SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2205.09707 · 2022

PLAID: An Efficient Engine for Late Interaction Retrieval

arXiv · PDF · Open in the Atlas

Code that ran

1 of 12 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 1 has no contract verdict; 0 failures of our instrument; 11 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/2205.09707.json · badge: /api/ran/2205.09707.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
stanford-futuredata/ColBERT canonical (classifier label)
classifier basis not recorded on the link
1 ran · 1 no recorded run
lightonai/pylate community (classifier label)
classifier basis not recorded on the link
0 ran · 10 no recorded run
FunctionStatusWhere it lives
pool_embeddings_hierarchical Ran
ran on a synthesized input
stanford-futuredata/ColBERT/colbert/modeling/checkpoint.py
code served (permissive licence) · get_code("94647c3b01781c9d")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
from scipy.cluster.hierarchy import linkage, fcluster

torch.manual_seed(0)

# Create small embedding tensor: 2 passages with varying token lengths
# Total tokens needed: first passage has 6 tokens, second has 4 tokens = 10 total
# Embedding dimension: 4
embedding_dim = 4
total_tokens = 10
p_embeddings = torch.randn(total_tokens, embedding_dim)

# Token lengths for 2 passages
token_lengths = [6, 4]

# Pool factor: reduce tokens by this factor
pool_factor = 2

# Number of protected tokens (kept at start of each passage)
protected_tokens = 1

# Show progress bar
showprogress = False

ARGS = {
    'p_embeddings': p_embeddings,
    'token_lengths': token_lengths,
    'pool_factor': pool_factor,
    'protected_tokens': protected_tokens,
    'showprogress': showprogress
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
scipy, torch, tqdm — 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
colbert_kd_scores Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/colbert.py
code served (permissive licence) · get_code("1ebe0cfae0199d86")
How it ran

no invocation recorded.

colbert_kd_scores_flash Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/_flash_backend.py
code served (permissive licence) · get_code("224990e9b3f9ef65")
How it ran

no invocation recorded.

colbert_kd_scores_lik Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/_lik_backend.py
code served (permissive licence) · get_code("c92e9e137a01d1ae")
How it ran

no invocation recorded.

colbert_scores Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/colbert.py
code served (permissive licence) · get_code("a91e1fd19b4e7554")
How it ran

no invocation recorded.

colbert_scores_flash Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/_flash_backend.py
code served (permissive licence) · get_code("2d0901aeb6f8d0f2")
How it ran

no invocation recorded.

colbert_scores_lik Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/_lik_backend.py
code served (permissive licence) · get_code("45019a2cb403e497")
How it ran

no invocation recorded.

colbert_scores_pairwise Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/colbert.py
code served (permissive licence) · get_code("25c15dd2b88dd6a3")
How it ran

no invocation recorded.

colbert_scores_pairwise_flash Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/_flash_backend.py
code served (permissive licence) · get_code("118934c70d50bc00")
How it ran

no invocation recorded.

colbert_scores_pairwise_lik Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/scores/_lik_backend.py
code served (permissive licence) · get_code("172113c1696193c3")
How it ran

no invocation recorded.

extract_skiplist_mask Not yet run
unverified (harvested, no recorded run)
lightonai/pylate/pylate/losses/contrastive.py
code served (permissive licence) · get_code("097db9fa1b9248b1")
How it ran

no invocation recorded.

find_class_names Not yet run
unverified (harvested, no recorded run)
stanford-futuredata/ColBERT/colbert/modeling/hf_colbert.py
code served (permissive licence) · get_code("92f33f7543114633")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 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 12 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

Pre-trained language models are increasingly important components across multiple information retrieval (IR) paradigms. Late interaction, introduced with the ColBERT model and recently refined in ColBERTv2, is a popular paradigm that holds state-of-the-art status across many benchmarks. To dramatically speed up the search latency of late interaction, we introduce the Performance-optimized Late Interaction Driver (PLAID). Without impacting quality, PLAID swiftly eliminates low-scoring passages using a novel centroid interaction mechanism that treats every passage as a lightweight bag of centroids. PLAID uses centroid interaction as well as centroid pruning, a mechanism for sparsifying the bag of centroids, within a highly-optimized engine to reduce late interaction search latency by up to 7$\times$ on a GPU and 45$\times$ on a CPU against vanilla ColBERTv2, while continuing to deliver state-of-the-art retrieval quality. This allows the PLAID engine with ColBERTv2 to achieve latency of tens of milliseconds on a GPU and tens or just few hundreds of milliseconds on a CPU at large scale, even at the largest scales we evaluate with 140M passages.

For agents

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

get_harvested_code_for_paper("2205.09707")
get_code_for_paper("2205.09707")
have("2205.09707")

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

curl https://syntology.ai/api/ran/2205.09707.json

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

[![Syntology run record](https://syntology.ai/api/ran/2205.09707.svg)](https://syntology.ai/paper/2205.09707)

Connect an agent — have() is free.