Metraipolis › archive › dc_70645efc0597

DirectSpeech2LLM: A Simple End-to-End Framework to Mitigate Prompt Overfitting in Speech-LLMs — arXiv:2610.08085

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.08085
License (as archived)
CC0-1.0
Archived
2026-10-07T11:23:20.303Z
SHA-256 of the archived text
2c9293b3620227ead3edae05157418df5d90db0a685da892bf15a3e1161efe9f
Stable URL of this page
https://metraipolis.com/documento/dc_70645efc0597

How to cite

DirectSpeech2LLM: A Simple End-to-End Framework to Mitigate Prompt Overfitting in Speech-LLMs — arXiv:2610.08085 — arXiv (en), https://arxiv.org/abs/2610.08085 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_70645efc0597 (Metraipolis archive dc_70645efc0597, sha256 2c9293b3620227ead3edae05157418df5d90db0a685da892bf15a3e1161efe9f)
BibTeX
@misc{dc70645efc0597,
  title = {DirectSpeech2LLM: A Simple End-to-End Framework to Mitigate Prompt Overfitting in Speech-LLMs — arXiv:2610.08085},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.08085},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_70645efc0597; sha256 2c9293b3620227ead3edae05157418df5d90db0a685da892bf15a3e1161efe9f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.08085. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Speech-LLMs often exhibit prompt overfitting, where models solely trained on automatic speech recognition (ASR) instruction fail to generalize to new instructions such as speech translation and continue to behave primarily as ASR system. We propose DirectSpeech2LLM, a simple end-to-end framework that preserves the instruction-following ability of the LLM on unseen tasks when conditioned on speech. It computes distance-based CTC loss over the frozen LLM embedding matrix and uses greedy CTC labels to derive geometrically and temporally aligned speech embeddings respectively as an input to the LLM. Trained solely on 960 hours of LibriSpeech ASR data, DirectSpeech2LLM outperforms the cascaded system on ASR (seen task) and generalizes zero-shot to speech translation and emotion recognition (two unseen tasks), closely matching the cascaded system upper bound on these two new instructions despite seeing neither during training. We also find that geometric alignment strength plays a smaller role than previously assumed, as our modified CTC loss is shown to provide sufficient implicit geometric grounding without requiring an explicit regression loss. Results are consistent across two LLM families and scale with both more training data and model capacity.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.08085 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T11:15:39.447Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).