Our approach requires no additional training and enables highly flexible module composition.

Many industries now deploy high-fidelity simulators (digital twins) to represent physical systems, yet their parameters must be calibrated to match the true system. This motivated the construction of simulation-driven parameter estimators, built by generating synthetic observations for sampled parameter values and learning a supervised mapping from observations to parameters. However, when the true parameters lie outside the sampled range, predictions suffer from an out-of-distribution (OOD) error. This paper introduces a fine-tuning approach for the Two-Stage estimator that mitigates OOD effects and improves accuracy. The effectiveness of the proposed method is verified through numerical simulations.

Generalization is a key challenge in machine learning, specifically in reasoning tasks, where models are expected to solve problems more complex than those encountered during training. Existing approaches typically train reasoning models in an end-to-end fashion, directly mapping input instances to solutions. While this allows models to learn useful heuristics from data, it often results in limited generalization beyond the training distribution. In this work, we propose a novel approach to reasoning generalization by learning energy landscapes over the solution spaces of smaller, more tractable subproblems. At test time, we construct a global energy landscape for a given problem by combining the energy functions of multiple subproblems. This compositional approach enables the incorporation of additional constraints during inference, allowing the construction of energy landscapes for problems of increasing difficulty. To improve the sample quality from this newly constructed energy landscape, we introduce Parallel Energy Minimization (PEM). We evaluate our approach on a wide set of reasoning problems. Our method outperforms existing state-of-the-art methods, demonstrating its ability to generalize to larger and more complex problems. Project website can be found at: https://alexoarga.github.io/compositional_reasoning/

Our approach requires no additional training and enables highly flexible module composition.

Large language model (LLM) unlearning has demonstrated effectiveness in removing the influence of undesirable data (also known as forget data). Existing approaches typically assume full access to the forget dataset, overlooking two key challenges: (1) Forget data is often privacy-sensitive, rare, or legally regulated, making it expensive or impractical to obtain (2) The distribution of available forget data may not align with how that information is represented within the model. To address these limitations, we propose a ``Reveal-and-Release'' method to unlearn with self-generated data, where we prompt the model to reveal what it knows using optimized instructions. To fully utilize the self-generated forget data, we propose an iterative unlearning framework, where we make incremental adjustments to the model's weight space with parameter-efficient modules trained on the forget data. Experimental results demonstrate that our method balances the tradeoff between forget quality and utility preservation.

Parameter-Efficient Fine-Tuning (PEFT) is an efficient alternative to full scale fine-tuning, gaining popularity recently. With pre-trained model sizes growing exponentially, PEFT can be effectively utilized to fine-tune compact modules, Parameter-Efficient Modules (PEMs), trained to be domain experts over diverse domains. In this project, we explore composing such individually fine-tuned PEMs for distribution generalization over the composite domain. To compose PEMs, simple composing functions are used that operate purely on the weight space of the individually fine-tuned PEMs, without requiring any additional fine-tuning. The proposed method is applied to the task of representing the 16 Myers-Briggs Type Indicator (MBTI) composite personalities via 4 building block dichotomies, comprising of 8 individual traits which can be merged (composed) to yield a unique personality. We evaluate the individual trait PEMs and the composed personality PEMs via an online MBTI personality quiz questionnaire, validating the efficacy of PEFT to fine-tune PEMs and merging PEMs without further fine-tuning for domain composition.

Retrieval-Augmented Generation (RAG) systems for Large Language Models (LLMs) hold promise in knowledge-intensive tasks but face limitations in complex multi-step reasoning. While recent methods have integrated RAG with chain-of-thought reasoning or test-time search using Process Reward Models (PRMs), these approaches encounter challenges such as a lack of explanations, bias in PRM training data, early-step bias in PRM scores, and insufficient post-training optimization of reasoning potential. To address these issues, we propose Retrieval-Augmented Reasoning through Trustworthy Process Rewarding (ReARTeR), a framework that enhances RAG systems' reasoning capabilities through post-training and test-time scaling. At test time, ReARTeR introduces Trustworthy Process Rewarding via a Process Reward Model for accurate scalar scoring and a Process Explanation Model (PEM) for generating natural language explanations, enabling step refinement. During post-training, it utilizes Monte Carlo Tree Search guided by Trustworthy Process Rewarding to collect high-quality step-level preference data, optimized through Iterative Preference Optimization. ReARTeR addresses three core challenges: (1) misalignment between PRM and PEM, tackled through off-policy preference learning; (2) bias in PRM training data, mitigated by balanced annotation methods and stronger annotations for challenging examples; and (3) early-step bias in PRM, resolved through a temporal-difference-based look-ahead search strategy. Experimental results on multi-step reasoning benchmarks demonstrate significant improvements, underscoring ReARTeR's potential to advance the reasoning capabilities of RAG systems.

To handle the complexities of real-world traffic, learning planners for self-driving from data is a promising direction. While recent approaches have shown great progress, they typically assume a setting in which the ground-truth world state is available as input. However, when deployed, planning needs to be robust to the long-tail of errors incurred by a noisy perception system, which is often neglected in evaluation. To address this, previous work has proposed drawing adversarial samples from a perception error model (PEM) mimicking the noise characteristics of a target object detector. However, these methods use simple PEMs that fail to accurately capture all failure modes of detection. In this paper, we present EMPERROR, a novel transformer-based generative PEM, apply it to stress-test an imitation learning (IL)-based planner and show that it imitates modern detectors more faithfully than previous work. Furthermore, it is able to produce realistic noisy inputs that increase the planner's collision rate by up to 85%, demonstrating its utility as a valuable tool for a more complete evaluation of self-driving planners.

Recent transformer-based architectures have shown impressive results in the field of image segmentation. Thanks to their flexibility, they obtain outstanding performance in multiple segmentation tasks, such as semantic and panoptic, under a single unified framework. To achieve such impressive performance, these architectures employ intensive operations and require substantial computational resources, which are often not available, especially on edge devices. To fill this gap, we propose Prototype-based Efficient MaskFormer (PEM), an efficient transformer-based architecture that can operate in multiple segmentation tasks. PEM proposes a novel prototype-based cross-attention which leverages the redundancy of visual features to restrict the computation and improve the efficiency without harming the performance. In addition, PEM introduces an efficient multi-scale feature pyramid network, capable of extracting features that have high semantic content in an efficient way, thanks to the combination of deformable convolutions and context-based self-modulation. We benchmark the proposed PEM architecture on two tasks, semantic and panoptic segmentation, evaluated on two different datasets, Cityscapes and ADE20K. PEM demonstrates outstanding performance on every task and dataset, outperforming task-specific architectures while being comparable and even better than computationally-expensive baselines.

As an efficient alternative to conventional full finetuning, parameter-efficient finetuning (PEFT) is becoming the prevailing method to adapt pretrained language models. In PEFT, a lightweight module is learned on each dataset while the underlying pretrained language model remains unchanged, resulting in multiple compact modules representing diverse skills when applied to various domains and tasks. In this paper, we propose to compose these parameter-efficient modules through linear arithmetic operations in the weight space, thereby integrating different module capabilities. Specifically, we first define addition and negation operators for the module, and then further compose these two basic operators to perform flexible arithmetic. Our approach requires \emph{no additional training} and enables highly flexible module composition. We apply different arithmetic operations to compose the parameter-efficient modules for (1) distribution generalization, (2) multi-tasking, (3) unlearning, and (4) domain transfer. Additionally, we extend our approach to detoxify Alpaca-LoRA, the latest instruction-tuned large language model based on LLaMA. Empirical results demonstrate that our approach produces new and effective parameter-efficient modules that significantly outperform existing ones across all settings.