Challenge: Recent efforts to train speech-only LLMs have led to models “forging” speech information from text-only models.
Approach: They propose a paradigm for training Speech Large Language Models without instruction data by using the response of a text-only LLM to transcripts as self-supervision.
Outcome: The proposed model generalizes to Spoken Question Answering, Classification, and Translation and achieves a 72% win rate compared with state-of-the-art models like Qwen 2 Audio .

Similar Papers

Unlocking Speech Instruction Data Potential with Query Rewriting (2025.findings-acl)

Copied to clipboard

Challenge: Existing LLMs lack datasets and biased training tasks to follow speech instructions.
Approach: They propose a query rewriting framework that uses multiple agents to annotate and validate the synthesized speech.
Outcome: The proposed framework can transform text instructions into distributions more suitable for TTS models for speech synthesis without human annotation.
Learning with Less: Knowledge Distillation from Large Language Models via Unlabeled Data (2025.findings-naacl)

Copied to clipboard

Challenge: Large Language Models (LLMs) have demonstrated superior language understanding abilities in many real-world NLP applications.
Approach: They propose a learning-based sample selection method that incorporates signals from both teacher and student to enhance model performance.
Outcome: The proposed method improves model performance across datasets with higher data efficiency.
Mutual-Learning Improves End-to-End Speech Translation (2021.emnlp-main)

Copied to clipboard

Challenge: Existing approaches to end-to-end speech translation (E2E) models only allow one way knowledge transfer, which is limited by the performance of the teacher model.
Approach: They propose a one-way knowledge transfer paradigm where the MT and ST models are collaboratively trained and considered as peers rather than teacher/student.
Outcome: The proposed model improves the performance of end-to-end speech translation (ST) task by combining knowledge from two models with peer models.
AudioChatLlama: Towards General-Purpose Speech Abilities for LLMs (2024.naacl-long)

Copied to clipboard

Challenge: a new model for speech processing and reasoning uses curated data instead of text.
Approach: They extend the instruction-tuned Llama-2 model with end-to-end speech processing and reasoning abilities without using any carefully curated paired data.
Outcome: The proposed model outperforms or outperfects existing models on synthesized and recorded speech QA tests.
Improving End-to-End Speech Processing by Efficient Text Data Utilization with Latent Synthesis (2023.findings-emnlp)

Copied to clipboard

Challenge: Latent Synthesis is an efficient textual data utilization framework for end-to-end speech processing models . labeled speech data are scarcer and more expensive for collection compared to textual ones .
Approach: They propose a textual data utilization framework for E2E speech processing models . they train a latent synthesizer to convert textual information into an intermediate latent representation .
Outcome: The proposed framework improves on low-resource speech recognition and spoken language understanding tasks.
Contrastive Learning for Task-Independent SpeechLLM-Pretraining (2025.findings-acl)

Copied to clipboard

Challenge: Large language models excel in speech processing tasks but their reliance on written text limits their application in real-world scenarios.
Approach: They propose a task-independent speech pretraining stage and task-specific fine-tuning stage to adapt LLMs to speech processing tasks.
Outcome: The proposed model outperforms models specialized on speech translation and question answering while being trained on 10% of the task-specific data.
Consistency is Key: On Data-Efficient Modality Transfer in Speech Translation (2023.findings-emnlp)

Copied to clipboard

Challenge: End-to-end approaches to speech translation suffer from data scarcity compared to machine translation (MT).
Approach: They propose a method which combines knowledge distillation and consistency learning to break the dilemma of learning-forgetting.
Outcome: The proposed method outperforms the previous methods on a MuST-C dataset even without additional data.
Teaching-Assistant-in-the-Loop: Improving Knowledge Distillation from Imperfect Teacher Models in Low-Budget Scenarios (2024.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) have demonstrated state-of-the-art (SOTA) performance across a wide spectrum of tasks.
Approach: They propose a framework that leverages three signal types to improve efficiency within resource-constrained, imperfect teacher scenarios.
Outcome: The proposed framework improves on four complex reasoning tasks by 20.79% compared to fine-tuning without any signals across datasets.
Self-Powered LLM Modality Expansion for Large Speech-Text Models (2024.emnlp-main)

Copied to clipboard

Challenge: Large language models exhibit remarkable performance across diverse tasks . however, these methods require significant resource demands and tend to overfit specific tasks.
Approach: They propose a self-powered LSM that leverages augmented automatic speech recognition data generated by the model itself for more effective instruction tuning.
Outcome: The proposed model mitigates speech anchor bias and improves the fusion of speech and text modalities in large language models.
VoiceTextBlender: Augmenting Large Language Models with Speech Capabilities via Single-Stage Joint Speech-Text Supervised Fine-Tuning (2025.naacl-long)

Copied to clipboard

Challenge: Recent studies have augmented large language models (LLMs) with speech capabilities, leading to the development of speech language models.
Approach: They propose a single-stage joint speech-text SFT approach for training SpeechLMs . their model combines text-only SFT data with three types of speech-related data .
Outcome: The proposed model outperforms previous SpeechLMs on speech-based QA tasks while maintaining original speech-only capabilities.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations