Aligning large language models (LLMs) with human objectives is crucial for real-world applications. However, fine-tuning LLMs for alignment often suffers from unstable training and requires substantial computing resources. Test-time alignment techniques, such as prompting and guided decoding, do not modify the underlying model, and their performance remains dependent on the original model's capabilities. To address these challenges, we propose aligning LLMs through representation editing. The core of our method is to view a pre-trained autoregressive LLM as a discrete-time stochastic dynamical system. To achieve alignment for specific objectives, we introduce external control signals into the state space of this language dynamical system. We train a value function directly on the hidden states according to the Bellman equation, enabling gradient-based optimization to obtain the optimal control signals at test time. Our experiments demonstrate that our method outperforms existing test-time alignment techniques while requiring significantly fewer resources compared to fine-tuning methods.

When machine learning training is outsourced to third parties, backdoor attacks become practical as the third party who trains the model may act maliciously to inject hidden behaviors into the otherwise accurate model. Until now, the mechanism to inject backdoors has been limited to poisoning. We argue that a supplychain attacker has more attack techniques available by introducing a handcrafted attack that directly manipulates a model's weights. This direct modification gives our attacker more degrees of freedom compared to poisoning, and we show it can be used to evade many backdoor detection or removal defenses effectively. Across four datasets and four network architectures our backdoor attacks maintain an attack success rate above 96%. Our results suggest that further research is needed for understanding the complete space of supply-chain backdoor attacks.

A.1 Convergences We show the convergences on validation in terms of timing (seconds) in Figure 1 and Figure 2. Basically, our algorithms converge to much better results in nearly same duration compared with other "layer sampling" approaches. Note that we cannot complete the training of AS-GAT on Reddit because of memory issues. A.2 Discussions on Timings between Layer Sampling and Graph Sampling Paradigms Note that the comparisons of timing between "graph sampling" and "layer sampling" paradigms have been studied recently in [1, 3]. As a result, we do not compare the timing with "graph sampling" approaches. Under certain conditions, the graph sampling approaches should be faster than layer sampling approaches.

Multimodal models trained on modality-complete data are plagued with severe performance degradation when encountering modality-missing data. Prevalent cross-modal knowledge distillation-based methods precisely align the representation of modality-missing data and that of its modality-complete counterpart to enhance robustness. However, due to the irreparable information asymmetry, this determinate alignment is too stringent, easily inducing modality-missing features to capture spurious factors erroneously. In this paper, a novel multimodal Probabilistic Conformal Distillation (PCD) method is proposed, which considers the inherent indeterminacy in this alignment. Given a modality-missing input, our goal is to learn the unknown Probability Density Function (PDF) of the mapped variables in the modality-complete space, rather than relying on the brute-force point alignment. Specifically, PCD models the modality-missing feature as a probabilistic distribution, enabling it to satisfy two characteristics of the PDF. One is the extremes of probabilities of modality-complete feature points on the PDF, and the other is the geometric consistency between the modeled distributions and the peak points of different PDFs. Extensive experiments on a range of benchmark datasets demonstrate the superiority of PCD over state-of-the-art methods. Code is available at: https://github.com/mxchen-mc/PCD.

Learning quickly is of great importance for machine intelligence deployed in online platforms. With the capability of transferring knowledge from learned tasks, meta-learning has shown its effectiveness in online scenarios by continuously updating the model with the learned prior. However, current online meta-learning algorithms are limited to learn a globally-shared meta-learner, which may lead to sub-optimal results when the tasks contain heterogeneous information that are distinct by nature and difficult to share. We overcome this limitation by proposing an online structured meta-learning (OSML) framework. Inspired by the knowledge organization of human and hierarchical feature representation, OSML explicitly disentangles the meta-learner as a meta-hierarchical graph with different knowledge blocks. When a new task is encountered, it constructs a meta-knowledge pathway by either utilizing the most relevant knowledge blocks or exploring new blocks. Through the meta-knowledge pathway, the model is able to quickly adapt to the new task. In addition, new knowledge is further incorporated into the selected blocks. Experiments on three datasets demonstrate the effectiveness and interpretability of our proposed framework in the context of both homogeneous and heterogeneous tasks.

We thank the reviewers for their insightful feedback! This is an important step towards a framework which targets computational resources "to reduce uncertainty as Bayesian since it relies on Bayes' theorem. This can be seen by recognizing that the posterior (see Section 2.1) This results in the proposed prior class in eq. We will clarify this in the final version. Similar reasoning applies for the inverse.

Molecular relational learning (MRL) is crucial for understanding the interaction behaviors between molecular pairs, a critical aspect of drug discovery and development. However, the large feasible model space of MRL poses significant challenges to benchmarking, and existing MRL frameworks face limitations in flexibility and scope. To address these challenges, avoid repetitive coding efforts, and ensure fair comparison of models, we introduce FlexMol, a comprehensive toolkit designed to facilitate the construction and evaluation of diverse model architectures across various datasets and performance metrics. FlexMol offers a robust suite of preset model components, including 16 drug encoders, 13 protein sequence encoders, 9 protein structure encoders, and 7 interaction layers. With its easy-to-use API and flexibility, FlexMol supports the dynamic construction of over 70, 000 distinct combinations of model architectures. Additionally, we provide detailed benchmark results and code examples to demonstrate FlexMol's effectiveness in simplifying and standardizing MRL model development and comparison.

In this paper, we propose a novel conceptual framework to detect outliers using optimal transport with a concave cost function. Conventional outlier detection approaches typically use a two-stage procedure: first, outliers are detected and removed, and then estimation is performed on the cleaned data. However, this approach does not inform outlier removal with the estimation task, leaving room for improvement. To address this limitation, we propose an automatic outlier rectification mechanism that integrates rectification and estimation within a joint optimization framework. We take the first step to utilize the optimal transport distance with a concave cost function to construct a rectification set in the space of probability distributions. Then, we select the best distribution within the rectification set to perform the estimation task. Notably, the concave cost function we introduced in this paper is the key to making our estimator effectively identify the outlier during the optimization process. We demonstrate the effectiveness of our approach over conventional approaches in simulations and empirical analyses for mean estimation, least absolute regression, and the fitting of option implied volatility surfaces.

Neural networks lack adversarial robustness, i.e., they are vulnerable to adversarial examples that through small perturbations to inputs cause incorrect predictions. Further, trust is undermined when models give miscalibrated predictions, i.e., the predicted probability is not a good indicator of how much we should trust our model. In this paper, we study the connection between adversarial robustness and calibration and find that the inputs for which the model is sensitive to small perturbations (are easily attacked) are more likely to have poorly calibrated predictions. Based on this insight, we examine if calibration can be improved by addressing those adversarially unrobust inputs. To this end, we propose Adversarial Robustness based Adaptive Label Smoothing (AR-AdaLS) that integrates the correlations of adversarial robustness and calibration into training by adaptively softening labels for an example based on how easily it can be attacked by an adversary. We find that our method, taking the adversarial robustness of the in-distribution data into consideration, leads to better calibration over the model even under distributional shifts. In addition, AR-AdaLS can also be applied to an ensemble model to further improve model calibration.

We propose a novel method for learning representations of poses for 3D deformable objects, which specializes in 1) disentangling pose information from the object's identity, 2) facilitating the learning of pose variations, and 3) transferring pose information to other object identities. Based on these properties, our method enables the generation of 3D deformable objects with diversity in both identities and poses, using variations of a single object. It does not require explicit shape parameterization such as skeletons or joints, point-level or shape-level correspondence supervision, or variations of the target object for pose transfer. We first design the pose extractor to represent the pose as a keypoint-based hybrid representation and the pose applier to learn an implicit deformation field. To better distill pose information from the object's geometry, we propose the implicit pose applier to output an intrinsic mesh property, the face Jacobian. Once the extracted pose information is transferred to the target object, the pose applier is fine-tuned in a self-supervised manner to better describe the target object's shapes with pose variations. The extracted poses are also used to train a cascaded diffusion model to enable the generation of novel poses. Our experiments with the DeformThings4D and Human datasets demonstrate state-of-the-art performance in pose transfer and the ability to generate diverse deformed shapes with various objects and poses.