Research map / Large language models / Language, safety and interpretability
End-to-End Contrastive Language-Speech Pretraining Model for Long-Form Spoken Question Answering
Jiliang Hu, Zuchao Li, Baoyuan Qi, Guoming Liu, Ping Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add VQ adaptor (quantizer) (ablation) | CLSR · Spoken-SQuAD | Q-C R@1 (%): 15.29 -> 42.52 | Improved |
| add pre-trained BGE in joint training (ablation) | CLSR · Spoken-SQuAD | Q-C R@1 (%): 42.52 -> 48.10 | Improved |
| add ASR (Paraformer) pre-training (ablation) | CLSR · Spoken-SQuAD | WER (%): 17.00 -> 16.18 | Improved |
| add ASR (Paraformer) pre-training (ablation) | CLSR · Spoken-SQuAD | Q-C R@1 (%): 48.10 -> 49.00 | Improved |
| add sampler (ablation) | CLSR · Spoken-SQuAD | WER (%): 16.18 -> 15.01 | Improved |
| add sampler (ablation) | CLSR · Spoken-SQuAD | Q-C R@1 (%): 49.00 -> 49.65 | Improved |
| add post-training of BGE (ablation) | CLSR · Spoken-SQuAD | Q-C R@1 (%): 49.65 -> 49.82 | No clear effect |
| add post-training of BGE (ablation) | CLSR · Spoken-SQuAD | Q-C R@10 (%): 85.91 -> 85.83 | No clear effect |
| use CLSR text-like representation instead of direct speech-text alignment (ParaBGE) (ablation) | CLSR · Spoken-SQuAD | Q-C R@1 (%): 38.68 -> 49.82 | Improved |
| use CLSR instead of ParaBGE | CLSR · SLUE-SQA-5 | WER (%): 7.31 -> 16.69 | Hurt |
| add CLSR retrieval before Qwen-Audio for long audio | Qwen-Audio · Long-audio SQA | EM: 18.00 -> 27.60 | Improved |
| add CLSR retrieval before Qwen-Audio for long audio | Qwen-Audio · Long-audio SQA | cost time (s): 7935.00 -> 783.00 | Improved |
Closest papers by meaning
- Long-Form Speech Generation with Spoken Language Models (ICML 2025 · similarity 0.86)
- FastLongSpeech: Enhancing Large Speech-Language Models for Efficient Long-Speech Processing (NeurIPS 2025 · similarity 0.86)
- Spoken Question Answering and Speech Continuation Using Spectrogram-Powered LLM (ICLR 2024 · similarity 0.85)
- CCFQA: A Benchmark for Cross-Lingual and Cross-Modal Speech and Text Factuality Evaluation (AAAI 2026 · similarity 0.85)
- Measuring Audio's Impact on Correctness: Audio-Contribution-Aware Post-Training of Large Audio Language Models (ICLR 2026 · similarity 0.85)
- Speech Recognition Meets Large Language Model: Benchmarking, Models, and Exploration (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
