A critical flaw of existing inverse reinforcement learning (IRL) methods is their inability to significantly outperform the demonstrator. This is because IRL typically seeks a reward function that makes the demonstrator appear near-optimal, rather than inferring the underlying intentions of the demonstrator that may ha…
New method optimizes language model performance for test-time strategies.
problem Mismatch between training objectives and test-time deployment of large language models.
method Tail-Extrapolated estimators to approximate best-of-N performance from limited training rollouts.
result Improved performance of best-of-N deployment across various models and datasets.
The performance of imitation learning is typically upper-bounded by the performance of the demonstrator. While recent empirical results demonstrate that ranked demonstrations allow for better-than-demonstrator performance, preferences over demonstrations may be difficult to obtain, and little is known theoretically abo…
BRAID fine-tunes diffusion models to optimize reward models in offline scenarios.
problem Combining generative modeling and model-based optimization in offline scenarios.
method Conservative fine-tuning of diffusion models using RL to optimize reward models.
result BRAID outperforms existing methods in offline data, avoiding invalid designs.
Automated algorithm selection and hyperparameter tuning facilitates the application of machine learning. Traditional multi-armed bandit strategies look to the history of observed rewards to identify the most promising arms for optimizing expected total reward in the long run. When considering limited time budgets and c…
Kernel method estimates long-term effects from short-term data.
problem Estimating long-term effects from short-term data in continuous actions.
method Kernel ridge regression to embed and extrapolate long-term effects.
result Uniform consistency and nonasymptotic error bounds for the estimator.
Bayesian reward learning from demonstrations enables rigorous safety and uncertainty analysis when performing imitation learning. However, Bayesian reward learning methods are typically computationally intractable for complex control problems. We propose Bayesian Reward Extrapolation (Bayesian REX), a highly efficient …
Study validates metrics for offline MBO using diffusion models.
problem Evaluate metrics for offline MBO without ground truth oracle.
method Propose and quantify validation metrics over datasets.
result Identify most effective validation metrics.
Contextual bandit methods fail with deficient support data.
problem Learning from support-deficient data in contextual bandits.
method Three approaches to IPS-based learning: action space restriction, reward extrapolation, and policy space restriction.
result Systematic analysis and empirical evaluation of approaches to IPS-based learning.
Bayesian inverse reinforcement learning (IRL) methods are ideal for safe imitation learning, as they allow a learning agent to reason about reward uncertainty and the safety of a learned policy. However, Bayesian IRL is computationally intractable for high-dimensional problems because each sample from the posterior req…
New algorithm uses machine learning to predict rewards for decision-making problems.
problem Sequential decision-making under uncertainty with scarce online data.
method Machine Learning-Assisted Upper Confidence Bound (MLA-UCB) algorithm.
result Proves to improve cumulative regret even with biased surrogate rewards.
RTE enables extrapolation to new tasks by learning task transformations.
problem Learning systems struggle to generalize to unseen tasks.
method Relational Task Extrapolator (RTE) learns task transformations to enable extrapolation.
result RTE substantially outperforms existing approaches on extrapolation tasks.
We extend nonparametric models to handle extrapolation, providing bounds for inference.
problem Challenges in nonparametric statistical inference when evaluating outside the conditioning variable's support.
method Introduced a class of extrapolation assumptions and a consistent estimation procedure to handle extrapolation.
result Validated extrapolation-aware conclusions through various applications and real-world data.
Binary encoding enables neural networks to extrapolate periodic functions.
problem Extrapolating periodic functions without prior knowledge of their form.
method Normalized Base-2 Encoding (NB2E) for continuous numerical values.
result MLPs using NB2E can successfully extrapolate diverse periodic signals.
Method controls extrapolation in prediction profiles for statistical and machine learning models.
problem Avoiding invalid predictions due to extrapolation in prediction profiles.
method Genetic algorithm optimization over constrained factor regions.
result Optimal factor settings without constraint are often invalid and extrapolated.
Neural networks struggle with extrapolation, but a new framework allows them to learn counterfactual invariances.
problem Neural networks' inability to extrapolate beyond training data distribution.
method Introduces a learning framework that allows neural networks to extrapolate over group transformations based on counterfactual invariances.
result Neural networks can learn counterfactual invariances from a single environment, overcoming their limitations in extrapolation.
Study shows LLMs can extrapolate rules from out-of-distribution prompts.
problem Understanding LLMs' ability to generalize from unexpected inputs.
method Formal languages and rule-based scenarios to evaluate LLMs' OOD behavior.
result LLMs can extrapolate rules from out-of-distribution prompts, even in complex scenarios.
New methods reduce extrapolation errors in feature importance.
problem Flawed feature importance methods using unrestricted permutations lead to extrapolation errors.
method Three new approaches: conditional model reliance, Knockoffs with Gaussian transformation, and restricted ALE plot designs.
result Theoretical and numerical results show our strategies reduce/eliminate extrapolation.
New RL method handles diverse human operator data.
problem Handling data from multiple, unknown policies.
method Domain-Invariant Model-based Offline RL (DIMORL) with Risk Extrapolation (REx).
result Models trained with REx generalize better across different demonstrators.
New models extrapolate false alarms in ASV without new data.
problem Reliable extrapolation of false alarm rates in ASV without new speaker data.
method Generative models in ASV score space for arbitrary systems.
result Models accurately extrapolate false alarm rates for large speaker populations.
Neural networks extrapolate poorly in simple tasks but succeed in complex ones.
problem Understanding neural networks' extrapolation capabilities and conditions for success.
method Analyzing ReLU MLPs and GNNs, connecting to neural tangent kernel.
result ReLU MLPs learn linear functions but not most nonlinear ones, while GNNs succeed in complex tasks due to task-specific non-linearities.
A new principle for extrapolating regression outside training data.
problem Regression extrapolation when predictions are outside training data range.
method Data-adaptive marginal transformation and simple relationship assumption.
result Progression method offers guarantees on approximation error beyond training data range.
We extend return extrapolation to incorporate asymmetry and saturation, finding that asymmetric nonlinear extrapolation leads to lower welfare loss.
problem Optimal portfolio choice under stochastic volatility
method Smooth, nonlinear extrapolation function with sentiment and variance hedging
result Lower welfare loss with asymmetric nonlinear extrapolation
Logical neural networks solve mazes by filling dead ends, but not all methods generalize well.
problem Understanding how logical neural networks extrapolate solutions to mazes.
method Examined recurrent and implicit neural networks trained on maze-solving tasks.
result Models fail to generalize well to diverse maze sizes, suggesting limitations in learning scalable algorithms.
Accelerates Riemannian gradient methods with extrapolation.
problem Optimizing functions on manifolds efficiently.
method Extrapolating iterates in Riemannian gradient descent.
result Achieves optimal convergence rate and computational advantage.
Improves reinforcement learning extrapolation in Gridworlds.
problem Generalizing to unseen states in reinforcement learning.
method Avoiding deterministic action choice, using ego-centric representation, incorporating symmetry, and adding an entropy term.
result Significant improvement in extrapolation performance in Gridworld environments.
New methods improve neural network extrapolation to long sequences.
problem Neural networks struggle with extrapolation to very long or adversarial sequences.
method Activation binning and localized differentiable memory architecture.
result No extrapolation errors detected within memory constraints.
Concept modulation models unify identifiability and extrapolation in conditional latent variable models.
problem Reliable generalization in conditional latent variable models
method Concept modulation models (CMMs) with structure AoΛoCoX result Lifts identifiability to conditional settings and controls extrapolation through attribute potentials.
New framework shows ERM is optimal for both interpolation and extrapolation in domain generalization.
problem Formalizing and solving the challenges of domain generalization.
method Reformulated domain generalization as an online game between a risk-minimizing player and an adversary.
result ERM is minimax-optimal for both interpolation and extrapolation in domain generalization.
NSR enables neural networks to reason with continuous numbers and extrapolate.
problem Quantitative reasoning and extrapolation in neural networks.
method Proposes Neural Status Register (NSR) for continuous number reasoning.
result NSR achieves extrapolation to numbers many orders of magnitude larger than training data.
Extends return extrapolation to nonlinear, asymmetric functions under stochastic volatility.
problem Behavioral anomalies in portfolio choice under stochastic volatility.
method Smooth, nonlinear, asymmetric extrapolation function; CRRA investor; Heston stochastic volatility; Hamilton-Jacobi-Bellman equation; Numerical solutions (finite-difference ADI, deep learning-driven iterative).
result Saturation acts as an endogenous correction mechanism, reducing welfare loss.
This paper explores conditions for neural networks to extrapolate to new domains.
problem Understanding when neural networks can extrapolate to unseen domains.
method Analyzes conditions for nonlinear models to extrapolate under specific distribution shifts.
result Neural networks of the form f(x)=∑fi(xi) can extrapolate if feature covariance is well-conditioned. The paper explores how to extrapolate from limited data points using causal mechanisms.
problem Handling distribution shifts with limited target samples.
method Formulates the extrapolation problem with a latent-variable model embodying the minimal change principle in causal mechanisms, and identifies conditions for identification.
result Theoretical understanding and practical methods for extrapolation without requiring an on-support target distribution.
Diverging Flows detects extrapolations in flow models, ensuring reliable predictions.
problem Flow models extrapolate into invalid data, leading to silent failures.
method Structurally enforce inefficient transport for off-manifold inputs.
result Effective detection of extrapolations without compromising predictive fidelity or inference latency.
Proposes a new method to avoid model extrapolation in Shapley values.
problem Model extrapolation in marginal Shapley values leads to unreliable explanations.
method Proposes a new approach that avoids model extrapolation using marginal averaging and causal information.
result Demonstrates the impacts of model extrapolation on Shapley values and proposes a new method to avoid it.
Paper tackles intervention extrapolation using identifiable representations.
problem Predicting effects of unseen interventions on outcomes.
method Combines identifiable representation learning with autoencoders to enforce linear invariance.
result Identifiable representations enable non-linear extrapolation of interventions.
Paper improves PINNs' extrapolation by TL and adaptive AFs.
problem PINNs' poor extrapolation performance and sensitivity to AFs.
method Transfer learning within an extended domain and adaptive activation functions.
result Average 40% reduction in relative L2 error and 50% in mean absolute error in extrapolation domain.
We present a framework on how to hedge the interest rate sensitivity of liabilities discounted by an extrapolated yield curve. The framework is based on functional analysis in that we consider the extrapolated yield curve as a functional of an observed yield curve and use its Gâteaux variation to understand the sensiti…
The paper tackles extrapolation in generative models by enforcing independence of mechanisms.
problem How to make generative models extrapolate to new, unseen environments?
method Developed a theoretical framework for independence of mechanisms, demonstrated on toy examples and real-world data.
result Extrapolation capabilities of generative models can be improved by enforcing independence of mechanisms explicitly during training.
Methodology extrapolates wind fields from sparse data with uncertainty quantification.
problem Extrapolating wind fields from limited measurements with uncertainty.
method Nonparametric Bayesian dictionary learning for sparse/incomplete data.
result Enhanced extrapolation accuracy, even in high-dimensional data.
BiPE blends intra-segment and inter-segment encodings for better length extrapolation.
problem Improving length extrapolation in language models.
method Bilevel Positional Encoding (BiPE) that separates intra-segment and inter-segment encodings.
result BiPE enhances length extrapolation across various text modalities.
Neural networks are surprisingly good at interpolating and perform remarkably well when the training set examples resemble those in the test set. However, they are often unable to extrapolate patterns beyond the seen data, even when the abstractions required for such patterns are simple. In this paper, we first review …
STMT predicts compounds in unknown areas with trend reflection.
problem Predicting properties of unknown compounds in the extrapolation area.
method Stochastic Threshold Model Trees (STMT) that reflects data trends.
result STMT improves prediction accuracy for one compound in the extrapolation area.
A new method improves language model extrapolation without changing training.
problem Improving language model performance on sequences not seen during training.
method Perturbation-based training procedure for language models.
result The method consistently improves out-of-support prediction.
We win EVA2025 by estimating extreme precipitation events using Peaks Over Thresholds and martingale testing.
problem Estimating the probability of extreme precipitation events with limited data.
method Modeling Peaks Over Thresholds with an exponential distribution and using martingale testing for evaluation.
result Our method outperforms other approaches in estimating extreme precipitation events.
We propose probabilistic models that can extrapolate learning curves of iterative machine learning algorithms, such as stochastic gradient descent for training deep networks, based on training data with variable-length learning curves. We study instantiations of this framework based on random forests and Bayesian recur…
Extreme value theory enhances statistical learning extrapolation for rare events.
problem Challenges in traditional machine learning methods for extreme data.
method Asymptotic theory and statistical tools for tail behavior.
result Effective extrapolation methods for extreme quantiles and anomalies.
Accelerates coordinate descent methods for machine learning problems.
problem Slowness of coordinate descent methods in machine learning.
method Extrapolation-based accelerated coordinate descent.
result Significant speed-up in practice compared to existing methods.