ATHENA is a cutting-edge research framework that revolutionizes parameter-efficient fine-tuning (PEFT) for large language and vision–language models. By introducing a novel PolyAdapter layer that ...

Figure 1: Overall architecture of our proposed Omni-AVSR method. Table 1: Main results on LRS2 and LRS3. Table 2: Comparison with unified methods on LRS3. Figure 2: Comparison with SoTA methods for ...