SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.16163 · ICLR · 2022

The MultiBERTs: BERT Reproductions for Robustness Analysis

Alexander D'amour, Steve Yadlowsky, Jacob Eisenstein, Thibault Sellam, Ian Tenney, Jason Wei, Naomi Saphra, Tal Linzen, Jasmijn Bastings, Iulia Turc, Dipanjan Das, Ellie Pavlick

arXiv · PDF · Open in the Atlas

Code that ran

2 of 2 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; 0 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/2106.16163.json · badge: /api/ran/2106.16163.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
google-research/language marked official in Papers with Code
Papers with Code archive
2 ran · 0 no recorded run
FunctionStatusWhere it lives
_bsample Ran
ran; our contract draft was wrong, not the code
google-research/language/language/multiberts/multibootstrap.py
code served (permissive licence) · get_code("8d26a19c56d47dc6")
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 ndarray, ndim 1, dtype family int, shape (5,) 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 _bsample takes:
# 1. items: a sequence to sample from
# 2. rng: a random number generator (defaults to np.random)

# Looking at the function body:
# - It calls rng.choice(items, size=len(items), replace=True)
# - This means items needs to be a sequence with a len()
# - rng needs to have a choice method compatible with numpy's

# Create a simple list of items to sample from
items = [1, 2, 3, 4, 5]

# Create a custom RNG or use numpy's default
rng = np.random.RandomState(42)

ARGS = {
    "items": items,
    "rng": rng,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (5,) — shape (5,) 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
multibootstrap Ran
ran; our contract draft was wrong, not the code
google-research/language/language/multiberts/multibootstrap.py
code served (permissive licence) · get_code("bb9fa00b3da530ad")
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 list, ndim 2, 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.

import numpy as np
import pandas as pd

def score_fn(y_true, y_pred):
    """Simple accuracy-like score function."""
    return np.mean(np.array(y_true) == np.array(y_pred))

# Create small run_info DataFrame
# 4 runs total: 2 base (intervention=False), 2 expt (intervention=True)
# Seeds: 'seed_a', 'seed_b' for base; 'seed_a', 'seed_b' for expt (paired)
run_info = pd.DataFrame({
    'seed': ['seed_a', 'seed_a', 'seed_b', 'seed_b'],
    'intervention': [False, False, True, True]
})

# Predictions: [num_runs=4, num_examples=3]
preds = np.array([
    [0, 1, 0],
    [0, 1, 1],
    [1, 1, 0],
    [1, 0, 0]
], dtype=np.int32)

# Labels: [num_examples=3]
labels = np.array([0, 1, 0], dtype=np.int32)

# Random number generator
rng = np.random.RandomState(42)

ARGS = {
    'run_info': run_info,
    'preds': preds,
    'labels': labels,
    'score_fn': score_fn,
    'paired_seeds': True,
    'sample_seeds': True,
    'sample_examples': True,
    'nboot': 10,
    'rng': rng,
    'verbose': False,
    'progress_indicator': lambda x: x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 2, 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

Also present in other papers' code

Of 2 samples lifted from this paper's linked repositories, 2 carry a function identity and 0 do not. Of the 2 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 2 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

Experiments with pre-trained models such as BERT are often based on a single checkpoint. While the conclusions drawn apply to the artifact tested in the experiment (i.e., the particular instance of the model), it is not always clear whether they hold for the more general procedure which includes the architecture, training data, initialization scheme, and loss function. Recent work has shown that repeating the pre-training process can lead to substantially different performance, suggesting that an alternate strategy is needed to make principled statements about procedures. To enable researchers to draw more robust conclusions, we introduce the MultiBERTs, a set of 25 BERT-Base checkpoints, trained with similar hyper-parameters as the original BERT model but differing in random weight initialization and shuffling of training data. We also define the Multi-Bootstrap, a non-parametric bootstrap method for statistical inference designed for settings where there are multiple pre-trained models and limited test data. To illustrate our approach, we present a case study of gender bias in coreference resolution, in which the Multi-Bootstrap lets us measure effects that may not be detected with a single checkpoint. We release our models and statistical library, 1 along with an additional set of 140 intermediate checkpoints captured during pre-training to facilitate research on learning dynamics. * Equal contribution. † Work done as a Google AI resident. ‡ Work done during an internship at Google.

For agents

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

get_harvested_code_for_paper("2106.16163")
get_code_for_paper("2106.16163")
have("2106.16163")

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

curl https://syntology.ai/api/ran/2106.16163.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.16163.svg)](https://syntology.ai/paper/2106.16163)

Connect an agent — have() is free.