KIT submission to IWSLT 2026 cross-lingual voice cloning track with language tag prompting and RL fine-tuning
Researchers from KIT describe their system for the IWSLT 2026 Cross-Lingual Voice Cloning shared task, which aims to synthesize speech in a target language while preserving source-speaker identity. The system builds on FishAudio-S2-Pro, a multilingual TTS model, and introduces language tag prompting to reduce accent leakage, RL fine-tuning for intelligibility, and a reference-conditioned lexical matching method for domain-specific pronunciation. Language prompting yields the largest gains; lexical matching provides consistent improvements on matched subsets.
Related guides (1)
Related events (8)
AlignAtt4LLM adapts simultaneous speech translation policy to decoder-only LLMs for IWSLT 2026
Researchers present AlignAtt4LLM, a simultaneous speech translation system for IWSLT 2026 covering English to German, Italian, and Chinese. The system cascades Qwen3-ASR for incremental transcription with Gemma-4 E4B-it for translation, applying a novel AlignAtt policy adapted for decoder-only LLMs that lack encoder-decoder cross-attention. Key contributions include explicit source span prompting, offline alignment head selection, and query/key capture to recover a usable attention-based read/write policy. The system outperforms IWSLT 2026 baselines for European language pairs in both low- and high-latency regimes.
CUNI submits 1B-parameter simultaneous speech translation system to IWSLT 2026
Researchers from CUNI submit a simultaneous speech translation system to the IWSLT 2026 shared task, built on the offline Canary model with the AlignAtt policy. The system covers Czech-English and English-German/Italian translation pairs, supports 25 source and 25 target languages, and outperforms similarly sized baselines in both low- and high-latency regimes. At 1B parameters, it is positioned as a compact, multilingual, computationally efficient solution.
RL-based alignment improves interactivity in full-duplex spoken dialogue models
Researchers propose a post-training alignment method using reinforcement learning to improve interactivity in full-duplex spoken dialogue models, which can listen and speak simultaneously. The method addresses four canonical axes of interactivity—pause handling, turn-taking, backchanneling, and user interruption—each with axis-specific reward functions, plus an LLM-based reward to prevent semantic degradation. The approach is applied to two open-source models, Moshi and PersonaPlex, showing consistent improvements in both offline and real-time multi-turn evaluation.
Thaka Wins KSAA-2026 Arabic Speech Diacritization Task with Regularized Fine-Tuning of CATT-Whisper
The Thaka team describes their winning system for Task 2 of the KSAA-2026 Shared Task on Arabic Speech Dictation with Automatic Diacritization, which requires producing fully diacritized Arabic text from speech audio and undiacritized transcripts. Their approach fine-tunes CATT-Whisper, a multimodal model combining a CATT text encoder with a frozen Whisper speech encoder, under severe data constraints (2,327 training samples, no external data). Key techniques include R-Drop consistency regularization, Optuna-optimized hyperparameters with high weight decay, Focal Loss, and Monte Carlo Dropout inference averaging over 200 stochastic forward passes across four checkpoints. The system achieves 23.26% WER on the primary metric, placing first among all participants.
BayLing-Duplex: Native full-duplex speech dialogue using a single autoregressive LLM
Researchers introduce BayLing-Duplex, a speech language model that achieves native full-duplex interaction — simultaneous listening and speaking — using a single autoregressive LLM with no auxiliary VAD or turn-taking module. Built by fine-tuning GLM-4-Voice on 400K samples plus a lightweight DPO stage, it reaches 92% turn-taking success and 100% interruption success on InstructS2S-Eval, and improves speech-response quality substantially over Moshi. The approach adds only special tokens to the standard vocabulary, making it portable across LLM architectures without architectural changes.
Expanding on how Voice Engine works and our safety research
OpenAI published additional technical details about Voice Engine, its text-to-speech model capable of voice cloning from short audio samples. The post covers the underlying technology and safety research accompanying the system. Voice Engine has been in limited preview, with OpenAI citing concerns about misuse of voice cloning as a reason for controlled rollout.
Cross-lingual in-context learning source language selection challenges fine-tuning assumptions
A new arXiv paper conducts a broad empirical study of cross-lingual transfer in few-shot in-context learning (ICL), spanning seven tasks, six models, and a typologically diverse set of languages. The study finds that conventional heuristics from supervised fine-tuning — such as relying on linguistic similarity or data availability — do not consistently transfer to the ICL regime. The authors also analyze language confusion as a key obstacle in generative cross-lingual ICL and propose alternative heuristics for source language selection.
Thinking Machines' TML-Interaction-Small 276B-A12B Advances SOTA Realtime Voice and VAD
Thinking Machines has released TML-Interaction-Small, a 276B-A12B mixture-of-experts model targeting native interaction capabilities including realtime voice. The model is reported to advance state-of-the-art in realtime voice interaction and supersedes standard voice activity detection (VAD) approaches. The item is a brief AINews digest entry from Latent Space with minimal technical detail beyond the headline claims.