The study uncovers latent capabilities of language models via causal representation learning.
problem Rigorous causal evaluations of language model capabilities are challenging due to confounding effects and computational costs.
method Proposes a causal representation learning framework to identify latent capability factors as causally interrelated after controlling for a common confounder (base model).
result Identifies a three-node linear causal structure explaining performance variations across 1500 models and six benchmarks.
LaTRO optimizes latent reasoning in LLMs without external reward.
problem Training LLMs to perform complex reasoning tasks.
method Formulates reasoning as latent distribution sampling and optimizes via variational approaches.
result LLMs improve reasoning and evaluation quality through self-improvement.
New framework transfers latent knowledge from weak to strong models.
problem Aligning superhuman LLMs with human feedback.
method Transfer learning framework using refinement approach.
result Proves weak-to-strong generalization is possible.
ContextBench benchmarks methods for generating linguistically fluent inputs that activate specific latent features in language models.
problem Identifying inputs that trigger specific behaviours or latent features in language models.
method Context modification and benchmarking methods like Evolutionary Prompt Optimisation (EPO) with LLM-assistance and diffusion model inpainting.
result Enhanced methods achieve state-of-the-art performance in balancing elicitation effectiveness and fluency.
Variational autoencoders (VAEs) are widely used deep generative models capable of learning unsupervised latent representations of data. Such representations are often difficult to interpret or control. We consider the problem of unsupervised learning of features correlated to specific labels in a dataset. We propose a …
Music Inpainting is the task of filling in missing or lost information in a piece of music. We investigate this task from an interactive music creation perspective. To this end, a novel deep learning-based approach for musical score inpainting is proposed. The designed model takes both past and future musical context i…
CIPNN model tackles continuous latent variables, solving intractable posterior problems.
problem Solving intractable posterior calculation for continuous latent variables.
method Derives analytical solution for posterior of continuous latent variables, proposes CIPNN and CIPAE.
result CIPNN model demonstrates great classification capability, solving problems for continuous latent variables.
Quantitatively assessing relationships between latent variables and observed variables is important for understanding and developing generative models and representation learning. In this paper, we propose latent-observed dissimilarity (LOD) to evaluate the dissimilarity between the probabilistic characteristics of lat…
An ability to model a generative process and learn a latent representation for speech in an unsupervised fashion will be crucial to process vast quantities of unlabelled speech data. Recently, deep probabilistic generative models such as Variational Autoencoders (VAEs) have achieved tremendous success in modeling natur…
Proposes a VAE with a discrete bottleneck for better text generation.
problem VAEs struggle with latent variable auto-regressive decoding in text generation.
method Introduces a discretized bottleneck to enforce latent feature matching in a compact space.
result Demonstrates improved text generation capabilities across various tasks.
Model infers latent variables in sparse coding models using Langevin dynamics.
problem Sampling posterior distribution in sparse coding models.
method Langevin dynamics for inference and simultaneous learning of parameters.
result Langevin dynamics efficiently sample from 'L0 sparse' posterior distribution.
Improved noise estimation in latent neural SDEs enhances model accuracy.
problem Latent neural SDEs underestimate noise, limiting their stochastic dynamics modeling.
method Explicit additional noise regularization in the loss function.
result Model accurately captures diffusion component of stochastic time series data.
We present the Mixed Likelihood Gaussian process latent variable model (GP-LVM), capable of modeling data with attributes of different types. The standard formulation of GP-LVM assumes that each observation is drawn from a Gaussian distribution, which makes the model unsuited for data with e.g. categorical or nominal a…
New framework aligns latent representations over-the-air using intelligent metasurfaces.
problem Heterogeneous transmitter-receiver models produce misaligned latent representations in semantic communication.
method Intelligent metasurfaces (SIM) emulate supervised and zero-shot semantic aligners directly in the wave domain.
result SIMs achieve up to 90% task accuracy in high SNR regimes, robust to low SNR.
We extend the framework of variational autoencoders to represent transformations explicitly in the latent space. In the family of hierarchical graphical models that emerges, the latent space is populated by higher order objects that are inferred jointly with the latent representations they act on. To explicitly demonst…
As deep Variational Auto-Encoder (VAE) frameworks become more widely used for modeling biomolecular simulation data, we emphasize the capability of the VAE architecture to concurrently maximize the timescale of the latent space while inferring a reduced coordinate, which assists in finding slow processes as according t…
In this paper we present a fully Bayesian latent variable model which exploits conditional nonlinear(in)-dependence structures to learn an efficient latent representation. The latent space is factorized to represent shared and private information from multiple views of the data. In contrast to previous approaches, we i…
LSI enables joint learning of latent variables and generative models.
problem Joint optimization of latent variables and generative models.
method Developed a principled ELBO objective in continuous time for joint learning.
result LSI learns effective latent representations and generative transformations.
Improves NF for complex data distributions with multiple modes.
problem Difficulty in handling data distributions with multiple isolated modes.
method Proposes a new framework using variational latent representation to improve NF.
result Significantly more powerful for generating data distributions with multiple modes.
DLFM models complex systems with uncertainty, outperforming traditional methods.
problem Modeling highly nonlinear dynamical systems with robust uncertainty quantification.
method Deep latent force model (DLFM) using physics-informed kernels derived from ODEs.
result DLFM achieves comparable performance to non-physics-informed models on univariate tasks and captures dynamics in real-world data.
This paper uses Factored Latent Analysis (FLA) to learn a factorized, segmental representation for observations of tracked objects over time. Factored Latent Analysis is latent class analysis in which the observation space is subdivided and each aspect of the original space is represented by a separate latent class mod…
Proves identifiability of deep latent variable models without auxiliary information.
problem Identify deep generative models without side information.
method Analyzes a broad class of deep latent variable models with universal approximation capabilities.
result Identifiability of generative models without side information u. We address the problem of one-to-many mappings in supervised learning, where a single instance has many different solutions of possibly equal cost. The framework of conditional variational autoencoders describes a class of methods to tackle such structured-prediction tasks by means of latent variables. We propose to in…
Developing a dialogue agent that is capable of making autonomous decisions and communicating by natural language is one of the long-term goals of machine learning research. Traditional approaches either rely on hand-crafting a small state-action set for applying reinforcement learning that is not scalable or constructi…
Novel approach detects early warning indicators in complex systems.
problem Detecting abrupt transitions in complex systems.
method Directed anisotropic diffusion map and latent stochastic dynamical systems.
result Early warning indicators can detect tipping points in state transitions.
VAELLS learns latent manifold structure to improve VAE model accuracy.
problem VAEs struggle with mismatched latent structure and global structure.
method Integrates learnable manifold model into latent space of VAE.
result Improves model accuracy by matching prior to data manifold structure.
New method uses cycle consistency to enforce invariance in latent space.
problem Learning meaningful and independent factors of variation in datasets.
method Two separate latent subspaces, cycle consistency constraints, deep information bottleneck.
result Identifies more meaningful factors leading to sparser and interpretable models.
IntroVAC learns interpretable latent subspaces for better image quality.
problem Difficulties in interpreting latent spaces and limitations in image generation.
method Introspective Variational Classifier (IntroVAC) using additional labels and adversarial training.
result Improved image quality and meaningful latent directions for fine-grained manipulation.
Paper presents a reparameterized DP-DLGMM for clustering.
problem Non-parametric DP priors in DLGMM are hard to couple with variational inference.
method Closed-form updates for DP-DLGMM's variational posterior.
result Model generates realistic samples and performs competitively in semi-supervised settings.
Proposes CTSDG model for better vehicle intention prediction across domains.
problem Domain generalization for vehicle intention prediction in dynamic environments.
method Structural causal model with recurrent latent variable integration.
result Consistent improvement in prediction accuracy compared to state-of-the-art methods.
We present a non-parametric Bayesian latent variable model capable of learning dependency structures across dimensions in a multivariate setting. Our approach is based on flexible Gaussian process priors for the generative mappings and interchangeable Dirichlet process priors to learn the structure. The introduction of…
Model-free reinforcement learning has recently been shown to successfully learn navigation policies from raw sensor data. In this work, we address the problem of learning driving policies for an autonomous agent in a high-fidelity simulator. Building upon recent research that applies deep reinforcement learning to navi…
Method learns latent dynamics of complex systems from noisy data.
problem Challenging to construct ROMs from noisy high-dimensional data.
method Recurrent stochastic variational deep kernel learning (SVDKL).
result Framework accurately predicts system evolution in low-dimensional latent spaces.
Learning representations that disentangle the underlying factors of variability in data is an intuitive way to achieve generalization in deep models. In this work, we address the scenario where generative factors present a multimodal distribution due to the existence of class distinction in the data. We propose N-VAE, …
LTMs use latent vectors for efficient autoregressive generation.
problem Efficient autoregressive generation in language models.
method Dual-rate optimization in variational Bayes framework.
result LTMs achieve superior sample and parameter efficiency.
Hybrid model improves traffic flow prediction accuracy.
problem Predicting traffic flow with high accuracy in short-term future.
method A hybrid model combining hidden Markov model and LSTM.
result Significant performance gains over conventional methods.
A new method uses hyperspherical latent spaces to disentangle data with periodic structures.
problem Disentangling data with periodic or cyclic underlying factors in Euclidean space.
method Diffusion Variational Autoencoder with a modified Evidence Lower Bound.
result The method can recover periodic true factors effectively.
In this work we present Discrete Attend Infer Repeat (Discrete-AIR), a Recurrent Auto-Encoder with structured latent distributions containing discrete categorical distributions, continuous attribute distributions, and factorised spatial attention. While inspired by the original AIR model andretaining AIR model's capabi…
A standard Variational Autoencoder, with a Euclidean latent space, is structurally incapable of capturing topological properties of certain datasets. To remove topological obstructions, we introduce Diffusion Variational Autoencoders with arbitrary manifolds as a latent space. A Diffusion Variational Autoencoder uses t…
Improved exploration in RL with latent state marginalization.
problem Complexity of deep probabilistic models limits their practical use in reinforcement learning.
method Adopting latent variable policies within the MaxEnt framework, with low-cost marginalization of latent states.
result Effective marginalization leads to better exploration and more robust training.
Unified method learns latent spaces from labeled data.
problem Identifying low-dimensional latent structures in high-dimensional data.
method Nonlinear multiple-response regression within an index model context.
result Unified method offers better interpretability and reduced computational complexity.
New method disentangles shared and private latent factors in multimodal data.
problem Challenges in disentangling shared and private latent factors in multimodal data.
method Proposes a modification to existing multimodal Variational Autoencoders (MMVAE) to better handle modality-specific variation.
result Demonstrates improved robustness of modified MMVAE to modality-specific variation.
A novel variational autoencoder is developed to model images, as well as associated labels or captions. The Deep Generative Deconvolutional Network (DGDN) is used as a decoder of the latent image features, and a deep Convolutional Neural Network (CNN) is used as an image encoder; the CNN is used to approximate a distri…
New score-based methods identify causal structures with latent variables.
problem Identifying causal structures involving latent variables.
method Score-based methods with identifiability guarantees.
result Score equivalence and consistency for latent variable causal models.
Stable long-term predictions for fluid flows using neural networks.
problem Predicting complex dynamics of fluid flows with high temporal stability.
method End-to-end trained neural network architecture combining CNN for spatial compression and LSTM for temporal prediction.
result Novel latent space subdivision (LSS) allows stable and controllable long-term predictions.
p3VAE combines physics and machine learning for robust data representations.
problem Improving machine learning models' robustness to environmental factors of variation.
method Physics-informed variational autoencoder integrating physical knowledge with neural networks.
result p3VAE outperforms competing models in extrapolation and interpretability. Integrating deep learning with latent state space models has the potential to yield temporal models that are powerful, yet tractable and interpretable. Unfortunately, current models are not designed to handle missing data or multiple data modalities, which are both prevalent in real-world data. In this work, we introdu…
Prediction and interpolation for long-range video data involves the complex task of modeling motion trajectories for each visible object, occlusions and dis-occlusions, as well as appearance changes due to viewpoint and lighting. Optical flow based techniques generalize but are suitable only for short temporal ranges. …