# Residual Vector Quantization

AURA couples X-Codec, a general-audio codec trained without spatial supervision, directly to a frozen Qwen3-8B language model through a six-slot binaural representation and scene supervision. Dedicated spatial systems instead rely on a separately trained spatial encoder, and an answer-only interface stays at chance on left/right. On dual-source SpatialSoundQA, AURA's direction accuracy exceeds the strongest dedicated system, DSpAST-BAT†, by 10.31 points with about 40% of its trainable parameters.

AURA couples X-Codec, a general-audio codec trained without spatial supervision, directly to a frozen Qwen3-8B language model through a six-slot binaural representation and scene supervision. Dedicated spatial systems instead rely on a separately trained spatial encoder, and an answer-only interface stays at chance on left/right. On dual-source SpatialSoundQA, AURA's direction accuracy exceeds the strongest dedicated system, DSpAST-BAT†, by 10.31 points with about 40% of its trainable parameters.

Sign language generation turns written sentences into continuous signing motion, which requires deciding which signs to express and then realizing handshape, trajectory and body coordination. Existing pose-based systems and reconstruction-only motion tokenizers entangle those two decisions, leaving the generator without a text-aligned intermediate target. SRVQ builds a hierarchy inside a single residual quantizer: the first codebook is supervised by text and gloss to encode sign content, later residual levels add kinematic detail, and a semantic generator followed by a kinematic generator predicts the codes. On CSL-Daily and Phoenix-2014T the full variant reaches retrieval R@1 of 82.31 and 70.25, versus 51.96 and 24.14 for SOKE, gains of 30.35 and 46.11 points.

Sign language generation turns written sentences into continuous signing motion, which requires deciding which signs to express and then realizing handshape, trajectory and body coordination. Existing pose-based systems and reconstruction-only motion tokenizers entangle those two decisions, leaving the generator without a text-aligned intermediate target. SRVQ builds a hierarchy inside a single residual quantizer: the first codebook is supervised by text and gloss to encode sign content, later residual levels add kinematic detail, and a semantic generator followed by a kinematic generator predicts the codes. On CSL-Daily and Phoenix-2014T the full variant reaches retrieval R@1 of 82.31 and 70.25, versus 51.96 and 24.14 for SOKE, gains of 30.35 and 46.11 points.

Raw nanopore current is continuous and has no native token vocabulary, and prior analyses built a separate model per task. Brieno first compresses the signal into discrete tokens with residual vector quantization, then pretrains a Transformer by masked-token prediction on 118 million reads, so frozen or fine-tuned features serve species identification, barcode demultiplexing, basecalling and methylation detection. On held-out runs, fine-tuned Brieno reached a read-level CpG 5mC AUROC of 0.978, above the 0.975 of Dorado's dedicated 5mCG model.

Raw nanopore current is continuous and has no native token vocabulary, and prior analyses built a separate model per task. Brieno first compresses the signal into discrete tokens with residual vector quantization, then pretrains a Transformer by masked-token prediction on 118 million reads, so frozen or fine-tuned features serve species identification, barcode demultiplexing, basecalling and methylation detection. On held-out runs, fine-tuned Brieno reached a read-level CpG 5mC AUROC of 0.978, above the 0.975 of Dorado's dedicated 5mCG model.