acceptodds
Withdrawn from ICLR 2027

Confounder or Mediator? Nuisance Probing Needs a Null Derived from the Design

Withdrawn from ICLR 2027, so left out of the paper lists and the map.

Abstract

Asking whether a representation encodes a nuisance variable while holding a variable of interest fixed is conditional effect estimation, and it inherits every requirement of one. Probing practice meets none: the conditioning set is hand-picked, overlap is never checked, the null is chosen for convenience, and effects are read against zero. The cost is not hypothetical. On a corpus built to contain no producer effect, naive probing declares one in 100% of trials and one-covariate-at-a-time conditioning still flags 43.1% of cells, where our procedure sits at 8.2% against a nominal 5% and still recovers a genuine direct effect in 61–63% of testable cells. The difficulty is that the controlled covariate is usually a mediator, not a confounder: an ENCODE lab chooses its measurement protocol ( versus 0.097 given biosample), so conditioning on protocol blocks part of the effect being measured. We derive the conditioning set from the measured aliasing structure, report an identifiability budget as an overlap diagnostic (22 of 46 cells testable on ENCODE RNA-seq), permute within strata, and score against a nuisance-blind surrogate. On 759 ENCODE tracks this separates what neither alternative can: the producer effect is real under joint control (18 of 22 cells, balanced accuracy 0.994 against 0.402 chance), yet the model reproduces only the protocol-mediated component and essentially none of the direct one. The procedure retracted five of our own published-style claims along the way; we report them, because the pattern is the evidence.

Primary area:
unsupervised, self-supervised, semi-supervised, and supervised representation learning
Keywords:
Nuisance probing, Representation learning, Conditional effect estimation, Mediation analysis, Design-based null hypotheses

Reviews

Among:
Shape:
1234this paper, Bottom 1%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 1.00 over 3 reviews, above 0% of the others.
Reviewer_VD8COpenReview ↗
Rating
1
Confidence
4
Informativeness
1
Presentation
1
Soundness
1
Read review
Paper summary

The uploaded paper is different from the submitted title and abstract.

Critical strengths

The uploaded paper is different from the submitted title and abstract. I cannot review the intended paper.

Critical weaknesses

See above

Questions for authors

See above

Reviewer_kwyGOpenReview ↗
Rating
1
Confidence
1
Informativeness
1
Presentation
1
Soundness
1
Read review
Paper summary

The uploaded PDF is an entirely different paper from the submission title and abstract. This submission should be considered for desk rejection.

Critical strengths

N/A

Critical weaknesses

N/A

Questions for authors

N/A

Reviewer_PGJMOpenReview ↗
Rating
1
Confidence
1
Informativeness
1
Presentation
1
Soundness
1
Read review
Paper summary

The PDF title/abstract do not match the OpenReview title/abstract, and the paper is 15 pages long. The paper should be desk-rejected.

Critical strengths

The PDF title/abstract do not match the OpenReview title/abstract, and the paper is 15 pages long. The paper should be desk-rejected.

Critical weaknesses

The PDF title/abstract do not match the OpenReview title/abstract, and the paper is 15 pages long. The paper should be desk-rejected.

Questions for authors

The PDF title/abstract do not match the OpenReview title/abstract, and the paper is 15 pages long. The paper should be desk-rejected.

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Discussion (0)

Sign in to comment.