Determine whether compression-progress failures persist with supplied world-class information

Determine whether compression-progress objectives can still select a complete-return policy that is deficient in counterfactual information when the possible-world class is supplied to the compressor, accounting for how the compressor represents and uses that class.

Background

The paper constructs a deficient complete-return optimum for a compression-progress objective when the description language, initialization, and background information are fixed before selecting the possible-world class. The authors explicitly limit that result: if the world class is supplied to the compressor, the compressor’s representation and use of the class may change the score. Whether the same comparison remains valid in that setting is left unresolved.

References

The comparison remains open when that class is supplied to the compressor: its representation and use can change the score.

— When Do Intrinsic Rewards Lead to Exploration?  (2610.02159 - Viteri et al., 1 Oct 2026) in Section 3, paragraph on compression progress and acquired evidence