Foundation Reinforcement Learning: towards Embodied Generalist Agents with Foundation Prior Assistance

Recently, people have shown that large-scale pre-training from internet-scale data is the key to building generalist models, as witnessed in NLP. To build embodied generalist agents, we and many other researchers hypothesize that such foundation prior is also an indispensable component. However, it is unclear what is the proper concrete form to represent those embodied foundation priors and how they should be used in the downstream task. In this paper, we propose an intuitive and effective set of embodied priors that consist of foundation policy, value, and success reward. The proposed priors are based on the goal-conditioned MDP. To verify their effectiveness, we instantiate an actor-critic method assisted by the priors, called Foundation Actor-Critic (FAC). We name our framework as Foundation Reinforcement Learning (FRL), since it completely relies on embodied foundation priors to explore, learn and reinforce. The benefits of FRL are threefold. (1) Sample efficient. With foundation priors, FAC learns significantly faster than traditional RL. Our evaluation on the Meta-World has proved that FAC can achieve 100% success rates for 7/8 tasks under less than 200k frames, which outperforms the baseline method with careful manual-designed rewards under 1M frames. (2) Robust to noisy priors. Our method tolerates the unavoidable noise in embodied foundation models. We show that FAC works well even under heavy noise or quantization errors. (3) Minimal human intervention: FAC completely learns from the foundation priors, without the need of human-specified dense reward, or providing teleoperated demos. Thus, FAC can be easily scaled up. We believe our FRL framework could enable the future robot to autonomously explore and learn without human intervention in the physical world. In summary, our proposed FRL is a novel and powerful learning paradigm, towards achieving embodied generalist agents.

Graph Neural Architecture Search with GPT-4

Graph Neural Architecture Search (GNAS) has shown promising results in automatically designing graph neural networks. However, GNAS still requires intensive human labor with rich domain knowledge to design the search space and search strategy. In this paper, we integrate GPT-4 into GNAS and propose a new GPT-4 based Graph Neural Architecture Search method (GPT4GNAS for short). The basic idea of our method is to design a new class of prompts for GPT-4 to guide GPT-4 toward the generative task of graph neural architectures. The prompts consist of descriptions of the search space, search strategy, and search feedback of GNAS. By iteratively running GPT-4 with the prompts, GPT4GNAS generates more accurate graph neural networks with fast convergence. Experimental results show that embedding GPT-4 into GNAS outperforms the state-of-the-art GNAS methods.

CasIL: Cognizing and Imitating Skills via a Dual Cognition-Action Architecture

Enabling robots to effectively imitate expert skills in longhorizon tasks such as locomotion, manipulation, and more, poses a long-standing challenge. Existing imitation learning (IL) approaches for robots still grapple with sub-optimal performance in complex tasks. In this paper, we consider how this challenge can be addressed within the human cognitive priors. Heuristically, we extend the usual notion of action to a dual Cognition (high-level)-Action (low-level) architecture by introducing intuitive human cognitive priors, and propose a novel skill IL framework through human-robot interaction, called Cognition-Action-based Skill Imitation Learning (CasIL), for the robotic agent to effectively cognize and imitate the critical skills from raw visual demonstrations. CasIL enables both cognition and action imitation, while high-level skill cognition explicitly guides low-level primitive actions, providing robustness and reliability to the entire skill IL process. We evaluated our method on MuJoCo and RLBench benchmarks, as well as on the obstacle avoidance and point-goal navigation tasks for quadrupedal robot locomotion. Experimental results show that our CasIL consistently achieves competitive and robust skill imitation capability compared to other counterparts in a variety of long-horizon robotic tasks.

OpenMSD: Towards Multilingual Scientific Documents Similarity Measurement

We develop and evaluate multilingual scientific documents similarity measurement models in this work. Such models can be used to find related works in different languages, which can help multilingual researchers find and explore papers more efficiently. We propose the first multilingual scientific documents dataset, Open-access Multilingual Scientific Documents (OpenMSD), which has 74M papers in 103 languages and 778M citation pairs. With OpenMSD, we pretrain science-specialized language models, and explore different strategies to derive "related" paper pairs to fine-tune the models, including using a mixture of citation, co-citation, and bibliographic-coupling pairs. To further improve the models' performance for non-English papers, we explore the use of generative language models to enrich the non-English papers with English summaries. This allows us to leverage the models' English capabilities to create better representations for non-English papers. Our best model significantly outperforms strong baselines by 7-16% (in mean average precision).

Exploring Flat Minima for Domain Generalization with Large Learning Rates

Domain Generalization (DG) aims to generalize to arbitrary unseen domains. A promising approach to improve model generalization in DG is the identification of flat minima. One typical method for this task is SWAD, which involves averaging weights along the training trajectory. However, the success of weight averaging depends on the diversity of weights, which is limited when training with a small learning rate. Instead, we observe that leveraging a large learning rate can simultaneously promote weight diversity and facilitate the identification of flat regions in the loss landscape. However, employing a large learning rate suffers from the convergence problem, which cannot be resolved by simply averaging the training weights. To address this issue, we introduce a training strategy called Lookahead which involves the weight interpolation, instead of average, between fast and slow weights. The fast weight explores the weight space with a large learning rate, which is not converged while the slow weight interpolates with it to ensure the convergence. Besides, weight interpolation also helps identify flat minima by implicitly optimizing the local entropy loss that measures flatness. To further prevent overfitting during training, we propose two variants to regularize the training weight with weighted averaged weight or with accumulated history weight. Taking advantage of this new perspective, our methods achieve state-of-the-art performance on both classification and semantic segmentation domain generalization benchmarks. The code is available at https://github.com/koncle/DG-with-Large-LR.

Theoretical Explanation of Activation Sparsity through Flat Minima and Adversarial Robustness

A recent empirical observation of activation sparsity in MLP layers offers an opportunity to drastically reduce computation costs for free. Despite several works attributing it to training dynamics, the theoretical explanation of activation sparsity's emergence is restricted to shallow networks, small training steps well as modified training, even though the sparsity has been found in deep models trained by vanilla protocols for large steps. To fill the three gaps, we propose the notion of gradient sparsity as the source of activation sparsity and a theoretical explanation based on it that explains gradient sparsity and then activation sparsity as necessary steps to adversarial robustness w.r.t. hidden features and parameters, which is approximately the flatness of minima for well-learned models. The theory applies to standardly trained LayerNorm-ed pure MLPs, and further to Transformers or other architectures if noises are added to weights during training. To eliminate other sources of flatness when arguing sparsities' necessity, we discover the phenomenon of spectral concentration, i.e., the ratio between the largest and the smallest non-zero singular values of weight matrices is small. We utilize random matrix theory (RMT) as a powerful theoretical tool to analyze stochastic gradient noises and discuss the emergence of spectral concentration. With these insights, we propose two plug-and-play modules for both training from scratch and sparsity finetuning, as well as one radical modification that only applies to from-scratch training. Another under-testing module for both sparsity and flatness is also immediate from our theories. Validational experiments are conducted to verify our explanation. Experiments for productivity demonstrate modifications' improvement in sparsity, indicating further theoretical cost reduction in both training and inference.

InsertNeRF: Instilling Generalizability into NeRF with HyperNet Modules

Generalizing Neural Radiance Fields (NeRF) to new scenes is a significant challenge that existing approaches struggle to address without extensive modifications to vanilla NeRF framework. We introduce InsertNeRF, a method for INStilling gEneRalizabiliTy into NeRF. By utilizing multiple plug-and-play HyperNet modules, InsertNeRF dynamically tailors NeRF's weights to specific reference scenes, transforming multi-scale sampling-aware features into scene-specific representations. This novel design allows for more accurate and efficient representations of complex appearances and geometries. Experiments show that this method not only achieves superior generalization performance but also provides a flexible pathway for integration with other NeRF-like systems, even in sparse input settings. Code will be available https://github.com/bbbbby-99/InsertNeRF.

IOMatch: Simplifying Open-Set Semi-Supervised Learning with Joint Inliers and Outliers Utilization

Semi-supervised learning (SSL) aims to leverage massive unlabeled data when labels are expensive to obtain. Unfortunately, in many real-world applications, the collected unlabeled data will inevitably contain unseen-class outliers not belonging to any of the labeled classes. To deal with the challenging open-set SSL task, the mainstream methods tend to first detect outliers and then filter them out. However, we observe a surprising fact that such approach could result in more severe performance degradation when labels are extremely scarce, as the unreliable outlier detector may wrongly exclude a considerable portion of valuable inliers. To tackle with this issue, we introduce a novel open-set SSL framework, IOMatch, which can jointly utilize inliers and outliers, even when it is difficult to distinguish exactly between them. Specifically, we propose to employ a multi-binary classifier in combination with the standard closed-set classifier for producing unified open-set classification targets, which regard all outliers as a single new class. By adopting these targets as open-set pseudo-labels, we optimize an open-set classifier with all unlabeled samples including both inliers and outliers. Extensive experiments have shown that IOMatch significantly outperforms the baseline methods across different benchmark datasets and different settings despite its remarkable simplicity. Our code and models are available at https://github.com/nukezil/IOMatch.

Efficient Last-iterate Convergence Algorithms in Solving Games

No-regret algorithms are popular for learning Nash equilibrium (NE) in two-player zero-sum normal-form games (NFGs) and extensive-form games (EFGs). Many recent works consider the last-iterate convergence no-regret algorithms. Among them, the two most famous algorithms are Optimistic Gradient Descent Ascent (OGDA) and Optimistic Multiplicative Weight Update (OMWU). However, OGDA has high per-iteration complexity. OMWU exhibits a lower per-iteration complexity but poorer empirical performance, and its convergence holds only when NE is unique. Recent works propose a Reward Transformation (RT) framework for MWU, which removes the uniqueness condition and achieves competitive performance with OMWU. Unfortunately, RT-based algorithms perform worse than OGDA under the same number of iterations, and their convergence guarantee is based on the continuous-time feedback assumption, which does not hold in most scenarios. To address these issues, we provide a closer analysis of the RT framework, which holds for both continuous and discrete-time feedback. We demonstrate that the essence of the RT framework is to transform the problem of learning NE in the original game into a series of strongly convex-concave optimization problems (SCCPs). We show that the bottleneck of RT-based algorithms is the speed of solving SCCPs. To improve the their empirical performance, we design a novel transformation method to enable the SCCPs can be solved by Regret Matching+ (RM+), a no-regret algorithm with better empirical performance, resulting in Reward Transformation RM+ (RTRM+). RTRM+ enjoys last-iterate convergence under the discrete-time feedback setting. Using the counterfactual regret decomposition framework, we propose Reward Transformation CFR+ (RTCFR+) to extend RTRM+ to EFGs. Experimental results show that our algorithms significantly outperform existing last-iterate convergence algorithms and RM+ (CFR+).