technique

Rank-to-Distill

techniqueactiveprovisionalrank-to-distill-c950ddfd·1 events·first seen 5d ago

Aliases: Rank-to-Distill

Co-occurring entities

Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data SpeechMatrix CVSS-C

More like this (12)

Weak-to-Strong Distillation distilabel Generalized Distillation ensemble distillation DeepSeek-R1-Distill-Qwen rank-1 approximation Self-Distillation DistilBERT reranking distillation DistilGPT-2 Model Distillation

Recent events (1)

4arXiv · cs.CL·5d ago·source ↗

Audio-LLM-based data filtering for speech-to-speech translation via Rank-to-Distill

A new arXiv paper proposes using audio large language models to filter noisy training data for end-to-end speech-to-speech translation (S2ST). The authors introduce a two-stage Rank-to-Distill strategy: a lightweight ranker generates pseudo-labels from noisy speech pairs, which then supervise an audio-LLM to make keep/drop decisions directly from raw audio. Experiments on CVSS-C and SpeechMatrix benchmarks show up to +1.4 ASR-BLEU improvement over unfiltered baselines.

Evaluation and Benchmarking Multimodal Progress Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data SpeechMatrix CVSS-C +1 more