TL;DR
The findings suggest that staged hybrid optimization can offer a practically deployable alternative for low-resource, domain-specific dialog generation under latency, privacy, and infrastructure constraints.
This paper describes the system for Task~2 of the second Multilingual Conversational Speech Language Model (MLC-SLM) Challenge, which adapt Qwen3-Omni-30B-A3B-Instruct with a segment-evidence-aware data and post-training pipeline and obtains 90.92% accuracy on the final official evaluation set.
A novel reward function is introduced, designed to guide LLMs toward a targeted simplification style with Group Relative Policy Optimization (GRPO), that combines the SARI metric with specific penalty components.
This work studies how to make LLMs natively generate TTS-friendly text, which is frame as a preference alignment problem: instead of relying on downstream rewriting modules, this work directly align LLMs to generate text optimized for spoken delivery.
Experiments with six LLAMA and MISTRAL models show that LoRA substantially improves dialogue generation while requiring only 25-50% of the training data to recover over 90% of the final performance gains, and zero-shot evaluation with frontier LLMs shows that TalkFa remains a challenging benchmark.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
