Research Studios Austria

Finite-sample decision risk in unseeded quantum compiler benchmarking

Version 4 — audited release. Archival technical preprint; not submitted to a venue. Quantum compiler benchmarks are used to accept or reject changes to production transpilers, but the compilers they measure are stochastic. At the pinned revision of IBM's Benchpress benchmark suit...

Finite-sample decision risk in unseeded quantum compiler benchmarking

Version 4 — audited release. Archival technical preprint; not submitted to a venue. Quantum compiler benchmarks are used to accept or reject changes to production transpilers, but the compilers they measure are stochastic. At the pinned revision of IBM's Benchpress benchmark suit...

Continuous vocal-attribute traversals fall below the path resolution of current speech representations: a pre-registered validity gate across three speaker encoders and three neural codecs

Study 3 of the BedVibe vocal-representation series. Voice conversion, attribute editing and speaker-blend conditioning all move along a straight line between a source and a target point in a learned representation, and treat the intervening coordinates as natural intermediate voi...

The Loss Curve Is Not a Sufficient Statistic: Silent Objective Failures from Sentinel-Class Collisions in Neural Codec Language Models

Masked cross-entropy is the standard training objective for sequence models. Positions that are not real targets are excluded by comparing each target against a sentinel value, conventionally ignore_index = -100 in PyTorch. We show that when this sentinel is also a valid class in...

The Loss Curve Is Not a Sufficient Statistic: Silent Objective Failures from Sentinel-Class Collisions in Neural Codec Language Models

Masked cross-entropy is the standard training objective for sequence models. Positions that are not real targets are excluded by comparing each target against a sentinel value, conventionally ignore_index = -100 in PyTorch. We show that when this sentinel is also a valid class in...

Fusion at scale: Critical enablers in global energy transitions

Fusion is a promising sustainable energy source that can potentially contribute to achieving the most ambitious climate goals with firm carbon-free baseload electricity. Due to recent advancements, fusion is transitioning from a purely research stage to a pre-commercialisation ph...

Deterministic Language Architecture: A Formal Category Definition and Architectural Specification for Non-Probabilistic Language Generation Systems

Canon³ — The Lume Synthetic Organism Canon (L-SOC) Language Architecture Volume I. This paper formally defines Deterministic Language Architecture (DLA) — a new category of language generation system in which output is governed by deterministic architecture rather than probabilis...

Deterministic Language Architecture: A Formal Category Definition and Architectural Specification for Non-Probabilistic Language Generation Systems

Canon³ — The Lume Synthetic Organism Canon (L-SOC) Language Architecture Volume I. This paper formally defines Deterministic Language Architecture (DLA) — a new category of language generation system in which output is governed by deterministic architecture rather than probabilis...