arXiv subjects

Gabriel Skantze

Publications and source records attributed to Gabriel Skantze.

At least 19 recordsLinked to original sources

arXiv2026

MuVAP: Multimodal Multiparty Voice Activity Projection for Turn-taking Prediction in the Wild

Current multiparty turn-taking models often rely on complex microphone arrays or multi-camera setups, limiting their applicability in human-robot interaction scenarios. We introduce MuVAP, a causal multimodal framework that extends Voice Activity Projection by grounding acoustic predictions in face tracks, enabling speaker-aware turn-taking predictions from a monaural audio stream and a single camera view. To address the combinatorial complexity of modeling multiple speakers, we propose Role-Relative Projection, which maps any N-speaker interaction onto a fixed current versus next floor-holder state. Because existing audiovisual datasets contain disruptive editing cuts that break causal tracking, we introduce the Audio-Visual Conversation Corpus, a 31-hour dataset of unedited, single-camera multiparty conversations. Evaluations demonstrate that MuVAP outperforms strong baselines on Shift-Hold and next-speaker prediction tasks across two- and three-speaker settings.

cs.SD↗

arXiv2026

Face versus Body Tracking for Human-Robot Interaction: An Egocentric Dataset

Meaningful human-robot interaction (HRI) requires a robot to continuously assess user engagement through persistent user tracking. However, state-of-the-art Multi-Object Tracking models are heavily optimized for surveillance or autonomous driving. A social robot faces distinct egocentric challenges, such as humans moving in unpredictable nonlinear patterns, obstructing each other, or leaving and reentering the scene. These dynamics trigger frequent identity switches (IDSW), causing the robot to lose its footing mid-conversation. To address this, we introduce a focused, custom-annotated egocentric dataset collected via the Furhat robot. We present a systematic evaluation isolating detection errors from tracking logic, comparing face versus body tracking, and assessing the impact of extended memory and appearance re-identification (ReID). Results indicate that increasing temporal memory mitigates prolonged occlusions but fails on complex dynamic events. Integrating ReID resolves complex switches but exhibits opposing effects: it substantially improves body tracking stability, yet causes facial IDSW to spike due to profile angle sensitivity. Ultimately, our optimized pipeline reduces IDSW by 49% compared to a standard tracking-by-detection baseline, effectively mitigating interaction breakdowns. As standard benchmarks lack dense, close-quarter occlusions, this work highlights the critical need for natively captured social dynamics to truly validate HRI perception models.

cs.RO↗

arXiv2026

Do Factual Recall Mechanisms Carry over from Text to Speech in Multimodal Language Models?

In recent years, several Speech Language Models (SLMs) that represent speech and written text jointly have been presented. The question then emerges about how model-internal mechanisms are similar and different when operating in the two modalities. We focus on how these systems encode, store, and retrieve factual knowledge, which has previously been investigated for text-only models. To investigate mechanisms behind the storage and recall of factual association in SLMs, we leverage Causal Mediation Analysis, a technique previously applied to text-based models. Initial results using SpiritLM, a multimodal model integrating discrete speech tokens reveal discrepancies between text-to-text and speech-to-text results, suggesting that the emergent mechanisms for factual recall are only partially carried over from the text to the speech modality. These results advance our understanding of how internal mechanisms encode factual associations in SLMs while contributing insights for improving speech-enabled AI systems.

cs.CL↗

arXiv2026

Aligning Backchannel and Dialogue Context Representations via Contrastive LLM Fine-Tuning

Backchannels (e.g., `yeah', `mhm', and `right') are short, non-interruptive feedback signals whose lexical form and prosody jointly convey pragmatic meaning. While prior computational research has largely focused on predicting backchannel timing, the relationship between lexico-prosodic form and meaning remains underexplored. We propose a two-stage framework: first, fine-tuning large language models on dialogue transcripts to derive rich contextual representations; and second, learning a joint embedding space for dialogue contexts and backchannel realizations. We evaluate alignment with human perception via triadic similarity judgments (prosodic and cross-lexical) and a context-backchannel suitability task. Our results demonstrate that the learned projections substantially improve context-backchannel retrieval compared to previous methods. In addition, they reveal that backchannel form is highly sensitive to extended conversational context and that the learned embeddings align more closely with human judgments than raw WavLM features.

cs.CL↗

arXiv2026

VoXtream2: Full-stream TTS with dynamic speaking rate control

Full-stream text-to-speech (TTS) for interactive systems must start speaking with minimal delay while remaining controllable as text arrives incrementally. We present VoXtream2, a zero-shot full-stream TTS model with dynamic speaking-rate control that can be updated mid-utterance on the fly. VoXtream2 combines a distribution matching mechanism over duration states with classifier-free guidance across conditioning signals to improve controllability and synthesis quality. Prompt-text masking enables textless audio prompting, removing the need for prompt transcription. Across standard zero-shot benchmarks and a dedicated speaking-rate test set, VoXtream2 achieves competitive objective and subjective results against public baselines despite a smaller model and less training data. In full-stream mode, it runs 4 times faster than real time with 74 ms first-packet latency on a consumer GPU.

eess.AS↗

arXiv2026

VoXtream: Full-Stream Text-to-Speech with Extremely Low Latency

We present VoXtream, a fully autoregressive, zero-shot streaming text-to-speech (TTS) system for real-time use that begins speaking from the first word. VoXtream directly maps incoming phonemes to audio tokens using a monotonic alignment scheme and a limited look-ahead that does not delay onset. Built around an incremental phoneme transformer, a temporal transformer predicting semantic and duration tokens, and a depth transformer producing acoustic tokens, VoXtream achieves, to our knowledge, the lowest initial delay among publicly available streaming TTS: 102 ms on GPU. Despite being trained on a mid-scale 9k-hour corpus, it matches or surpasses larger baselines on several metrics, while delivering competitive quality in both output- and full-streaming settings. Demo and code are available at https://herimor.github.io/voxtream.

eess.AS↗

arXiv2025

Investigating the Representation of Backchannels and Fillers in Fine-tuned Language Models

Backchannels and fillers are important linguistic expressions in dialogue, but often treated as 'noise' to be bypassed in modern transformer-based language models (LMs). Here, we study how they are represented in LMs using three fine-tuning strategies on three dialogue corpora in English and Japanese, in which backchannels and fillers are both preserved and annotated. This allows us to investigate how fine-tuning can help LMs learn these representations. We first apply clustering analysis to the learnt representation of backchannels and fillers, and find increased silhouette scores in representations from fine-tuned models, which suggests that fine-tuning enables LMs to distinguish the nuanced semantic variation in different backchannel and filler use. We also employ natural language generation metrics and qualitative analyses to verify that utterances produced by fine-tuned LMs resemble those produced by humans more closely. Our findings suggest the potential for transforming general LMs into conversational LMs that can produce human-like language more adequately.

cs.CL↗

arXiv2025

Into the Wild: When Robots Are Not Welcome

Social robots are increasingly being deployed in public spaces, where they face not only technological difficulties and unexpected user utterances, but also objections from stakeholders who may not be comfortable with introducing a robot into those spaces. We describe our difficulties with deploying a social robot in two different public settings: 1) Student services center; 2) Refugees and asylum seekers drop-in service. Although this is a failure report, in each use case we eventually managed to earn the trust of the staff and form a relationship with them, allowing us to deploy our robot and conduct our studies.

cs.RO↗

arXiv2025

Human-Robot Interaction Conversational User Enjoyment Scale (HRI CUES)

Understanding user enjoyment is crucial in human-robot interaction (HRI), as it can impact interaction quality and influence user acceptance and long-term engagement with robots, particularly in the context of conversations with social robots. However, current assessment methods rely solely on self-reported questionnaires, failing to capture interaction dynamics. This work introduces the Human-Robot Interaction Conversational User Enjoyment Scale (HRI CUES), a novel 5-point scale to assess user enjoyment from an external perspective (e.g. by an annotator) for conversations with a robot. The scale was developed through rigorous evaluations and discussions among three annotators with relevant expertise, using open-domain conversations with a companion robot that was powered by a large language model, and was applied to each conversation exchange (i.e. a robot-participant turn pair) alongside overall interaction. It was evaluated on 25 older adults' interactions with the companion robot, corresponding to 174 minutes of data, showing moderate to good alignment between annotators. Although the scale was developed and tested in the context of older adult interactions with a robot, its basis in general and non-task-specific indicators of enjoyment supports its broader applicability. The study further offers insights into understanding the nuances and challenges of assessing user enjoyment in robot interactions, and provides guidelines on applying the scale to other domains and populations. The dataset is available online.

cs.RO↗

arXiv2025

Detecting Referring Expressions in Visually Grounded Dialogue with Autoregressive Language Models

In this paper, we explore the use of a text-only, autoregressive language modeling approach for the extraction of referring expressions from visually grounded dialogue. More specifically, the aim is to investigate the extent to which the linguistic context alone can inform the detection of mentions that have a (visually perceivable) referent in the visual context of the conversation. To this end, we adapt a pretrained large language model (LLM) to perform a relatively course-grained annotation of mention spans in unfolding conversations by demarcating mention span boundaries in text via next-token prediction. Our findings indicate that even when using a moderately sized LLM, relatively small datasets, and parameter-efficient fine-tuning, a text-only approach can be effective, highlighting the relative importance of the linguistic context for this task. Nevertheless, we argue that the task represents an inherently multimodal problem and discuss limitations fundamental to unimodal approaches.

cs.CL↗

arXiv2025

"Dyadosyncrasy", Idiosyncrasy and Demographic Factors in Turn-Taking

Turn-taking in dialogue follows universal constraints but also varies significantly. This study examines how demographic (sex, age, education) and individual factors shape turn-taking using a large dataset of US English conversations (Fisher). We analyze Transition Floor Offset (TFO) and find notable interspeaker variation. Sex and age have small but significant effects female speakers and older individuals exhibit slightly shorter offsets - while education shows no effect. Lighter topics correlate with shorter TFOs. However, individual differences have a greater impact, driven by a strong idiosyncratic and an even stronger "dyadosyncratic" component - speakers in a dyad resemble each other more than they resemble themselves in different dyads. This suggests that the dyadic relationship and joint activity are the strongest determinants of TFO, outweighing demographic influences.

eess.AS↗

arXiv2025

Representation of perceived prosodic similarity of conversational feedback

Vocal feedback (e.g., `mhm', `yeah', `okay') is an important component of spoken dialogue and is crucial to ensuring common ground in conversational systems. The exact meaning of such feedback is conveyed through both lexical and prosodic form. In this work, we investigate the perceived prosodic similarity of vocal feedback with the same lexical form, and to what extent existing speech representations reflect such similarities. A triadic comparison task with recruited participants is used to measure perceived similarity of feedback responses taken from two different datasets. We find that spectral and self-supervised speech representations encode prosody better than extracted pitch features, especially in the case of feedback from the same speaker. We also find that it is possible to further condense and align the representations to human perception through contrastive learning.

cs.CL↗

arXiv2025

Perception of Emotions in Human and Robot Faces: Is the Eye Region Enough?

The increased interest in developing next-gen social robots has raised questions about the factors affecting the perception of robot emotions. This study investigates the impact of robot appearances (humanlike, mechanical) and face regions (full-face, eye-region) on human perception of robot emotions. A between-subjects user study (N = 305) was conducted where participants were asked to identify the emotions being displayed in videos of robot faces, as well as a human baseline. Our findings reveal three important insights for effective social robot face design in Human-Robot Interaction (HRI): Firstly, robots equipped with a back-projected, fully animated face - regardless of whether they are more human-like or more mechanical-looking - demonstrate a capacity for emotional expression comparable to that of humans. Secondly, the recognition accuracy of emotional expressions in both humans and robots declines when only the eye region is visible. Lastly, within the constraint of only the eye region being visible, robots with more human-like features significantly enhance emotion recognition.

cs.RO↗

arXiv2025

Yeah, Un, Oh: Continuous and Real-time Backchannel Prediction with Fine-tuning of Voice Activity Projection

In human conversations, short backchannel utterances such as "yeah" and "oh" play a crucial role in facilitating smooth and engaging dialogue. These backchannels signal attentiveness and understanding without interrupting the speaker, making their accurate prediction essential for creating more natural conversational agents. This paper proposes a novel method for real-time, continuous backchannel prediction using a fine-tuned Voice Activity Projection (VAP) model. While existing approaches have relied on turn-based or artificially balanced datasets, our approach predicts both the timing and type of backchannels in a continuous and frame-wise manner on unbalanced, real-world datasets. We first pre-train the VAP model on a general dialogue corpus to capture conversational dynamics and then fine-tune it on a specialized dataset focused on backchannel behavior. Experimental results demonstrate that our model outperforms baseline methods in both timing and type prediction tasks, achieving robust performance in real-time environments. This research offers a promising step toward more responsive and human-like dialogue systems, with implications for interactive spoken dialogue applications such as virtual assistants and robots.

cs.CL↗

arXiv2025

What Can You Say to a Robot? Capability Communication Leads to More Natural Conversations

When encountering a robot in the wild, it is not inherently clear to human users what the robot's capabilities are. When encountering misunderstandings or problems in spoken interaction, robots often just apologize and move on, without additional effort to make sure the user understands what happened. We set out to compare the effect of two speech based capability communication strategies (proactive, reactive) to a robot without such a strategy, in regard to the user's rating of and their behavior during the interaction. For this, we conducted an in-person user study with 120 participants who had three speech-based interactions with a social robot in a restaurant setting. Our results suggest that users preferred the robot communicating its capabilities proactively and adjusted their behavior in those interactions, using a more conversational interaction style while also enjoying the interaction more.

cs.RO↗

arXiv2025

Applying General Turn-taking Models to Conversational Human-Robot Interaction

Turn-taking is a fundamental aspect of conversation, but current Human-Robot Interaction (HRI) systems often rely on simplistic, silence-based models, leading to unnatural pauses and interruptions. This paper investigates, for the first time, the application of general turn-taking models, specifically TurnGPT and Voice Activity Projection (VAP), to improve conversational dynamics in HRI. These models are trained on human-human dialogue data using self-supervised learning objectives, without requiring domain-specific fine-tuning. We propose methods for using these models in tandem to predict when a robot should begin preparing responses, take turns, and handle potential interruptions. We evaluated the proposed system in a within-subject study against a traditional baseline system, using the Furhat robot with 39 adults in a conversational setting, in combination with a large language model for autonomous response generation. The results show that participants significantly prefer the proposed system, and it significantly reduces response delays and interruptions.

cs.CL↗

arXiv2024

Referring Expression Generation in Visually Grounded Dialogue with Discourse-aware Comprehension Guiding

We propose an approach to referring expression generation (REG) in visually grounded dialogue that is meant to produce referring expressions (REs) that are both discriminative and discourse-appropriate. Our method constitutes a two-stage process. First, we model REG as a text- and image-conditioned next-token prediction task. REs are autoregressively generated based on their preceding linguistic context and a visual representation of the referent. Second, we propose the use of discourse-aware comprehension guiding as part of a generate-and-rerank strategy through which candidate REs generated with our REG model are reranked based on their discourse-dependent discriminatory power. Results from our human evaluation indicate that our proposed two-stage approach is effective in producing discriminative REs, with higher performance in terms of text-image retrieval accuracy for reranked REs compared to those generated using greedy decoding.

cs.CL↗

arXiv2024

Joint Learning of Context and Feedback Embeddings in Spoken Dialogue

Short feedback responses, such as backchannels, play an important role in spoken dialogue. So far, most of the modeling of feedback responses has focused on their timing, often neglecting how their lexical and prosodic form influence their contextual appropriateness and conversational function. In this paper, we investigate the possibility of embedding short dialogue contexts and feedback responses in the same representation space using a contrastive learning objective. In our evaluation, we primarily focus on how such embeddings can be used as a context-feedback appropriateness metric and thus for feedback response ranking in U.S. English dialogues. Our results show that the model outperforms humans given the same ranking task and that the learned embeddings carry information about the conversational function of feedback responses.

cs.CL↗