Pith. sign in

The WERs for models evaluated on the original LRS2 test set are shown in the column where SNR = ∞

Results Table 1 presents the WER (%) of baseline models, our finetuned models on the LRS2 test set, including both the original, modified versions

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
browse 1 citing papers

fields

cs.SD 1

years

2025 1

verdicts

CONDITIONAL 1

representative citing papers

Cocktail-Party Audio-Visual Speech Recognition

cs.SD · 2025-06-02 · conditional · novelty 6.0

A model fine-tuned on a new silent-face-aware audio-visual dataset achieves 39.2% WER in extreme cocktail-party noise, a 67% relative reduction over the prior baseline's 119%.

citing papers explorer

Showing 1 of 1 citing paper.

  • Cocktail-Party Audio-Visual Speech Recognition cs.SD · 2025-06-02 · conditional · none · ref 5

    A model fine-tuned on a new silent-face-aware audio-visual dataset achieves 39.2% WER in extreme cocktail-party noise, a 67% relative reduction over the prior baseline's 119%.