This paper tackles belief-state selection in simulators with latent states.
problem Selecting among approximate belief-state samplers for simulators with latent variables.
method Reduces belief-state selection to conditional distribution selection, develops algorithms and analyses.
result Different formulations of belief-state selection have varying guarantees under different roll-out methods.
FORBES learns flexible belief states for POMDPs using normalizing flows.
problem Accurately modeling belief states in POMDPs for high-dimensional, continuous spaces.
method Integrates normalizing flows into variational inference for continuous belief state learning.
result FORBES learns flexible belief states that enable multi-modal predictions and high-quality reconstructions.
Task-oriented dialog presents a difficult challenge encompassing multiple problems including multi-turn language understanding and generation, knowledge retrieval and reasoning, and action prediction. Modern dialog systems typically begin by converting conversation history to a symbolic object referred to as belief sta…
Agent decides when to measure latent states in RL to improve efficiency.
problem Costly state measurement in RL negatively affects future outcomes.
method Introduces AOMDP with measurement action, uses online RL and sequential Monte Carlo.
result Reduced uncertainty improves sample efficiency and policy value.
Frequently, acquiring training data has an associated cost. We consider the situation where the learner may purchase data during training, subject TO a budget. IN particular, we examine the CASE WHERE each feature label has an associated cost, AND the total cost OF ALL feature labels acquired during training must NOT e…
New theorems show agents need specific internal structures to perform well under uncertainty.
problem How do agents need to be structured to perform well under uncertainty?
method Proved selection theorems showing strong task performance forces specific internal structures.
result Strong task performance forces world models, belief-like memory, and persistent regime-tracking variables.
When agents interact with a complex environment, they must form and maintain beliefs about the relevant aspects of that environment. We propose a way to efficiently train expressive generative models in complex environments. We show that a predictive algorithm with an expressive generative model can form stable belief-…
A new concept of confidence in learning is defined and analyzed.
problem Understanding and quantifying trust in learning processes.
method Formal axioms, continuum measures, vector fields, loss functions.
result Confidence can be represented and optimized in learning.
The study proves necessary conditions for robust decision-making in uncertain environments.
problem Conditions for robust decision-making in uncertain environments.
method Quantitative selection theorems and binary betting decisions.
result World models, belief-like memory, and persistent variables are necessary for strong task performance.
POLAR learns efficient data acquisition policies using pretrained belief representations.
problem Challenges in learning effective policies for adaptive data acquisition.
method POLAR decouples representation learning from policy learning by leveraging pretrained predictive foundation models as belief-state encoders.
result POLAR outperforms state-of-the-art methods across diverse tasks while requiring fewer training samples.
Unsupervised representation learning has succeeded with excellent results in many applications. It is an especially powerful tool to learn a good representation of environments with partial or noisy observations. In partially observable domains it is important for the representation to encode a belief state, a sufficie…
We study the restless bandit associated with an extremely simple scalar Kalman filter model in discrete time. Under certain assumptions, we prove that the problem is indexable in the sense that the Whittle index is a non-decreasing function of the relevant belief state. In spite of the long history of this problem, thi…
Latent-state environments with long horizons, such as those faced by recommender systems, pose significant challenges for reinforcement learning (RL). In this work, we identify and analyze several key hurdles for RL in such environments, including belief state error and small action advantage. We develop a general prin…
In statistical dialogue management, the dialogue manager learns a policy that maps a belief state to an action for the system to perform. Efficient exploration is key to successful policy optimisation. Current deep reinforcement learning methods are very promising but rely on epsilon-greedy exploration, thus subjecting…
A new model validation framework for agentic AI systems based on POMDPs.
problem Model validation of agentic AI systems.
method A POMDP-based framework for belief-state, forecast, and policy validation.
result The framework decomposes autonomous decision making into information, beliefs, forecasts, actions, and utility.
This paper provides an analysis of the tradeoff between asymptotic bias (suboptimality with unlimited data) and overfitting (additional suboptimality due to limited data) in the context of reinforcement learning with partial observability. Our theoretical analysis formally characterizes that while potentially increasin…
Benchmark evaluates financial misinformation detection models, revealing weaknesses without external context.
problem Detecting financial misinformation without external references.
method RFC Bench at paragraph level, two tasks: reference-free detection and comparison-based diagnosis.
result Performance improves with comparative context, revealing model weaknesses in reference-free settings.
Humans achieve efficient learning by relying on prior knowledge about the structure of naturally occurring tasks. There is considerable interest in designing reinforcement learning (RL) algorithms with similar properties. This includes proposals to learn the learning algorithm itself, an idea also known as meta learnin…
We present a new model, Predictive State Recurrent Neural Networks (PSRNNs), for filtering and prediction in dynamical systems. PSRNNs draw on insights from both Recurrent Neural Networks (RNNs) and Predictive State Representations (PSRs), and inherit advantages from both types of models. Like many successful RNN archi…
The paper tackles restless bandits with limited observation, proposing a method to analyze and approximate their optimal strategies.
problem Restless bandits with limited observation.
method General probabilistic model, PCL analysis, and approximation process.
result The proposed method can transform the problem into a finite-state problem, enabling the use of existing algorithms.
Study on natural actor-critic for POMDPs with finite memory.
problem Learning in partially observed Markov decision processes with noisy observations.
method Finite actor-critic method with multi-step temporal difference learning.
result First non-asymptotic global convergence for POMDPs with function approximation.
We solve POMDPs by approximating them as finite-state MDPs.
problem Computational challenges in learning optimal policies for POMDPs.
method Transform POMDP into a Superstate MDP, apply TD-learning and policy optimization.
result Finite-time bounds on TD-learning error for non-Markovian dynamics.
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.
New method infers human sensorimotor costs from behavior.
problem Inferring human sensorimotor costs from observed behavior.
method Inverse optimal control with signal-dependent noise.
result Recovering costs and benefits in sensorimotor behavior.
RL agents fail to generalize to unseen environments, even when dynamics are similar.
problem RL agents fail to generalize to unseen environments despite similar dynamics.
method Analyzed policy learning in POMDPs, formalized training dynamics as instances, and introduced a shared belief representation over an ensemble of specialized policies.
result Maximizing rewards induces instance-specific policies that are suboptimal on the training set.
Develops a framework for quantifying agentic AI model risk using LLM-inferred Bayesian state filters.
problem Quantifying the risk of agentic AI systems due to uncertain beliefs and actions.
method Representing the system as a partially observed Markov decision process with latent states, Bayesian belief updates, control-dependent losses, and tail-risk functionals.
result Develops a rigorous framework for separating uncertainty quantification from risk measurement.
Complex behaviors are often driven by an internal model, which integrates sensory information over time and facilitates long-term planning. Inferring an agent's internal model is a crucial ingredient in social interactions (theory of mind), for imitation learning, and for interpreting neural activities of behaving agen…
We introduce Recurrent Predictive State Policy (RPSP) networks, a recurrent architecture that brings insights from predictive state representations to reinforcement learning in partially observable environments. Predictive state policy networks consist of a recursive filter, which keeps track of a belief about the stat…
In the modern healthcare system, rapidly expanding costs/complexity, the growing myriad of treatment options, and exploding information streams that often do not effectively reach the front lines hinder the ability to choose optimal treatment decisions over time. The goal in this paper is to develop a general purpose (…
New method estimates selection event for black-box models.
problem Infeasibility of conditional selective inference due to unavailable selection event.
method Bootstrapping to estimate selection event and conditional distribution.
result Feasibility of conditional selective inference for models without exact selection event.
Cluster stability selection improves feature selection in correlated data.
problem Feature selection stability in correlated data.
method Cluster stability selection exploiting known cluster structure.
result Better predictive performance than lasso alone and stability selection.
New method corrects selection bias in post-selective inference for Group LASSO.
problem Inference after Group LASSO selection is unreliable.
method Develops a consistent, post-selective Bayesian method to adjust for selection bias.
result Corrects bias in recovering effects of selected variables.
Proposes PEMI for online selective conformal prediction with asymmetric rules.
problem Challenges of handling asymmetric selection mechanisms in online selective conformal prediction.
method PEMI: permutation-based framework for selective conformal prediction with arbitrary asymmetric selection rules.
result Achieves exact selection-conditional coverage for any asymmetric selection mechanism and any prediction model.
The paper addresses errors in online selective conformal prediction and proposes new strategies to ensure valid inference.
problem Online selective conformal prediction's exchangeability issues and false coverage rate control problems.
method Evaluation and correction of existing calibration selection strategies, proposing new ones that preserve exchangeability.
result Novel calibration selection strategies ensure both selection-conditional coverage and FCR control.
This work develops scalable model selection methods with fast update and selection.
problem Efficient model selection for large pools of candidate models.
method Isolated model embedding, which supports asymptotically fast update and selection.
result Standardized Embedder achieves competitive model selection performances.
Online method selects candidates from data streams, ensuring irreversible decisions.
problem Conformal selection's incompatibility with irreversible decisions in online scenarios.
method Online Conformal Selection with Accept-to-Reject Changes (OCS-ARC) incorporating online Benjamini-Hochberg procedure.
result OCS-ARC controls FDR at or below nominal level, improving selection power.
Paper improves k-NN predictive performance with efficient variable selection.
problem Improving predictive performance of k-NN models. method Efficient forward selection of predictor variables.
result Novel approach approaches outperformance of stepwise selection models.
Value selection reduces model size while maintaining accuracy.
problem Space efficiency in model size reduction.
method Two probabilistic methods based on information theory's metric: PVS and P + VS.
result Value selection achieves balance between accuracy and model size reduction.
Proposes HSIC-Lasso for selective inference in non-linear data.
problem Detecting influential features in non-linear and high-dimensional data.
method Model-free HSIC-Lasso based on truncated Gaussians and polyhedral lemma.
result Tight control of type-I error even for small sample sizes.
Paper proposes a statistical test for feature selection pipelines using selective inference.
problem Assessing the significance of feature selection pipelines in data analysis.
method Selective inference technique applied to feature selection pipelines composed of various algorithms.
result The proposed statistical test controls false positive feature selection probabilities.
New method improves feature selection by integrating stability paths.
problem Improving feature selection with tighter false positive control.
method Integrating stability paths to strengthen theoretical bounds on E(FP).
result Significantly more true positives with same E(FP) control.
UniFeat is an open-source Java tool for feature selection.
problem Efficient feature selection in various research areas.
method Provides a set of advanced feature selection methods.
result Facilitates rapid development of new feature selection algorithms.
We develop a general approach to valid inference after model selection. At the core of our framework is a result that characterizes the distribution of a post-selection estimator conditioned on the selection event. We specialize the approach to model selection by the lasso to form valid confidence intervals for the sel…
Flexible selective inference using flow-based transport maps.
problem Selective inference with complex selection events.
method Flow-based generative modeling for conditional distribution approximation.
result Valid p-values and confidence sets for adaptively selected hypotheses and parameters.
Recently, many regularized procedures have been proposed for variable selection in linear regression, but their performance depends on the tuning parameter selection. Here a criterion for the tuning parameter selection is proposed, which combines the strength of both stability selection and cross-validation and therefo…
Develops a more powerful selective inference method for stepwise feature selection.
problem Loss of power in existing conditional SI methods due to over-conditioning.
method Uses homotopy continuation approach to overcome over-conditioning.
result Shows improved power and efficiency in selective inference for feature selection.
Paper tackles selective regression using uncertainty estimation.
problem Selective regression for machine learning models to avoid predictions when uncertain.
method Model-agnostic non-parametric uncertainty estimation.
result Superior performance compared to state-of-the-art selective regressors.
Online selection of dynamic features has attracted intensive interest in recent years. However, existing online feature selection methods evaluate features individually and ignore the underlying structure of feature stream. For instance, in image analysis, features are generated in groups which represent color, texture…