SpeechCombine

This repository contains the model checkpoints, sample training data, and evaluation data for
the paper [Unlocking Speech–Text Compositional Powers: Instruction-Following Speech Language Models without Instruction Tuning]

📁 Repository structure

• Exps/: SpeechCombine checkpoint and tokenizer
• eval/: Eevaluation data and evaluator checkpoint
• tts/: TTS checkpoint and speaker embedding
• training_data.jsonl: SpeechCombine training data

🔗 Citation

If this resource is useful for your research, please consider citing our paper.

---

License: CC BY-NC 4.0