The DreamerV3 algorithm recently obtained remarkable performance across diverse environment domains by learning an accurate world model based on Recurrent Neural Networks (RNNs). Following the success of model-based reinforcement learning algorithms and the rapid adoption of the Transformer architecture for its superior training efficiency and favorable scaling properties, recent works such as STORM have proposed replacing RNN-based world models with Transformer-based world models using masked self-attention. However, despite the improved training efficiency of these methods, their impact on performance remains limited compared to the Dreamer algorithm, struggling to learn competitive Transformer-based world models. In this work, we show that the next state prediction objective adopted in previous approaches is insufficient to fully exploit the representation capabilities of Transformers. We propose to extend world model predictions to longer time horizons by introducing TWISTER (Transformer-based World model wIth contraSTivE Representations), a world model using action-conditioned Contrastive Predictive Coding to learn high-level temporal feature representations and improve the agent performance. TWISTER achieves a human-normalized mean score of 162% on the Atari 100k benchmark, setting a new record among state-of-the-art methods that do not employ look-ahead search. We release our code at https://github.com/burchim/TWISTER.

The Dreamer algorithm has recently obtained remarkable performance across diverse environment domains by training powerful agents with simulated trajectories. However, the compressed nature of its world model’s latent space can result in the loss of crucial information, negatively affecting the agent’s performance. Recent approaches, such as $\Delta$-IRIS and DIAMOND, address this limitation by training more accurate world models. However, these methods require training agents directly from pixels, which reduces training efficiency and prevents the agent from benefiting from the inner representations learned by the world model. In this work, we propose an alternative approach to world modeling that is both accurate and efficient. We introduce EMERALD (Efficient MaskEd latent tRAnsformer worLD model), a world model using a spatial latent state with MaskGIT predictions to generate accurate trajectories in latent space and improve the agent performance. On the Crafter benchmark, EMERALD achieves new state-of-the-art performance, becoming the first method to surpass human experts performance within 10M environment steps. Our method also succeeds to unlock all 22 Crafter achievements at least once during evaluation.

Humans are adept at leveraging visual cues from lip movements for recognizing speech in adverse listening conditions. Audio-Visual Speech Recognition (AVSR) models follow similar approach to achieve robust speech recognition in noisy conditions. In this work, we present a multilingual AVSR model incorporating several enhancements to improve performance and audio noise robustness. Notably, we adapt the recently proposed Fast Conformer model to process both audio and visual modalities using a novel hybrid CTC/RNN-T architecture. We increase the amount of audio-visual training data for six distinct languages, generating automatic transcriptions of unlabelled multilingual datasets (VoxCeleb2 and AVSpeech). Our proposed model achieves new state-of-the-art performance on the LRS3 dataset, reaching WER of 0.8%. On the recently introduced MuAViC benchmark, our model yields an absolute average-WER reduction of 11.9% in comparison to the original baseline. Finally, we demonstrate the ability of the proposed model to perform audio-only, visual-only, and audio-visual speech recognition at test time.

End-to-end Automatic Speech Recognition (ASR) systems based on neural networks have seen large improvements in recent years. The availability of large scale hand-labeled datasets and sufficient computing resources made it possible to train powerful deep neural networks, reaching very low Word Error Rate (WER) on academic benchmarks. However, despite impressive performance on clean audio samples, a drop of performance is often observed on noisy speech. In this work, we propose to improve the noise robustness of the recently proposed Efficient Conformer Connectionist Temporal Classification (CTC)-based architecture by processing both audio and visual modalities. We improve previous lip reading methods using an Efficient Conformer back-end on top of a ResNet-18 visual front-end and by adding intermediate CTC losses between blocks. We condition intermediate block features on early predictions using Inter CTC residual modules to relax the conditional independence assumption of CTC-based models. We also replace the Efficient Conformer grouped attention by a more efficient and simpler attention mechanism that we call patch attention. We experiment with publicly available Lip Reading Sentences 2 (LRS2) and Lip Reading Sentences 3 (LRS3) datasets. Our experiments show that using audio and visual modalities allows to better recognize speech in the presence of environmental noise and significantly accelerate training, reaching lower WER with 4 times less training steps. Our Audio-Visual Efficient Conformer (AVEC) model achieves state-of-the-art performance, reaching WER of 2.3% and 1.8% on LRS2 and LRS3 test sets. Code and pretrained models are available at https://github.com/burchim/AVEC.

The recently proposed Conformer architecture has shown state-of-the-art performances in Automatic Speech Recog-nition by combining convolution with attention to model both local and global dependencies. In this paper, we study how to reduce the Conformer architecture complexity with a limited computing budget, leading to a more efficient archi-tecture design that we call Efficient Conformer. We introduce progressive downsampling to the Conformer encoder and propose a novel attention mechanism named grouped attention, allowing us to reduce attention complexity from$O(n^{2}d)$to$O(n^{2}d/g)$for sequence length$n$, hidden dimension$d$and group size parameter$g$. We also experiment the use of strided multi-head self-attention as a global downsampling operation. Our experiments are performed on the LibriSpeech dataset with CTC and RNN-Transducer losses. We show that within the same computing budget, the proposed architecture achieves better performances with faster training and decoding compared to the Conformer. Our 13M parameters CTC model achieves competitive WERs of 3.6%/9.0% without using a language model and 2.7%/6.7% with an external n-gram language model on the test-clean/test-other sets while being 29%11Inference time on a single Intel Core i9-9940X 3.3GHz CPU thread faster than our CTC Conformer baseline at inference and 36% faster to train.