SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1908.10396 · 2019

Accelerating Large-Scale Inference with Anisotropic Vector Quantization

arXiv · PDF · Open in the Atlas

Code that ran

7 of 8 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 7 have no contract verdict; 0 failures of our instrument; 1 no recorded run), measured 2026-11-21; 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/1908.10396.json · badge: /api/ran/1908.10396.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
datastax/jvector community (classifier label)
classifier basis not recorded on the link
7 ran · 1 no recorded run
FunctionStatusWhere it lives
build_footer_text Ran
ran on a synthesized input
datastax/jvector/jvector-examples/src/main/resources/visualize_benchmarks.py
code served (permissive licence) · get_code("4835100e17256550")
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 textwrap
from collections import defaultdict
from typing import Dict, List, Any, Tuple, Optional

ARGS = {
    'instance_type': 'r5.2xlarge',
    'jdk_version': '11.0.15',
    'jvm_flags': '-Xmx16g -XX:+UseG1GC',
    'dataset_config_map': {
        'wikipedia': 'prod_config.yaml',
        'common_crawl': 'prod_config.yaml',
        'bookcorpus': 'dev_config.yaml',
        'arxiv': 'test_config.yaml'
    },
    'max_line_width': 120
}
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
extract_release_from_filename Ran
ran on a synthesized input
datastax/jvector/jvector-examples/src/main/resources/compare_benchmark_iterations.py
code served (permissive licence) · get_code("bc967aa9df2e48e3")
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 re

def extract_release_from_filename(filename: str) -> str:
    """
    Extract release number from filename
    """
    version_match = re.search('v?(\\d+\\.\\d+\\.\\d+)', filename)
    if version_match:
        return version_match.group(1)
    return os.path.splitext(filename)[0]

# Test with a filename containing a version number
ARGS = {
    "filename": "release_v1.2.3_final.tar.gz"
}
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
filter_pareto_optimal Ran
ran on a synthesized input
datastax/jvector/plot_output.py
code served (permissive licence) · get_code("6a1c22f297b5f8c6")
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.

class Point:
    def __init__(self, recall, throughput):
        self.recall = recall
        self.throughput = throughput

# Create a list of points with different recall and throughput values
data = [
    Point(recall=0.8, throughput=100.0),
    Point(recall=0.7, throughput=150.0),
    Point(recall=0.9, throughput=80.0),
    Point(recall=0.85, throughput=90.0),
    Point(recall=0.75, throughput=120.0),
]

ARGS = {
    "data": data
}
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 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_dataset_config_map Ran
ran on a synthesized input
datastax/jvector/jvector-examples/src/main/resources/visualize_benchmarks.py
code served (permissive licence) · get_code("91e821c26dd9da52")
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 dict. 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 shutil

# Create a temporary directory structure that mimics the expected layout
temp_dir = tempfile.mkdtemp()
repo_root = temp_dir
jvector_examples_dir = os.path.join(repo_root, 'jvector-examples')
yaml_configs_dir = os.path.join(jvector_examples_dir, 'yaml-configs')
index_parameters_dir = os.path.join(yaml_configs_dir, 'index-parameters')

os.makedirs(index_parameters_dir, exist_ok=True)

# Create a .github directory so _find_repo_root can find the root
github_dir = os.path.join(repo_root, '.github')
os.makedirs(github_dir, exist_ok=True)

# Create some sample YAML files
sample_datasets = {'dataset1', 'dataset2', 'dataset3'}
for dataset in sample_datasets:
    yaml_file = os.path.join(index_parameters_dir, f'{dataset}.yml')
    with open(yaml_file, 'w') as f:
        f.write(f'# Config for {dataset}\n')

# Create a default.yml as well
default_yaml_file = os.path.join(index_parameters_dir, 'default.yml')
with open(default_yaml_file, 'w') as f:
    f.write('# Default config\n')

# Set up the ARGS dictionary
ARGS = {
    'datasets': sample_datasets,
    'config_dir': index_parameters_dir
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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_jvm_flags Ran
ran on a synthesized input
datastax/jvector/jvector-examples/src/main/resources/visualize_benchmarks.py
code served (permissive licence) · get_code("dd7e8c51e7331624")
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 re

# Create a temporary workflow file that mimics the structure the function expects
temp_dir = tempfile.mkdtemp()
workflows_dir = os.path.join(temp_dir, '.github', 'workflows')
os.makedirs(workflows_dir, exist_ok=True)

workflow_file = os.path.join(workflows_dir, 'run-bench.yml')

# Create a minimal but realistic run-bench.yml file
workflow_content = """name: Benchmark

on:
  pull_request:
    branches: [main]
  push:
    branches: [main]

jobs:
  benchmark:
    runs-on: ubuntu-latest
    strategy:
      matrix:
        jdk: [11, 17, 21]
    steps:
      - uses: actions/checkout@v3
      
      - name: Run benchmark
        run: |
          if [[ "${{ github.event_name }}" == "pull_request" ]]; then
            java -XX:+UseG1GC -XX:MaxGCPauseMillis=200 \\
              -Xmx512m -Xms256m \\
              ${{ matrix.jdk >= 17 && '-XX:+EnablePreview' || '' }} \\
              -Dbenchmark.threads=4 \\
              -XX:ReservedCodeCacheSize=${HALF_MEM_GB}m \\
              org.openjdk.jmh.Main
          else
            java -XX:+UseG1GC -XX:MaxGCPauseMillis=200 \\
              -Xmx1024m -Xms512m \\
              ${{ matrix.jdk >= 17 && '-XX:+EnablePreview -XX:+UnlockExperimentalVMOptions' || '' }} \\
              ${{ matrix.jdk >= 21 && '-XX:+TieredCompilation' || '' }} \\
              -Dbenchmark.threads=8 \\
              -Dbenchmark.mode=thrpt \\
              -XX:ReservedCodeCacheSize=${HALF_MEM_GB}m \\
              org.openjdk.jmh.Main
          fi
"""

with open(workflow_file, 'w') as f:
    f.write(workflow_content)

ARGS = {
    'jdk_major': 17,
    'workflow_path': workflow_file,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
psutil — 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
is_pareto_optimal Ran
ran on a synthesized input
datastax/jvector/plot_output.py
code served (permissive licence) · get_code("e53b1ba6db1713c4")
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.

class Point:
    def __init__(self, recall, throughput):
        self.recall = recall
        self.throughput = throughput

candidate = Point(recall=0.75, throughput=100.0)

others = [
    Point(recall=0.70, throughput=95.0),
    Point(recall=0.80, throughput=105.0),
    Point(recall=0.72, throughput=98.0),
]

ARGS = {
    "candidate": candidate,
    "others": others,
}
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
load_benchmark_data Ran
ran on a synthesized input
datastax/jvector/jvector-examples/src/main/resources/compare_benchmark_iterations.py
code served (permissive licence) · get_code("09cf386323967ddc")
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 BenchmarkData. 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 csv

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

# Create first CSV file (v1.0.0)
csv_file_1 = os.path.join(temp_dir, 'benchmarks_v1.0.0.csv')
with open(csv_file_1, 'w', newline='') as f:
    writer = csv.DictWriter(f, fieldnames=['Dataset', 'Accuracy', 'Loss'])
    writer.writeheader()
    writer.writerow({'Dataset': 'dataset_A', 'Accuracy': '0.95', 'Loss': '0.05'})
    writer.writerow({'Dataset': 'dataset_B', 'Accuracy': '0.92', 'Loss': '0.08'})

# Create second CSV file (v1.0.1)
csv_file_2 = os.path.join(temp_dir, 'benchmarks_v1.0.1.csv')
with open(csv_file_2, 'w', newline='') as f:
    writer = csv.DictWriter(f, fieldnames=['Dataset', 'Accuracy', 'Loss'])
    writer.writeheader()
    writer.writerow({'Dataset': 'dataset_A', 'Accuracy': '0.96', 'Loss': '0.04'})
    writer.writerow({'Dataset': 'dataset_B', 'Accuracy': '0.93', 'Loss': '0.07'})

# Create third CSV file (main)
csv_file_3 = os.path.join(temp_dir, 'benchmarks_main.csv')
with open(csv_file_3, 'w', newline='') as f:
    writer = csv.DictWriter(f, fieldnames=['Dataset', 'Accuracy', 'Loss'])
    writer.writeheader()
    writer.writerow({'Dataset': 'dataset_A', 'Accuracy': '0.97', 'Loss': '0.03'})
    writer.writerow({'Dataset': 'dataset_B', 'Accuracy': '0.94', 'Loss': '0.06'})

ARGS = {
    'directory': temp_dir
}
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 BenchmarkData — a BenchmarkData output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a BenchmarkData 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
parse_data Not yet run
unverified (harvested, no recorded run)
datastax/jvector/plot_output.py
code served (permissive licence) · get_code("c0ae8333dcd0e9f8")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 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 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.

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Quantization based techniques are the current state-of-the-art for scaling maximum inner product search to massive databases. Traditional approaches to quantization aim to minimize the reconstruction error of the database points. Based on the observation that for a given query, the database points that have the largest inner products are more relevant, we develop a family of anisotropic quantization loss functions. Under natural statistical assumptions, we show that quantization with these loss functions leads to a new variant of vector quantization that more greatly penalizes the parallel component of a datapoint's residual relative to its orthogonal component. The proposed approach achieves state-of-the-art results on the public benchmarks available at \url{ann-benchmarks.com}.

For agents

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

get_harvested_code_for_paper("1908.10396")
get_code_for_paper("1908.10396")
have("1908.10396")

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

curl https://syntology.ai/api/ran/1908.10396.json

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

[![Syntology run record](https://syntology.ai/api/ran/1908.10396.svg)](https://syntology.ai/paper/1908.10396)

Connect an agent — have() is free.