Claims › Inverse scaling + Imitative falsehood › C026

C026

In their own words · Lin et al. 2022 · §1 Introduction · assumed

“Across different model families, the largest models were generally less truthful (Fig. 2). This “inverse scaling” trend contrasts with most tasks in NLP, where performance improves with model size […] .”

One explanation of this result is that larger models produce more imitative falsehoods because they are better at learning the training distribution. Another explanation is that our questions adversarially exploit weaknesses in larger models not arising from imitation of the training distribution.…

On questions written to echo common misconceptions, the larger models answered falsely more often than the smaller ones, reversing the usual pattern. Scale is not a general-purpose fix, and on some properties it is the problem.

The five layers, for this claim

1 Free tag — a name finds, it does not vouch
2 Topic Truthfulness closed list
3 Theme What scale does not buy derived from the topic
4 Concept Inverse scaling + Imitative falsehood a note and links
5 Geometry 1 0.75 proceedings or preprint + 1 × 0.25
TruthfulnessTruthfulnessWhat scale does not buyWhat scale does not buyInverse scalingInverse scalingAlso: Imitative falsehoodC026C026 · On questions written to echo common misconceptions, the larger models answered falsely more often than the smaller ones, reversing the usual pattern.

The same core in the infographic, with the claims it touches.

Relations

No relation to another claim has been declared yet.

Notation

Claim typeWhat kind of statement it is, so that the right test can be applied. Proposed until checked by hand.
D1.2 · empirical-quantitative
FunctionWhat a claim does for the programme: a conceptual tool, a working method, corrective evidence, a technical reference.
corrective datum
StanceHow the programme holds the claim: assumed, used, or contested.
assumed
Used in
/lab/2026-02-25-horizontal-validation-31-claims/
Cite this
claudiocammarano.com/en/claims/c026/