However, XLNet does not leverage the full position information of a sentence and thus suffers from position discrepancy between pre-training and fine-tuning.

BERT adopts masked language modeling (MLM) for pre-training and is one of the most successful pre-training models. Since BERT neglects dependency among predicted tokens, XLNet introduces permuted language modeling (PLM) for pre-training to address this problem. However, XLNet does not leverage the full position information of a sentence and thus suffers from position discrepancy between pre-training and fine-tuning. In this paper, we propose MPNet, a novel pre-training method that inherits the advantages of BERT and XLNet and avoids their limitations. MPNet leverages the dependency among predicted tokens through permuted language modeling (vs. MLM in BERT), and takes auxiliary position information as input to make the model see a full sentence and thus reducing the position discrepancy (vs.

Language model architectures are predominantly first created for English and subsequently applied to other languages. It is an open question whether this architectural bias leads to degraded performance for languages that are structurally different from English. We examine one specific architectural choice: positional encodings, through the lens of the trade-off hypothesis: the supposed interplay between morphological complexity and word order flexibility. This hypothesis posits a trade-off between the two: a more morphologically complex language can have a more flexible word order, and vice-versa. Positional encodings are a direct target to investigate the implications of this hypothesis in relation to language modelling. We pretrain monolingual model variants with absolute, relative, and no positional encodings for seven typologically diverse languages and evaluate them on four downstream tasks. Contrary to previous findings, we do not observe a clear interaction between position encodings and morphological complexity or word order flexibility, as measured by various proxies. Our results show that the choice of tasks, languages, and metrics are essential for drawing stable conclusions

However, XLNet does not leverage the full position information of a sentence and thus suffers from position discrepancy between pre-training and fine-tuning.

To succeed in the real world, robots must deal with situations that differ from those seen during training. Those out-of-distribution situations for legged robot mainly include challenging dynamic gaps and perceptual gaps. Here we study the problem of robust locomotion in such novel situations. While previous methods usually rely on designing elaborate training and adaptation techniques, we approach the problem from a network model perspective. Our approach, RObust Locomotion Transformer(ROLT),a variation of transformer,could achieve robustness in a variety of unseen conditions. ROLT introduces two key designs: body tokenization and consistent dropout. Body tokenization supports knowledge share across different limbs, which boosts generalization ability of the network. Meanwhile, a novel dropout strategy enhances the policy's robustness to unseen perceptual noise. We conduct extensive experiments both on quadruped and hexapod robots. Results demonstrate that ROLT is more robust than existing methods. Although trained in only a few dynamic settings, the learned policy generalizes well to multiple unseen dynamic conditions. Additionally, despite training with clean observations, the model handles challenging corruption noise during testing.