Despite the success of reinforcement learning from human feedback (RLHF) in aligning language models with human values, reward hacking, also termed reward overoptimization, remains a critical challenge. This issue primarily arises from reward misgeneralization, where reward models (RMs) compute reward using spurious features that are irrelevant to human preferences. In this work, we tackle this problem from an information-theoretic perspective and propose a framework for reward modeling, namely InfoRM, by introducing a variational information bottleneck objective to filter out irrelevant information.Notably, we further identify a correlation between overoptimization and outliers in the IB latent space of InfoRM, establishing it as a promising tool for detecting reward overoptimization.Inspired by this finding, we propose the Cluster Separation Index (CSI), which quantifies deviations in the IB latent space, as an indicator of reward overoptimization to facilitate the development of online mitigation strategies. Extensive experiments on a wide range of settings and RM scales (70M, 440M, 1.4B, and 7B) demonstrate the effectiveness of InfoRM. Further analyses reveal that InfoRM's overoptimization detection mechanism is not only effective but also robust across a broad range of datasets, signifying a notable advancement in the field of RLHF.

Also known as the phase retrieval problem, it is a fundamental challenge in nano-, bio-and astronomical imaging systems, and speech processing. The problem is ill-posed, and therefore additional assumptions on the signal and/or the measurements are necessary.

Estimation of functions of $ d $ variables is considered using ridge combinations of the form $ \textstyle\sum_{k=1}^m c_{1,k} \phi(\textstyle\sum_{j=1}^d c_{0,j,k}x_j-b_k) $ where the activation function $ \phi $ is a function with bounded value and derivative. These include single-hidden layer neural networks, polynomials, and sinusoidal models. From a sample of size $ n $ of possibly noisy values at random sites $ X \in B = [-1,1]^d $, the minimax mean square error is examined for functions in the closure of the $ \ell_1 $ hull of ridge functions with activation $ \phi $. It is shown to be of order $ d/n $ to a fractional power (when $ d $ is of smaller order than $ n $), and to be of order $ (\log d)/n $ to a fractional power (when $ d $ is of larger order than $ n $). Dependence on constraints $ v_0 $ and $ v_1 $ on the $ \ell_1 $ norms of inner parameter $ c_0 $ and outer parameter $ c_1 $, respectively, is also examined. Also, lower and upper bounds on the fractional power are given. The heart of the analysis is development of information-theoretic packing numbers for these classes of functions.

Doctor John Doe dictating") are removed, directives to transcribers Dictated medical reports pose many challenges to automatic (e.g., "thank you for writing this", "I'm sorry, remove speech recognition (ASR) due to the linguistic complexity this") are skipped, and explanatory phrases (e.g., "first of the domain, the acoustic environment, and the way name Kevin K E V for Victor I N") and repeated words (e.g., speakers handle dictation. First, highly complex, domainspecific "she uh she") are handled. Moreover, auto-formatting corrects medical terminology including thousands of drug partial phrases in context (e.g., "temp" to "temperature"); names render the use of standard language models ineffective.

We consider a distributed parameter estimation problem, in which multiple terminals send messages related to their local observations using limited rates to a fusion center who will obtain an estimate of a parameter related to observations of all terminals. It is well known that if the transmission rates are in the Slepian-Wolf region, the fusion center can fully recover all observations and hence can construct an estimator having the same performance as that of the centralized case. One natural question is whether Slepian-Wolf rates are necessary to achieve the same estimation performance as that of the centralized case. In this paper, we show that the answer to this question is negative. We establish our result by explicitly constructing an asymptotically minimum variance unbiased estimator (MVUE) that has the same performance as that of the optimal estimator in the centralized case while requiring information rates less than the conditions required in the Slepian-Wolf rate region.

A plan for a question required the composition of REQUEST and INFORM and led to the development of two new kinds of informing speech acts, INFORMREF To plan a yes/no question about some proposition P. one should think that the and INFORMIF, and their mediating acts. The INFORMREF acts lead to hearer knows whether P is true or false (or, at least "might know"). An approximate "what," "when," and "where" questions while INFORMIF results in a yes/no representation of AGT2's knowing whether P is true or false is OR (AGT2 question.2' The reason for these new acts is that, in planning a REQUEST that BELIEVE P, AGT2 BELIEVE -- P)).'9 Such goals are often created, as modelled someone else perform an INFORM act, one only has incomplete knowledge of by our type 4 inference, when a planner does not know the truth-value of P. their beliefs and goals; but an INFORM, as originally defined can only be Typical circumstances in which an agent may acquire such disjunctive beliefs planned when one knows what is to be said.

We consider the estimation of an i.i.d.\ vector $\xbf \in \R^n$ from measurements $\ybf \in \R^m$ obtained by a general cascade model consisting of a known linear transform followed by a probabilistic componentwise (possibly nonlinear) measurement channel. We present a method, called adaptive generalized approximate message passing (Adaptive GAMP), that enables joint learning of the statistics of the prior and measurement channel along with estimation of the unknown vector $\xbf$. The proposed algorithm is a generalization of a recently-developed method by Vila and Schniter that uses expectation-maximization (EM) iterations where the posteriors in the E-steps are computed via approximate message passing. The techniques can be applied to a large class of learning problems including the learning of sparse priors in compressed sensing or identification of linear-nonlinear cascade models in dynamical systems and neural spiking processes. We prove that for large i.i.d.\ Gaussian transform matrices the asymptotic componentwise behavior of the adaptive GAMP algorithm is predicted by a simple set of scalar state evolution equations. This analysis shows that the adaptive GAMP method can yield asymptotically consistent parameter estimates, which implies that the algorithm achieves a reconstruction quality equivalent to the oracle algorithm that knows the correct parameter values. The adaptive GAMP methodology thus provides a systematic, general and computationally efficient method applicable to a large range of complex linear-nonlinear models with provable guarantees.

A complex system is made up of many components with many interactions. So the design of systems such as simulation systems, cooperative systems or assistance systems includes a very accurate modelling of interactional and communicational levels. The agent-based approach provides an adapted abstraction level for this problem. After having studied the organizational context and communicative capacities of agentbased systems, to simulate the reorganization of a flexible manufacturing, to regulate an urban transport system, and to simulate an epidemic detection system, our thoughts on the interactional level were inspired by human-machine interface models, especially those in "cognitive engineering". To provide a general framework for agent-based complex systems modelling, we then proposed a scale of four behaviours that agents may adopt in their complex systems (reactive, routine, cognitive, and collective). To complete the description of multi-level agent models, which is the focus of this paper, we illustrate our modelling and discuss our ongoing work on each level.

Rough set theory, a mathematical tool to deal with inexact or uncertain knowledge in information systems, has originally described the indiscernibility of elements by equivalence relations. Covering rough sets are a natural extension of classical rough sets by relaxing the partitions arising from equivalence relations to coverings. Recently, some topological concepts such as neighborhood have been applied to covering rough sets. In this paper, we further investigate the covering rough sets based on neighborhoods by approximation operations. We show that the upper approximation based on neighborhoods can be defined equivalently without using neighborhoods. To analyze the coverings themselves, we introduce unary and composition operations on coverings. A notion of homomorphismis provided to relate two covering approximation spaces. We also examine the properties of approximations preserved by the operations and homomorphisms, respectively.

In Pawlak's rough set theory, a set is approximated by a pair of lower and upper approximations. To measure numerically the roughness of an approximation, Pawlak introduced a quantitative measure of roughness by using the ratio of the cardinalities of the lower and upper approximations. Although the roughness measure is effective, it has the drawback of not being strictly monotonic with respect to the standard ordering on partitions. Recently, some improvements have been made by taking into account the granularity of partitions. In this paper, we approach the roughness measure in an axiomatic way. After axiomatically defining roughness measure and partition measure, we provide a unified construction of roughness measure, called strong Pawlak roughness measure, and then explore the properties of this measure. We show that the improved roughness measures in the literature are special instances of our strong Pawlak roughness measure and introduce three more strong Pawlak roughness measures as well. The advantage of our axiomatic approach is that some properties of a roughness measure follow immediately as soon as the measure satisfies the relevant axiomatic definition.