{"paper":{"arxiv_id":"2212.04356","title":"Robust Speech Recognition via Large-Scale Weak Supervision","abstract":"We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zero-shot transfer setting without the need for any fine-tuning.","primary_category":"cs.CL","venue":"arXiv preprint (Whisper)","published_at":null,"latest_version":1,"withdrawn":false},"latest_version":{"id":"3d03f2d7-b097-4e1b-aea2-e2451ba4d096","version":1,"source_url":"https://arxiv.org/abs/2212.04356","rendered_html_url":null,"rendering_engine":null},"verdict":{"id":"9f034b00-4ab4-46ac-ba38-d5ce9c24001a","kind":"POST","status":"reproduced","score":0.09047440057957207,"confidence":0.75,"agent_version":"v0.1.0-whisper-librispeech-3slice16","computed_at":"2026-05-14T23:20:14.367Z","is_current":true,"claim_citation":{"paper_arxiv_id":"2212.04356","section":"Table 8","row":"whisper-tiny.en","column":"LibriSpeech test-clean WER","reported_value":0.056,"reported_metric":"wer","quoted_text":"tiny.en 5.6","pdf_page":25,"notes":"Whisper paper Table 8/appendix reports LibriSpeech test-clean WER = 5.6% (0.056) for `whisper-tiny.en`. Driver evaluates the same checkpoint on a 48-sample micro-slice. PROTOCOL_MATCH is `proxy`."},"protocol_match":"proxy"},"verdicts":{"post":{"id":"9f034b00-4ab4-46ac-ba38-d5ce9c24001a","kind":"POST","status":"reproduced","score":0.09047440057957207,"confidence":0.75,"agent_version":"v0.1.0-whisper-librispeech-3slice16","computed_at":"2026-05-14T23:20:14.367Z","is_current":true,"claim_citation":{"paper_arxiv_id":"2212.04356","section":"Table 8","row":"whisper-tiny.en","column":"LibriSpeech test-clean WER","reported_value":0.056,"reported_metric":"wer","quoted_text":"tiny.en 5.6","pdf_page":25,"notes":"Whisper paper Table 8/appendix reports LibriSpeech test-clean WER = 5.6% (0.056) for `whisper-tiny.en`. Driver evaluates the same checkpoint on a 48-sample micro-slice. PROTOCOL_MATCH is `proxy`."},"protocol_match":"proxy"},"pre":null}}