Related papers: Indirect jumps improve instruction sequence perfor…

In continuous control, exploration is often performed through undirected strategies in which parameters of the networks or selected actions are perturbed by random noise. Although the deep setting of undirected exploration has been shown to…

We consider a modulated process S which, conditional on a background process X, has independent increments. Assuming that S drifts to -infinity and that its increments (jumps) are heavy-tailed (in a sense made precise in the paper), we…

Contribution: This study examined student effort and performance in an introductory programming course with respect to student-held implicit theories and self-efficacy. Background: Implicit theories and self-efficacy shed a light into…

We present a formal system for proving the partial correctness of a single-pass instruction sequence as considered in program algebra by decomposition into proofs of the partial correctness of segments of the single-pass instruction…

Large language models' behavior is often shaped by instructions such as system prompts, refusal boundaries, privacy constraints, and tool-use rules that must hold at inference time. Yet in practice these constraints can be violated under…

We study asymptotic behaviours of a non-linear vertex-reinforced jump process defined on an arbitrary infinite graph with bounded degree. We prove that if the reinforcement function $w$ is reciprocally integrable and non-decreasing, then…

In this paper, we establish a new inequality tying together the effective length and the maximum correlation between the outputs of an arbitrary pair of Boolean functions which operate on two sequences of correlated random variables. We…

Large language models (LLMs) could be valuable personal AI agents across various domains, provided they can precisely follow user instructions. However, recent studies have shown significant limitations in LLMs' instruction-following…

For many types of learning, spaced training that involves repeated long inter-trial intervals (ITIs) leads to more robust memory formation than does massed training that involves short or no intervals. Several cognitive theories have been…

In-context learning (ICL) enables large language models to perform new tasks by conditioning on a sequence of examples. Most prior work reasonably and intuitively assumes that which examples are chosen has a far greater effect on…

Finding the optimal hyperparameters of a model can be cast as a bilevel optimization problem, typically solved using zero-order techniques. In this work we study first-order methods when the inner optimization problem is convex but…

Directed paths have been used by several authors to describe concurrent executions of a program. Spaces of directed paths in an appropriate state space contain executions with all possible legal schedulings. It is interesting to investigate…

In this paper, we consider a class of stochastic impulse control problem when there is a fixed delay $\Delta$ between the decision and execution times. The dynamics of the controlled system between two impulses is an arbitrary adapted…

Humans can learn several tasks in succession with minimal mutual interference but perform more poorly when trained on multiple tasks at once. The opposite is true for standard deep neural networks. Here, we propose novel computational…

We propose an optimization proxy in terms of iterative implicit gradient methods for solving constrained optimization problems with nonconvex loss functions. This framework can be applied to a broad range of machine learning settings,…

What kinds of instructional prompts are easier to follow for Language Models (LMs)? We study this question by conducting extensive empirical analysis that shed light on important features of successful instructional prompts. Specifically,…

We study the problem of sample efficient reinforcement learning, where prior data such as demonstrations are provided for initialization in lieu of a dense reward signal. A natural approach is to incorporate an imitation learning objective,…

Reward hacking arises when a model improves a proxy reward by exploiting shortcuts rather than solving the intended task. We study this failure mode through the geometry of reinforcement learning updates in language models and argue that…

Designing effective task-level prompts is crucial for improving the performance of Large Language Models (LLMs). While prior work on instruction induction demonstrates that LLMs can infer better instructions with limited examples, existing…

The influence of class orderings in the evaluation of incremental learning has received very little attention. In this paper, we investigate the impact of class orderings for incrementally learned classifiers. We propose a method to compute…