TextlessRAG retrieves document image pages and generates spoken answers directly from an audio query, without OCR, ASR, or TTS, and introduces a bilingual speech-document RAG benchmark.
Internvl3: Exploring advanced training and test-time recipes for open-source multimodal models,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TextlessRAG: End-to-End Visual Document RAG by Speech Without Text
TextlessRAG retrieves document image pages and generates spoken answers directly from an audio query, without OCR, ASR, or TTS, and introduces a bilingual speech-document RAG benchmark.