dummy_ckpt_path, 'wb') as f:
    f.write(dummy_content)

# Calculate the actual MD5 of our dummy file
with open(dummy_ckpt_path, 'rb') as f:
    actual_md5 = hashlib.md5(f.read()).hexdigest()

ARGS = {
    'name': 'vgg_lpips',
    'root': dummy_ckpt_dir,
    'check': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
requests, 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
get_inception_score Ran
ran, honoured the contract we drafted
zhaoyue-zephyrus/bsq-vit/transcoder/evaluations/inception_score.py
code served (permissive licence) · get_code("898e597a6eb684c2")
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 float32, 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 numpy as np

# The function takes logits as input and:
# 1. Applies softmax along axis=1
# 2. Computes entropy using scipy.stats.entropy
# 3. Returns exp of mean entropy
# 
# logits should be a 2D numpy array where:
# - axis 0 is batch dimension
# - axis 1 is class/logit dimension

# Create a small batch of logits
np.random.seed(0)
logits = np.random.randn(2, 4).astype(np.float32)

ARGS = {
    "logits": logits
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float32, 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_psnr Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zhaoyue-zephyrus/bsq-vit/transcoder/evaluations/psnr.py
code served (permissive licence) · get_code("18eb88ffa0fbed03")
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 1, dtype family float, shape (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)

# Create small tensors for image data
# Shape: (batch=2, channels=3, height=4, width=5)
x_input = torch.rand(2, 3, 4, 5)
x_recon = torch.rand(2, 3, 4, 5)

# Create small video tensors
# Shape: (batch=2, time=3, channels=3, height=4, width=5)
x_input_video = torch.rand(2, 3, 3, 4, 5)
x_recon_video = torch.rand(2, 3, 3, 4, 5)

ARGS = {
    "x_input": x_input,
    "x_recon": x_recon,
    "zero_mean": False,
    "is_video": False,
}
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 1, dtype family float, shape (2,) — shape (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
hinge_d_loss Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/losses/vqperceptual.py
pointer only (licence: MIT); the stored copy get_code serves, from joanrod/ocr-vqgan, is pointer only (licence: NONE) · get_code("22a83df891d750c6")
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 joanrod/ocr-vqgan, 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.

hinge_g_loss Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/losses/vqperceptual.py
code served (permissive licence) · get_code("39168ca3aac0a4d7")
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 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

torch.manual_seed(0)

# The function hinge_g_loss takes logits_fake (a tensor) and computes -mean(logits_fake)
# We need to create a small tensor with fake logits

logits_fake = torch.randn(2, 4)  # batch size 2, 4 logits per sample

ARGS = {
    "logits_fake": logits_fake,
}
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 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
md5_hash Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/losses/lpips.py
code served (permissive licence) · get_code("c8d9d38b144e401f")
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 tempfile
import os

# Create a temporary file with some content
temp_file = tempfile.NamedTemporaryFile(mode='wb', delete=False, suffix='.txt')
temp_file.write(b"This is test content for md5 hashing")
temp_file.close()
temp_path = temp_file.name

ARGS = {
    "path": temp_path
}
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 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
normalize_tensor Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/losses/lpips.py
pointer only (licence: MIT); the stored copy get_code serves, from zh460045050/V2L-Tokenizer, is pointer only (licence: NONE) · get_code("7257fdf155f5ef45")
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 zh460045050/V2L-Tokenizer, 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.

vanilla_d_loss Ran
ran; our contract draft was wrong, not the code
zhaoyue-zephyrus/bsq-vit/transcoder/losses/vqperceptual.py
pointer only (licence: MIT); the stored copy get_code serves, from joanrod/ocr-vqgan, is pointer only (licence: NONE) · get_code("263c54290c553ddd")
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 joanrod/ocr-vqgan, 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.

calculate_frechet_distance Not yet run
unverified (harvested, no recorded run)
zhaoyue-zephyrus/bsq-vit/transcoder/evaluations/fid/fid_score.py
code served (permissive licence) · get_code("0def50a351111624")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

get_lpips Not yet run
unverified (harvested, no recorded run)
zhaoyue-zephyrus/bsq-vit/transcoder/evaluations/lpips.py
code served (permissive licence) · get_code("f50362598da489ee")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 6 the same month, 6 later) and 8 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
calculate_frechet_distance 74 other papers
61 dated earlier, 3 the same month, 9 later, 1 same year, order not determinable
1610.09585 (2016-10)
Conditional Image Synthesis With Auxiliary Classifier GANs
1706.08500 (2017-06)
GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
1807.11458 (2018-07)
To learn image super-resolution, use a GAN to learn how to do image degradation first
and 71 more not named here
md5_hash 30 other papers
17 dated earlier, 2 the same month, 10 later, 1 same year, order not determinable
2008.01777 (2020-08)
Making Sense of CNNs: Interpreting Deep Representations & Their Invariances with INNs
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
and 27 more not named here
hinge_d_loss 15 other papers
7 dated earlier, 1 the same month, 7 later
2012.09841 (2020-12)
Taming Transformers for High-Resolution Image Synthesis
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
2211.13221 (2022-11)
Latent Video Diffusion Models for High-Fidelity Long Video Generation
and 12 more not named here
normalize_tensor 13 other papers
10 dated earlier, 1 the same month, 2 later
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
2203.10897 (2022-03)
Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression
2203.16521 (2022-03)
CoordGAN: Self-Supervised Dense Correspondences Emerge from GANs
and 10 more not named here
vanilla_d_loss 11 other papers
7 dated earlier, 1 the same month, 3 later
2012.09841 (2020-12)
Taming Transformers for High-Resolution Image Synthesis
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
2211.13221 (2022-11)
Latent Video Diffusion Models for High-Fidelity Long Video Generation
and 8 more not named here
get_ckpt_path 10 other papers
6 dated earlier, 1 the same month, 2 later, 1 same year, order not determinable
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
Li_MAGE_MAsked_Generative_Encoder_To_Unify_Representation_Learning_and_Image_CVPR_2023_paper (2023 (year only))
MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis
2308.07228 (2023-08)
RestoreFormer++: Towards Real-World Blind Face Restoration from Undegraded Key-Value Pairs
and 7 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

We propose a new transformer-based image and video tokenizer with Binary Spherical Quantization (BSQ). BSQ projects the high-dimensional visual embedding to a lower-dimensional hypersphere and then applies binary quantization. BSQ is (1) parameter-efficient without an explicit codebook, (2) scalable to arbitrary token dimensions, and (3) compact: compressing visual data by up to 100$\times$ with minimal distortion. Our tokenizer uses a transformer encoder and decoder with simple block-wise causal masking to support variable-length videos as input. The resulting BSQ-ViT achieves state-of-the-art visual reconstruction quality on image and video reconstruction benchmarks with 2.4$\times$ throughput compared to the best prior methods. Furthermore, by learning an autoregressive prior for adaptive arithmetic coding, BSQ-ViT achieves comparable results on video compression with state-of-the-art video compression standards. BSQ-ViT also enables masked language models to achieve competitive image synthesis quality to GAN- and diffusion-based methods.

For agents

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

get_harvested_code_for_paper("2406.07548")
get_code_for_paper("2406.07548")
have("2406.07548")

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

curl https://syntology.ai/api/ran/2406.07548.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.07548.svg)](https://syntology.ai/paper/2406.07548)

Connect an agent — have() is free.