Welcome to our monthly digest, where you can catch up with any AIhub stories you may have missed, peruse the latest news, recap recent events, and more. This month, we learn about drug and material design using generative models and Bayesian optimization, find out about a system for real-time monitoring for healthcare data, and explore domain-specific distribution shifts in volunteer-collected biodiversity datasets. Ananya Joshi recently completed her PhD, where she developed a system that experts have used for the past two years to identify respiratory outbreaks (like COVID-19) in large-scale healthcare streams across the United States. In this interview, she tells us more about this project, how healthcare applications inspire basic AI research, and her future plans. Onur Boyar is a PhD student at Nagoya university, working on generative models and Bayesian methods for materials and drug design.

Adaptation of pretrained vision-language models such as CLIP to various downstream tasks have raised great interest in recent researches. Previous works have proposed a variety of test-time adaptation (TTA) methods to achieve strong generalization without any knowledge of the target domain. However, existing trainingrequired TTA approaches like TPT necessitate entropy minimization that involves large computational overhead, while training-free methods like TDA overlook the potential for information mining from the test samples themselves. In this paper, we break down the design of existing popular training-required and training-free TTA methods and bridge the gap between them within our framework. Specifically, we maintain a light-weight key-value memory for feature retrieval from instance-agnostic historical samples and instance-aware boosting samples. The historical samples are filtered from the testing data stream and serve to extract useful information from the target distribution, while the boosting samples are drawn from regional bootstrapping and capture the knowledge of the test sample itself. We theoretically justify the rationality behind our method and empirically verify its effectiveness on both the out-of-distribution and the cross-domain datasets, showcasing its applicability in real-world situations.

The table below shows the notations grouped by the modules. This work was done when Ziwei Xu was visiting the Center for Research in Computer Vision. The classes on the horizontal axis are sorted based on the performance of the task model without DTL. Dashed line shows the median performance of all classes. The annotation above (below) the line indicates the averaged improvement for classes ranked at top (bottom) 50% in the baseline performance.

We propose Differentiable Temporal Logic (DTL), a model-agnostic framework that introduces temporal constraints to deep networks. DTL treats the outputs of a network as a truth assignment of a temporal logic formula, and computes a temporal logic loss reflecting the consistency between the output and the constraints. We propose a comprehensive set of constraints, which are implicit in data annotations, and incorporate them with deep networks via DTL. We evaluate the effectiveness of DTL on the temporal action segmentation task and observe improved performance and reduced logical errors in the output of different task models. Furthermore, we provide an extensive analysis to visualize the desirable effects of DTL.

Large Language Models (LLMs) often suffer from overconfidence during inference, particularly when adapted to downstream domain-specific tasks with limited data. Previous work addresses this issue by employing approximate Bayesian estimation after the LLMs are trained, enabling them to quantify uncertainty. However, such post-training approaches' performance is severely limited by the parameters learned during training. In this paper, we go beyond post-training Bayesianization and propose Bayesian Low-Rank Adaptation by Backpropagation (BLoB), an algorithm that continuously and jointly adjusts both the mean and covariance of LLM parameters throughout the whole fine-tuning process. Our empirical results verify the effectiveness of BLoB in terms of generalization and uncertainty estimation, when evaluated on both in-distribution and out-of-distribution data.

A.1 Tabular experiments A.1.1 Implementation Details For our experiments of the FourRooms and FourRoomsTraps domains we based our implementation on [Bacon et al., 2016] and ran the experiments for 300 episodes that last a maximum of 1000 steps. As these are tabular domains, each state is defined by a single feature for both the actor and the critic. Full hyperparameters are listed here: Hyperparameter Value Actor lr 1e-1 Critic lr 1e-1 Discount 0.99 Max Steps 1000 Temperature 1e-1 GCN: hidden size 64 GCN: α 0.6 GCN: η 1e1 Table 2: Hyperparameters for the FourRooms and FourRoomsTraps domain In Fig we notice a close ressemblance in the final output. In this particular environment, performing the messages obtained the forward-backward algorithm are therefore well approximated by the propagation mechanism of the GCN. This also results in very similar empirical performance.

Potential-based reward shaping provides an approach for designing good reward functions, with the purpose of speeding up learning. However, automatically finding potential functions for complex environments is a difficult problem (in fact, of the same difficulty as learning a value function from scratch). We propose a new framework for learning potential functions by leveraging ideas from graph representation learning. Our approach relies on Graph Convolutional Networks which we use as a key ingredient in combination with the probabilistic inference view of reinforcement learning. More precisely, we leverage Graph Convolutional Networks to perform message passing from rewarding states. The propagated messages can then be used as potential functions for reward shaping to accelerate learning. We verify empirically that our approach can achieve considerable improvements in both small and high-dimensional control problems.

Thank you for the constructive feedback. We averaged the results over 10 random seeds. We will add more discussion on this to the future work section. Bi-LSTM runs considerably slower than the PPO and GCN baseline. In contrast, an RNN's output would depend on potentially all past states (in the case of LSTM/GRU this depends on the Because we essentially want to make predictions on the state space graph, local connectivity leads to better results.

Large Language Models (LLMs) have the capacity to store and recall facts. Through experimentation with open-source models, we observe that this ability to retrieve facts can be easily manipulated by changing contexts, even without altering their factual meanings. These findings highlight that LLMs might behave like an associative memory model where certain tokens in the contexts serve as clues to retrieving facts. We mathematically explore this property by studying how transformers, the building blocks of LLMs, can complete such memory tasks. We study a simple latent concept association problem with a one-layer transformer and we show theoretically and empirically that the transformer gathers information using self-attention and uses the value matrix for associative memory.

See proof of Proposition 3 below for the form of the Jacobian. Theorem 4.7] and so is the product p Intermediate steps above used the following gradient identities. Following this, we then first step is the same as the forward procedure: we solve for recover x by inverting Step 2 of the forward procedure. Since will the same in all dimensions, we can simply pick a dimension in Equation (51). C.1 UCI Data sets The main preprocessing we did was to (i) remove the "label" attribute from each data set, and (ii) remove attributes that only ever take on one value.