We analyze the capabilities of Transformer language models in learning compositional discrete tasks. To this end, we evaluate training LLaMA models and prompting GPT-4 and Gemini on four tasks demanding to learn a composition of several discrete sub-tasks. In particular, we measure how well these models can reuse primitives observable in the sub-tasks to learn the composition task.

STOC 2023] obtained generic algorithms that work for various learning tasks. However, their focus was on the *example-rich* regime, where the users have so many examples that each user could themselves solve the problem. In this work we consider the *example-scarce* regime, where each user has only a few examples, and obtain the following results:* For approximate-DP, we give a generic transformation of any item-level DP algorithm to a user-level DP algorithm. Roughly speaking, the latter gives a (multiplicative) savings of $O_{\varepsilon,\delta}(\sqrt{m})$ in terms of the number of users required for achieving the same utility, where $m$ is the number of examples per user. This algorithm, while recovering most known bounds for specific problems, also gives new bounds, e.g., for PAC learning.

Learning controllers with offline data in decision-making systems is an essential area of research due to its potential to reduce the risk of applications in real-world systems. However, in responsibility-sensitive settings such as healthcare, decision accountability is of paramount importance, yet has not been adequately addressed by the literature.This paper introduces the Accountable Offline Controller (AOC) that employs the offline dataset as the Decision Corpus and performs accountable control based on a tailored selection of examples, referred to as the Corpus Subset. AOC operates effectively in low-data scenarios, can be extended to the strictly offline imitation setting, and displays qualities of both conservation and adaptability.We assess AOC's performance in both simulated and real-world healthcare scenarios, emphasizing its capability to manage offline control tasks with high levels of performance while maintaining accountability.

Out-of-distribution (OOD) testing is increasingly popular for evaluating a machine learning system's ability to generalize beyond the biases of a training set. OOD benchmarks are designed to present a different joint distribution of data and labels between training and test time. VQA-CP has become the standard OOD benchmark for visual question answering, but we discovered three troubling practices in its current use. First, most published methods rely on explicit knowledge of the construction of the OOD splits. They often rely on yes'' when the common training answer was ``no''.

Accurate proton dose calculation using Monte Carlo (MC) is computationally demanding in workflows like robust optimisation, adaptive replanning, and probabilistic inference, which require repeated evaluations. To address this, we develop a neural surrogate that integrates Monte Carlo dropout to provide fast, differentiable dose predictions along with voxelwise predictive uncertainty. The method is validated through a series of experiments, starting with a one-dimensional analytic benchmark that establishes accuracy, convergence, and variance decomposition. Two-dimensional bone-water phantoms, generated using TOPAS Geant4, demonstrate the method's behavior under domain heterogeneity and beam uncertainty, while a three-dimensional water phantom confirms scalability for volumetric dose prediction. Across these settings, we separate epistemic (model) from parametric (input) contributions, showing that epistemic variance increases under distribution shift, while parametric variance dominates at material boundaries. The approach achieves significant speedups over MC while retaining uncertainty information, making it suitable for integration into robust planning, adaptive workflows, and uncertainty-aware optimisation in proton therapy.

As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization---the capacity to combine learned skills in novel ways not encountered during training---has garnered significant attention. This type of generalization, particularly in scenarios beyond training data, is also of great interest in the study of AI safety and alignment. A recent study introduced the Skill-Mix evaluation, where models are tasked with composing a short paragraph demonstrating the use of a specified k -tuple of language skills. While small models struggled with composing even with k 3, larger models like GPT-4 performed reasonably well with k 5 and 6 .In this paper, we employ a setup akin to Skill-Mix to evaluate the capacity of smaller models to learn compositional generalization from examples. Utilizing a diverse set of language skills---including rhetorical, literary, reasoning, theory of mind, and common sense---GPT was used to generate text samples that exhibit random subsets of k skills.

Objective: The aim of this study was to build an effective co-reference resolution system tailored for the biomedical domain. Materials and Methods: Experiment materials used in this study is provided by the 2011 i2b2 Natural Language Processing Challenge. The 2011 i2b2 challenge involves coreference resolution in medical documents. Concept mentions have been annotated in clinical texts, and the mentions that co-refer in each document are to be linked by coreference chains. Normally, there are two ways of constructing a system to automatically discover co-referent links. One is to manually build rules for co-reference resolution, and the other category of approaches is to use machine learning systems to learn automatically from training datasets and then perform the resolution task on testing datasets. Results: Experiments show the existing co-reference resolution systems are able to find some of the co-referent links, and our rule based system performs well finding the majority of the co-referent links. Our system achieved 89.6% overall performance on multiple medical datasets. Conclusion: The experiment results show that manually crafted rules based on observation of training data is a valid way to accomplish high performance in this coreference resolution task for the critical biomedical domain.

Text-to-Image (T2I) models are capable of generating high-quality artistic creations and visual content. However, existing research and evaluation standards predominantly focus on image realism and shallow text-image alignment, lacking a comprehensive assessment of complex semantic understanding and world knowledge integration in text to image generation. To address this challenge, we propose $\textbf{WISE}$, the first benchmark specifically designed for $\textbf{W}$orld Knowledge-$\textbf{I}$nformed $\textbf{S}$emantic $\textbf{E}$valuation. WISE moves beyond simple word-pixel mapping by challenging models with 1000 meticulously crafted prompts across 25 sub-domains in cultural common sense, spatio-temporal reasoning, and natural science. To overcome the limitations of traditional CLIP metric, we introduce $\textbf{WiScore}$, a novel quantitative metric for assessing knowledge-image alignment. Through comprehensive testing of 20 models (10 dedicated T2I models and 10 unified multimodal models) using 1,000 structured prompts spanning 25 subdomains, our findings reveal significant limitations in their ability to effectively integrate and apply world knowledge during image generation, highlighting critical pathways for enhancing knowledge incorporation and application in next-generation T2I models. Code and data are available at https://github.com/PKU-YuanGroup/WISE.

Large Language Models (LLMs) have demonstrated remarkable instruction-following capabilities across various applications. However, their performance in multilingual settings remains poorly understood, as existing evaluations lack fine-grained constraint analysis. We introduce XIFBench, a comprehensive constraint-based benchmark for assessing multilingual instruction-following abilities of LLMs, featuring a novel taxonomy of five constraint categories and 465 parallel instructions across six languages spanning different resource levels. To ensure consistent cross-lingual evaluation, we develop a requirement-based protocol that leverages English requirements as semantic anchors. These requirements are then used to validate the translations across languages. Extensive experiments with various LLMs reveal notable variations in instruction-following performance across resource levels, identifying key influencing factors such as constraint categories, instruction complexity, and cultural specificity.