28 papers
DialFork: Going Beyond Factuality to Evaluate Procedural Hallucination in Conversational Agents
Sergio Burdisso, Esaú Villatoro-Tello, Dairazalia Sánchez-Cortés +4
Large language models (LLMs) deployed as conversational agents are increasingly expected to follow procedural logic. While factual hallucination has received extensive attention, a…
Trace Evolution Bench: Measuring Whether Agents Improve from Their Own Environment Traces
Fangfei Li, Chenyang Zhao, Long Wang +2
Self-evolving agents are usually scored by endpoint Gain: held-out success after an update minus success before. That difference can credit a frozen rescore as evolution. We introd…
Steering Speech-Language Models: Training-Free Task Specialization via Contrastive Activation Addition
Séverin Baroudi, Yanis Labrak, Pierfrancesco Melucci +5
Activation steering has proven effective for controlling the behavior of Large Language Models (LLMs) at inference time, but its application to SpeechLLMs remains new, and training…
Unified Shared Encoder in Spoof-Aware Speaker Verification with Hybrid Wavelet Prompt Tuning
Aref Farhadipour, Srikanth Madikeri, Teodora Vukovic +2
Spoofing-aware speaker verification (SASV) must confirm who is speaking and that the speech is genuine, but current systems gain one at the expense of the other. Modular cascades a…
UZH-CL at ArA-DF 2026: Prompt-Tuned Foundation Models and Track-Adaptive Score Fusion for Arabic Speech Deepfake Detection
Aref Farhadipour, Teodora Vukovic, Petr Motlicek
Detecting synthetic and voice-converted speech remains difficult for low-resource languages with dialectal diversity, where systems must generalize across regional dialects and uns…
Rethinking Human-Aligned Evaluation: An Analysis of Semantic Metrics Beyond WER
Hritika Sharma, Thibault Bañeras-Roux, Alessandra Pinto +4
Word Error Rate (WER), the most commonly used metric for Automatic Speech Recognition (ASR), treats every lexical deviation from the reference as equally costly, regardless of whet…
From Generative to Discriminative: Lightweight Adaptation of SpeechLLMs for Multimodal Emotion Recognition
Hasindri Watawana, Sergio Burdisso, Esaú Villatoro-Tello +4
SpeechLLMs can recognise emotion by generating a label token, but this approach relies on the LLM's existing output representations. We compare this generative approach with a disc…
Generative vs. Encoder Large Language Models for ASR Evaluation: A Comparative Study
Thibault Bañeras-Roux, Shashi Kumar, Driss Khalil +6
Automatic Speech Recognition (ASR) is typically evaluated using Word Error Rate (WER), which poorly reflects semantic similarity. While embedding-based metrics correlate better wit…
When Synthetic Speech Is All You Have: Better Call GRPO
Shashi Kumar, Yanis Labrak, Hasindri Watawana +5
LLM-based ASR adapted to regulated domains such as banking is bottlenecked by privacy: real speech is costly and legally constrained to collect, making synthetic text-to-speech (TT…
How to Leverage Synthetic Speech for LLM-Based ASR Systems?
Yanis Labrak, Dairazalia Sanchez-Cortes, Sergio Burdisso +9
In regulated domains such as banking and healthcare, where privacy constraints make real speech costly to collect and retain, synthetic speech from modern text-to-speech (TTS) is a…
Test-Time Compute Scaling for ASR with Depth-Conditioned Looped Transformers
Yacouba Kaloga, Shashi Kumar, Shakeel A. Sheikh +3
End-to-end ASR systems typically use fixed-depth acoustic encoders at inference, making it difficult to trade additional test-time computation for improved recognition without trai…
Geometric Latent Reasoning Induces Shorter Generations in LLMs
Shashi Kumar, Yacouba Kaloga, Petr Motlicek +2
Large language models solve complex problems by generating lengthy chains of explicit reasoning tokens. While effective, this makes reasoning expensive, length-sensitive, and const…
Evaluation of Automatic Speech Recognition Using Generative Large Language Models
Thibault Bañeras-Roux, Shashi Kumar, Driss Khalil +6
Automatic Speech Recognition (ASR) is traditionally evaluated using Word Error Rate (WER), a metric that is insensitive to meaning. Embedding-based semantic metrics are better corr…
Closing the Speech-Text Gap with Limited Audio for Effective Domain Adaptation in LLM-Based ASR
Thibault Bañeras-Roux, Sergio Burdisso, Esaú Villatoro-Tello +9
Conventional end-to-end automatic speech recognition (ASR) systems rely on paired speech-text data for domain adaptation. Recent LLM-based ASR architectures connect a speech encode…
Generating Synthetic Doctor-Patient Conversations for Long-form Audio Summarization
Yanis Labrak, David Grünert, Séverin Baroudi +11
Long-context audio reasoning is underserved in both training data and evaluation. Existing benchmarks target short-context tasks, and the open-ended generation tasks most relevant…
Abstract Compression of Conversational Audio Context for LLM-based ASR
Shashi Kumar, Esaú Villatoro-Tello, Sergio Burdisso +7
Standard LLM-based speech recognition systems process utterances in isolation, limiting their ability to leverage conversational context. Here we study whether multimodal context f…
When Consistency Becomes Bias: Interviewer Effects in Semi-Structured Clinical Interviews
Hasindri Watawana, Sergio Burdisso, Diego A. Moreno-Galván +4
Automatic depression detection from doctor-patient conversations has gained momentum thanks to the availability of public corpora and advances in language modeling. However, interp…
Reducing Prompt Sensitivity in LLM-based Speech Recognition Through Learnable Projection
Sergio Burdisso, Esaú Villatoro-Tello, Shashi Kumar +7
LLM-based automatic speech recognition (ASR), a well-established approach, connects speech foundation models to large language models (LLMs) through a speech-to-LLM projector, yiel…
Adaptive Multimodal Person Recognition: A Robust Framework for Handling Missing Modalities
Aref Farhadipour, Teodora Vukovic, Volker Dellwo +2
Person identification systems often rely on audio, visual, or behavioral cues, but real-world conditions frequently present with missing or degraded modalities. To address this cha…
Text-only adaptation in LLM-based ASR through text denoising
Andrés Carofilis, Sergio Burdisso, Esaú Villatoro-Tello +8
Adapting large language model (LLM)-based automatic speech recognition (ASR) systems to new domains using text-only data is a significant yet underexplored challenge. Standard fine…
SDialog: A Python Toolkit for End-to-End Agent Building, User Simulation, Dialog Generation, and Evaluation
Sergio Burdisso, Séverin Baroudi, Yanis Labrak +8
We present SDialog, an MIT-licensed open-source Python toolkit that unifies dialog generation, evaluation and mechanistic interpretability into a single end-to-end framework for bu…
Latent Space Factorization in LoRA
Shashi Kumar, Yacouba Kaloga, John Mitros +2
Low-rank adaptation (LoRA) is a widely used method for parameter-efficient finetuning. However, existing LoRA variants lack mechanisms to explicitly disambiguate task-relevant info…
TokenVerse++: Towards Flexible Multitask Learning with Dynamic Task Activation
Shashi Kumar, Srikanth Madikeri, Esaú Villatoro-Tello +8
Token-based multitasking frameworks like TokenVerse require all training utterances to have labels for all tasks, hindering their ability to leverage partially annotated datasets a…
Better Semi-supervised Learning for Multi-domain ASR Through Incremental Retraining and Data Filtering
Andres Carofilis, Pradeep Rangappa, Srikanth Madikeri +10
Fine-tuning pretrained ASR models for specific domains is challenging when labeled data is scarce. But unlabeled audio and labeled data from related domains are often available. We…