SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1909.01264 · 2019

On the Downstream Performance of Compressed Word Embeddings

arXiv · PDF · Open in the Atlas

Code that ran

2 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; 2 failures of our instrument; 3 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/1909.01264.json · badge: /api/ran/1909.01264.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
HazyResearch/smallfry canonical (classifier label)
classifier basis: classifier_2025
2 ran · 3 no recorded run
FunctionStatusWhere it lives
compress_long_mat Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HazyResearch/smallfry/src/smallfry/quant_embedding.py
code served (permissive licence) · get_code("7fbe6134eb036758")
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) and returned Tensor, ndim 2, dtype family int, shape (2, 1) 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 math

torch.manual_seed(0)

# Create a small int64 tensor with shape (2, 4)
# The function compresses along the last dimension
long_tensor = torch.randint(0, 256, (2, 4), dtype=torch.int64)

# nbit must divide LONG_BITS (64)
# Valid values: 1, 2, 4, 8, 16, 32, 64
nbit = 8

ARGS = {
    "long_tensor": long_tensor,
    "nbit": nbit,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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, 1) — shape (2, 1) 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
decompress_long_mat Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HazyResearch/smallfry/src/smallfry/quant_embedding.py
code served (permissive licence) · get_code("0a3f49fe8de06b20")
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) and returned Tensor, ndim 2, dtype family int, shape (2, 10) 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 decompresses a tensor of int64 values where multiple smaller integers
# are packed into each int64 value. It needs:
# - byte_tensor: an int64 tensor (despite the name, it's int64 not uint8)
# - nbit: number of bits per packed integer
# - dim: optional dimension to truncate to

# LONG_BITS = 64, so valid nbit values are divisors of 64: 1, 2, 4, 8, 16, 32, 64
# Let's use nbit=8, so we can pack 8 integers per int64

# Create a small int64 tensor with shape (2, 4) - batch of 2, 4 packed values each
byte_tensor = torch.tensor([
    [0x0102030405060708, 0x0a0b0c0d0e0f1011, 0x1213141516171819, 0x1a1b1c1d1e1f2021],
    [0x2223242526272829, 0x2a2b2c2d2e2f3031, 0x3233343536373839, 0x3a3b3c3d3e3f4041]
], dtype=torch.int64)

nbit = 8

# dim is optional, let's test with and without it
# When dim is provided, it truncates the output to that many elements in the last dimension
dim = 10  # This will truncate the output's last dimension to 10

ARGS = {
    "byte_tensor": byte_tensor,
    "nbit": nbit,
    "dim": dim
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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, 10) — shape (2, 10) 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
compress_and_compute_frob_squared_error Not yet run
unverified (harvested, no recorded run)
HazyResearch/smallfry/src/smallfry/compress.py
code served (permissive licence) · get_code("56ca64877b14f624")
How it ran

no invocation recorded.

compress_uniform Not yet run
unverified (harvested, no recorded run)
HazyResearch/smallfry/src/smallfry/compress.py
code served (permissive licence) · get_code("1cc1ad513150315f")
How it ran

no invocation recorded.

find_optimal_range Not yet run
unverified (harvested, no recorded run)
HazyResearch/smallfry/src/smallfry/compress.py
code served (permissive licence) · get_code("2f36b0864e68d40c")
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

Compressing word embeddings is important for deploying NLP models in memory-constrained settings. However, understanding what makes compressed embeddings perform well on downstream tasks is challenging---existing measures of compression quality often fail to distinguish between embeddings that perform well and those that do not. We thus propose the eigenspace overlap score as a new measure. We relate the eigenspace overlap score to downstream performance by developing generalization bounds for the compressed embeddings in terms of this score, in the context of linear and logistic regression. We then show that we can lower bound the eigenspace overlap score for a simple uniform quantization compression method, helping to explain the strong empirical performance of this method. Finally, we show that by using the eigenspace overlap score as a selection criterion between embeddings drawn from a representative set we compressed, we can efficiently identify the better performing embedding with up to $2\times$ lower selection error rates than the next best measure of compression quality, and avoid the cost of training a model for each task of interest.

For agents

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

get_harvested_code_for_paper("1909.01264")
get_code_for_paper("1909.01264")
have("1909.01264")

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

curl https://syntology.ai/api/ran/1909.01264.json

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

[![Syntology run record](https://syntology.ai/api/ran/1909.01264.svg)](https://syntology.ai/paper/1909.01264)

Connect an agent — have() is free.