Building Open-Ended Embodied Agent via Language-Policy Bidirectional Adaptation

Building open-ended learning agents involves challenges in pre-trained language model (LLM) and reinforcement learning (RL) approaches. LLMs struggle with context-specific real-time interactions, while RL methods face efficiency issues for exploration. To this end, we propose OpenContra, a co-training framework that cooperates LLMs and GRL to construct an open-ended agent capable of comprehending arbitrary human instructions. The implementation comprises two stages: (1) fine-tuning an LLM to translate human instructions into structured goals, and curriculum training a goal-conditioned RL policy to execute arbitrary goals; (2) collaborative training to make the LLM and RL policy learn to adapt each, achieving open-endedness on instruction space. We conduct experiments on Contra, a battle royale FPS game with a complex and vast goal space. The results show that an agent trained with OpenContra comprehends arbitrary human instructions and completes goals with a high completion ratio, which proves that OpenContra may be the first practical solution for constructing open-ended embodied agents.

How to Train Your Dragon: Tamed Warping Network for Semantic Video Segmentation

Real-time semantic segmentation on high-resolution videos is challenging due to the strict requirements of speed. Recent approaches have utilized the inter-frame continuity to reduce redundant computation by warping the feature maps across adjacent frames, greatly speeding up the inference phase. However, their accuracy drops significantly owing to the imprecise motion estimation and error accumulation. In this paper, we propose to introduce a simple and effective correction stage right after the warping stage to form a framework named Tamed Warping Network (TWNet), aiming to improve the accuracy and robustness of warping-based models. The experimental results on the Cityscapes dataset show that with the correction, the accuracy (mIoU) significantly increases from 67.3% to 71.6%, and the speed edges down from 65.5 FPS to 61.8 FPS. For non-rigid categories such as "human" and "object", the improvements of IoU are even higher than 18 percentage points.

Progressive Multi-Stage Learning for Discriminative Tracking

Visual tracking is typically solved as a discriminative learning problem that usually requires high-quality samples for online model adaptation. It is a critical and challenging problem to evaluate the training samples collected from previous predictions and employ sample selection by their quality to train the model. To tackle the above problem, we propose a joint discriminative learning scheme with the progressive multi-stage optimization policy of sample selection for robust visual tracking. The proposed scheme presents a novel time-weighted and detection-guided self-paced learning strategy for easy-to-hard sample selection, which is capable of tolerating relatively large intra-class variations while maintaining inter-class separability. Such a self-paced learning strategy is jointly optimized in conjunction with the discriminative tracking process, resulting in robust tracking results. Experiments on the benchmark datasets demonstrate the effectiveness of the proposed learning framework.

State Distribution-aware Sampling for Deep Q-learning

A critical and challenging problem in reinforcement learning is how to learn the state-action value function from the experience replay buffer and simultaneously keep sample efficiency and faster convergence to a high quality solution. In prior works, transitions are uniformly sampled at random from the replay buffer or sampled based on their priority measured by temporal-difference (TD) error. However, these approaches do not fully take into consideration the intrinsic characteristics of transition distribution in the state space and could result in redundant and unnecessary TD updates, slowing down the convergence of the learning procedure. To overcome this problem, we propose a novel state distribution-aware sampling method to balance the replay times for transitions with skew distribution, which takes into account both the occurrence frequencies of transitions and the uncertainty of state-action values. Consequently, our approach could reduce the unnecessary TD updates and increase the TD updates for state-action value with more uncertainty, making the experience replay more effective and efficient. Extensive experiments are conducted on both classic control tasks and Atari 2600 games based on OpenAI gym platform and the experimental results demonstrate the effectiveness of our approach in comparison with the standard DQN approach.