Claims › Concept · What scale does not buy

Imitative falsehood

Can a model be wrong because it learned its training data well?

A model can be wrong because it learned the training distribution well, not because it learned it badly.

Where it sits

1 Free tag — a name finds, it does not vouch
2 Topic Truthfulness closed list, one per claim
3 Theme What scale does not buy derived from the topics
4 Concept Imitative falsehood · 2 claims a note and links, or no build
5 Geometry 1 1.25 source, use, dissent

How its claims lean on each other

1 · Free tag2 · Topic3 · Theme4 · Concept5 · Geometryweight, 0.5 → 2The claimsTruthfulnessTruthfulnessWhat scale does not buyWhat scale does not buyInverse scalingInverse scalingImitative falsehoodImitative falsehoodAlso: Imitative falsehoodC026C026 · On questions written to echo common misconceptions, the larger models answered falsely more often than the smaller ones, reversing the usual pattern.C028C028 · Across the field, hallucination means generated content that is nonsensical or unfaithful to the source it was given: a definition about fidelity to the input, not about truth in the world.

The claims

  1. C026

    “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 […] .”

    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.

    Lin et al. 2022 · §1 Introduction · assumed · weight 1 · open C026

  2. C028

    “Within the context of NLP, the above definition of hallucination, the generated content that is nonsensical or unfaithful to the provided source content […] , is the most inclusive and standard.”

    Across the field, hallucination means generated content that is nonsensical or unfaithful to the source it was given: a definition about fidelity to the input, not about truth in the world. It is the broadest one in use and the one that travels best between tasks.

    Ji et al. 2022 · §2 · used · weight 1.25 · open C028

All 15 concepts →All the claims, one by one →