Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models

Full-duplex spoken dialogue models can listen and speak simultaneously, making them a promising architecture for natural conversation. However, current models are trained solely with supervised learning through token-level likelihood maximization, which does not directly optimize interaction-level behaviors, causing interactivity issues such as excessive silence and ill-timed turn-taking. Recent work has applied reinforcement learning (RL) to improve interactivity, but existing methods address only a limited set of interactive behaviors in their rewards. In this work, we propose a post-training alignment method that comprehensively improves the interactivity of full-duplex spoken dialogue models through RL. We address the four canonical axes of interactivity: pause handling, turn-taking, backchanneling, and user interruption. For each axis, we extract short audio segments from human conversation corpora and optimize the model with axis-specific reward functions. An extra LLM-based reward for response quality prevents semantic degradation. We apply our method to two open-source models, Moshi and PersonaPlex, demonstrating consistent improvements in interactivity on both offline evaluation with pre-recorded audio and real-time multi-turn dialogue evaluation.

AuRA: Internalizing Audio Understanding into LLMs as LoRA

Recent efforts to extend large language models (LLMs) to speech inputs typically rely on cascaded ASR-LLM pipelines, end-to-end speech-language models, or bridge/distillation-based adaptation. While these routes respectively reuse strong pretrained components, enable native speech-language interaction, or offer lightweight adaptation, they often suffer from transcript-interface latency, costly multimodal training, or sequential speech-language coupling. To address these limitations, we present AuRA, a method that distills audio encoding capability into the LLM. Specifically, AuRA feeds the same speech input to an ASR encoder (as a teacher) and a LoRA-adapted LLM (as a student) through a lightweight audio embedding layer, and uses layer-wise distillation to align the student's hidden states with corresponding teacher representations, thereby internalizing speech representations into lightweight LLM-side adaptations. Compared with cascaded and serial bridge methods, AuRA enables tighter speech-language joint modeling and efficient parallel end-to-end inference, while also reusing pretrained speech and language models rather than requiring large-scale multimodal training. On multiple speech-language benchmarks, AuRA consistently outperforms cascaded systems, speech-to-LLM adaptation baselines, and large-scale speech-language and multimodal models in both effectiveness and efficiency.

What Do Deepfake Speech Detectors Actually Hear?

Deepfake speech detectors often output a single score without explaining why an audio sample is flagged, where in the signal the evidence lies, or what cues drive the decision. We propose an audio-native explainability pipeline using Integrated Gradients on time-aligned self-supervised representations to localize decision evidence over time. We apply the proposed method to three WavLM-based detectors (AASIST, CA-MHFA, SLS) on ASVspoof 5 and manually annotate the highest-attribution regions to provide a semantic meaning of the most important cues. Despite similar performance, the detectors rely on different cues: AASIST emphasizes non-speech/environment cues, CA-MHFA focuses on localized phoneme artifacts, and SLS relies on word boundaries and spectral integrity. We move beyond speculative reasoning and validate our findings by causal masking of the primary detector cues. Observed performance degradation further supports the explained detector semantics.

Ethical and Technical Limits of Deepfake Speech Datasets

Claims about the robustness and fairness of deepfake speech detectors are only as credible as the datasets used to train and evaluate those systems. We present a dataset-level audit of the deepfake speech landscape. We compile and analyze 39 deepfake speech datasets, examining key attributes including accessibility, documentation, demographic and language coverage, dataset scale, and the underlying bona fide speech sources. Our audit reveals two important takeaways. Firstly, fairness assessment is largely infeasible because most datasets lack demographic metadata, and only a few contain gender or language labels. This prevents any meaningful subgroup analysis and leaves other demographic attributes unaddressed. Secondly, we identify substantial overlap in underlying bona fide source corpora across datasets, which can undermine cross-dataset evaluation and lead to overstated generalization claims.

Phoneme-First Prediction for LLM-Based Speech Recognition

Recent research has explored integrating Large Language Models (LLMs) with speech encoders to create speech-augmented LLMs capable of contextualized speech recognition. The main challenge lies in aligning the semantic embeddings of LLMs with the acoustic representations of speech encoders. We propose a novel approach that teaches the LLM to first predict phonemes from the speech features before generating the final transcript. By integrating a phoneme prediction step directly into the LLM, the model develops a fine-grained knowledge of pronunciation, reducing acoustic confusion and improving transcription accuracy and explainability. Our method is cheap and simple, as phoneme targets can be automatically derived from existing transcripts. Through comprehensive experiments, we show that intermediate phoneme prediction can improve speech recognition, particularly in low-resource settings, and yields outputs that are acoustically more faithful to the speech.

Speech Encoder Fusion for LLM-based Automatic Speech Recognition

Speech-aware large language models (LLMs) can incorporate speech through pre-trained acoustic encoders that project speech features into the LLM embedding space. While the choice of the speech encoder critically influences performance, different encoders often exhibit complementary strengths, motivating their combination. In this work, we investigate whether fusing multiple pre-trained speech encoders can enhance speech-aware LLMs for automatic speech recognition (ASR). We explore several fusion strategies beyond simple feature concatenation, including learned combinations and Transformer-based fusion architectures, and evaluate them across mono- and multilingual ASR settings as well as diarized speech recognition. Our results indicate that carefully fusing multiple parallel speech encoders improves downstream performance in all scenarios with limited computational overhead.

Towards Deep Contextual Reasoning from Broad Descriptions for ASR with Speech-LLM via Metadata-Driven Reasoning Chains

Speech recognition often fails on rare, domain-specific terms and context-related named entities. Existing contextualization techniques typically bias decoding with keywords or phrase lists, which does not scale well or exploit deeper knowledge. We propose a training method that teaches a speech-LLM to use broad descriptions (e.g. from videos) as weak semantic priors to perform contextual reasoning grounded in the audio. We build 400 hours of reasoning-augmented speech data by pairing erroneous hypotheses with video metadata and LLM-generated reasoning explanations that justify context-driven corrections. We finetune the speech-LLM to perform chain-of-thought reasoning: generate an initial transcript, then reason over the context, and finally return a corrected transcript. On held-out YouTube-derived test sets, our approach reduces errors, with specific improvements on rare words and named entities, and lays groundwork for deeper contextual reasoning in speech recognition.

Recovering the Zipfian Distribution in Unsupervised Term Discovery

Unsupervised term discovery involves segmenting unlabelled speech into word- or syllable-like units and clustering these into a lexicon of candidate types. True lexicons follow a Zipfian distribution, yet the dominant centre-based clustering approach -- K-means -- produces a more uniform distribution due to an inductive bias toward spherical clusters. In this paper we revisit graph-based clustering as a bottom-up alternative, where segment embeddings are connected by pairwise similarity and partitioned using the Leiden algorithm. We show that graph clustering substantially outperforms centre-based approaches (K-means, GMM, BIRCH) in both word- and syllable-level lexicon discovery across three languages, producing more Zipf-like distributions. Another bottom-up approach, agglomerative clustering with average linkage, also performs well, although it is computationally less efficient and allows for less control over the resulting distribution. Our work calls into question the dominance of centre-based clustering for term discovery, and promotes graph clustering as an attractive alternative.

Anchoring the Unknown: Open-Set Model Attribution via Proxy-Anchor Learning

The proliferation of text-to-speech (TTS) systems capable of generating realistic synthetic speech poses growing challenges for audio forensics. While binary deepfake detection has received considerable attention, source tracing (i.e., identifying which TTS system produced a given audio sample) remains underexplored, particularly in open-set scenarios where unknown systems may be encountered. We propose a metric learning framework based on the Proxy-Anchor loss function that operates on Wav2Vec2-BERT embeddings to learn a discriminative embedding space for TTS source attribution and out-of-distribution (OOD) detection of unseen systems. We evaluate it on the MLAAD v9 dataset spanning 140 TTS systems across 51 languages, and introduce an architecture merging strategy that groups TTS system versions into unified classes, reducing inter-class confusion. Our system achieves 99.76% accuracy on 110 in-distribution classes and a False Positive Rate (FPR@95) as low as 2.04% for OOD detection. Also, for a fair comparison against the current state of the art, we further evaluate it on the MLAAD v5 official dataset splits, improving the OOD accuracy by almost doubling it. These results demonstrate that Proxy-Anchor metric learning, combined with architecture-aware class design and post-hoc OOD scoring, provides an effective framework for forensic TTS source tracing in both closed-set and open-set settings.

Multilingual Word-Level Forced Alignment with Self-Supervised Representations and Learned Dynamic Programming

We present a method for accurate multilingual word-level forced alignment, consisting of an alignment encoder and a learned alignment decoder. The encoder integrates two representations: one from the Massively Multilingual Speech (MMS) model and another from a self-supervised phoneme boundary detector (UnSupSeg). It learns to fuse them and to estimate word-boundary probabilities over long temporal contexts. The alignment decoder is a learned dynamic programming that combines encoder outputs with segmental features over the MMS and UnSupSeg representations to infer final word boundaries. Trained iteratively on TIMIT and Buckeye, the proposed approach outperforms Montreal Forced Aligner (MFA) and MMS-based alignment on both datasets. On unseen languages (Dutch, German, and Hebrew), the proposed model achieves performance consistently better than or on par with existing alignment approaches, indicating its potential to scale to 1100+ languages supported by MMS without further training.