Papers › Learning to Model and Ignore Dataset Bias with Mixed Capacity Ensembles

Learning to Model and Ignore Dataset Bias with Mixed Capacity Ensembles

7 Nov 2020Findings of the Association for Computational Linguistics 2020arXiv:2011.03856archive 2025-07-28

Christopher Clark, Mark Yatskar, Luke Zettlemoyer

Many datasets have been shown to contain incidental correlations created by idiosyncrasies in the data collection process. For example, sentence entailment datasets can have spurious word-class correlations if nearly all contradiction sentences contain the word "not", and image recognition datasets can have tell-tale object-background correlations if dogs are always indoors. In this paper, we propose a method that can automatically detect and ignore these kinds of dataset-specific patterns, which we call dataset biases. Our method trains a lower capacity model in an ensemble with a higher capacity model. During training, the lower capacity model learns to capture relatively shallow correlations, which we hypothesize are likely to reflect dataset bias. This frees the higher capacity model to focus on patterns that should generalize better. We ensure the models learn non-overlapping approaches by introducing a novel method to make them conditionally independent. Importantly, our approach does not require the bias to be known in advance. We evaluate performance on synthetic datasets, and four datasets built to penalize models that exploit known biases on textual entailment, visual question answering, and image recognition tasks. We show improvement in all settings, including a 10 point gain on the visual question answering dataset.

PaperPDFConference PDFCodeCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 18 code samples harvested from 1 repository linked to this paper: 10 ran (of which 5 constructed an object rather than computing a result; 8 with no instrument failure: 2 honoured, 1 violated, 5 with no contract checked; 2 where Syntology's instrument failed) · 8 unverified. Where Syntology's instrument failed: 2 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 18 samples from 1 repository, 10 ran (8 with no instrument failure; 2 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

chrisc36/autobias officialmentioned in paperpytorchSyntology: 18 samples, 10 ran (2 instrument-failed) · 8 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

18 samples harvested: 10 ran (of which 5 constructed an object rather than computing a result; 8 with no instrument failure: 2 honoured, 1 violated, 5 with no contract checked; 2 where Syntology's instrument failed) · 8 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

2ran · honoured contract
1ran · violated contract
2ran · our draft was wrong
5ran
8unverified

Licence: 0 of the 18 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from chrisc36/autobias. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

chrisc36/autobias official repository 18 samples, 10 ran (2 instrument-failed) · 8 unverified
ArgminFunction autobias/modules/classifier_ensembles.py ran Apache-2.0 (permissive) · 66d942a22936224d · report fixture: returned ArgminFunction · level recorded 2026-09-22
ClassifierEnsembleOutput autobias/modules/classifier_ensembles.py ran Apache-2.0 (permissive) · 1517452197d65b98 · report fixture: returned ClassifierEnsembleOutput · level recorded 2026-09-22
Configuration autobias/modules/classifier_ensembles.py ran Apache-2.0 (permissive) · 4407441e018c9820 · report fixture: returned Configuration · level recorded 2026-09-22
Layer in 20 papers autobias/modules/classifier_ensembles.py ran · metamorphic tier: deterministic Apache-2.0 (permissive) · 1788407f38a54f42 · report fixture: returned Layer · level recorded 2026-09-22
TrainOutput autobias/modules/classifier_ensembles.py ran Apache-2.0 (permissive) · 329d6081e2b7090a · report fixture: returned TrainOutput · level recorded 2026-09-22
_get_param_names autobias/modules/classifier_ensembles.py ran · our draft was wrong Apache-2.0 (permissive) · c24ddaada1a1c755 · report fixture: signature cls:callable/0/? · returned list · level recorded 2026-09-22
_to_py autobias/modules/classifier_ensembles.py ran · violated contract fingerprinted Apache-2.0 (permissive) · 01b501ef47ff7315 · report fixture: signature obj:ndarray/1/int · returned list · level recorded 2026-09-22
compute_class_prior autobias/modules/classifier_ensembles.py ran · our draft was wrong Apache-2.0 (permissive) · 3126bcabe72c01de · report fixture: signature labels:ndarray/1/int, n_classes:int/0/int · returned Tensor · level recorded 2026-09-22
cross_entropy in 92 papers
census: 6 implementations from 8 papers, 5 distinct outputs across 2 buckets, one shared input per bucket
autobias/modules/classifier_ensembles.py ran · honoured contract Apache-2.0 (permissive) · fc370fba23e6b428 · report fixture: signature logits:Tensor/2/float, targets:Tensor/1/int, weights:Tensor/1/float · returned Tensor · level recorded 2026-09-22
weighted_mean in 10 papers
census: 5 implementations from 5 papers, 4 distinct outputs across 3 buckets, one shared input per bucket this sample: bucket 1 of 3, arg 1: rank 2, kind float, dtype float32 · arg 2: rank 2, kind float, dtype float32
autobias/modules/classifier_ensembles.py ran · honoured contract fingerprinted Apache-2.0 (permissive) · 94de706a508f6295 · report fixture: signature x:Tensor/2/float, weights:Tensor/2/float · returned Tensor · level recorded 2026-09-22
ArgminTransformFunction autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · 1965ee4e5a5c3bab · report
ClfArgminEnsemble autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · 881087663bfb8da6 · report
ClfHead autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · 9dccbb193bda0913 · report
Configured autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · 4a2c1d211e8c974d · report
Predictor in 24 papers autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · 8b2698ce36b58c25 · report
_ConfiguredJSONEncoder autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · d9dc3b79245d226b · report
config_to_json autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · 44eeb9ee4b01368a · report
get_configuration autobias/modules/classifier_ensembles.py unverified Apache-2.0 (permissive) · be373e2a8ecbf2c4 · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it.

1 assembly on this paper: 1 harvested, 0 generated.

harvested Assembly for Mixed Capacity Ensembles (graph method: Mixed Capacity Ensembles). 18 bricks; individually on a fixture: 10 ran (of which 5 constructed an object rather than computing a result; 8 with no instrument failure: 2 honoured, 1 violated, 5 with no contract checked; 2 where Syntology's instrument failed) · 8 unverified. The assembly has not been run end-to-end.

  1. ClfArgminEnsemble entry · level 0 · unverified · 881087663bfb8da6
  2. ClassifierEnsembleOutput helper · level 1 · ran · 1517452197d65b98
  3. ClfHead model · level 0 · unverified · 9dccbb193bda0913
  4. Predictor helper · level 0 · unverified · 8b2698ce36b58c25
  5. TrainOutput helper · level 1 · ran · 329d6081e2b7090a
  6. compute_class_prior helper · level 1 · ran · our draft was wrong · 3126bcabe72c01de
  7. cross_entropy helper · level 1 · ran · honoured contract · fc370fba23e6b428
  8. ArgminTransformFunction helper · level 0 · unverified · 1965ee4e5a5c3bab
  9. Layer helper · level 1 · ran · 1788407f38a54f42
  10. weighted_mean helper · level 1 · ran · honoured contract · 94de706a508f6295
  11. ArgminFunction helper · level 1 · ran · 66d942a22936224d
  12. Configured helper · level 0 · unverified · 4a2c1d211e8c974d
  13. Configuration helper · level 1 · ran · 4407441e018c9820
  14. _get_param_names helper · level 1 · ran · our draft was wrong · c24ddaada1a1c755
  15. config_to_json helper · level 0 · unverified · 44eeb9ee4b01368a
  16. get_configuration helper · level 0 · unverified · be373e2a8ecbf2c4
  17. _ConfiguredJSONEncoder helper · level 0 · unverified · d9dc3b79245d226b
  18. _to_py helper · level 1 · ran · violated contract · 01b501ef47ff7315

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2011.03856") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Natural Language InferenceQuestion AnsweringSentenceVisual Question AnsweringVisual Question Answering (VQA)

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections