Francesco Pappone

Francesco Pappone, Ruggero Marino Lazzaroni, Federico Califano et al.

While Large Language Models (LLMs) excel at generating human-like text, aligning their outputs with complex, qualitative goals like pedagogical soundness remains a significant challenge. Standard reinforcement learning techniques often rely on slow and expensive LLM-as-a-judge evaluations or on brittle, keyword-based metrics like ROUGE, which fail to capture the semantic essence of a high-quality explanation. In this work, we introduce a novel approach to reward shaping within the Group Relative Policy Optimisation (GRPO) framework. Our central contribution is the use of a small, efficient encoder-only transformer as a semantic reward model. This model provides a dense, semantically rich reward signal based on the cosine similarity between a generated explanation and a ground-truth reference, guiding the policy towards explanations that are not just factually correct but also structurally and conceptually aligned with expert reasoning. We apply this method to the task of training a model for the Italian medical-school entrance examinations, following standard domain-adaptive continued pre-training (CPT) and supervised fine-tuning (SFT). Our results demonstrate that GRPO with our proposed semantic reward significantly improves explanation faithfulness and clarity over a strong SFT baseline, showcasing the power of using lightweight encoder models for nuanced reward shaping in complex generation tasks

Francesco Pappone, Donato Crisostomi, Emanuele Rodolà

Recurrent-depth transformers scale test-time compute by iterating latent computations before emitting tokens. We study the geometry of these iterates and argue for a simple, two-scale operational picture: (i) within a looped block, updates act as small-scale refinements; (ii) across consecutive blocks, states undergo a larger-scale drift. Across training, our measurements show that loop steps become smaller and increasingly orthogonal to one another, indicating better local modeling of fine structure rather than merely pushing in a single direction. These dynamics motivate an early-exit mechanism based on the model's second-order difference in step-size, which we show is superior in terms of performance, stability and time-efficiency, when compared to the KL-divergence exit strategy of Geiping et al. and its naive first-order counterpart.

Francesco Pappone, Federico Califano, Marco Tafani

Accurately determining the geographic origin of mineral samples is pivotal for applications in geology, mineralogy, and material science. Leveraging the comprehensive Raman spectral data from the RRUFF database, this study introduces a novel machine learning framework aimed at geolocating mineral specimens at the country level. We employ a one-dimensional ConvNeXt1D neural network architecture to classify mineral spectra based solely on their spectral signatures. The processed dataset comprises over 32,900 mineral samples, predominantly natural, spanning 101 countries. Through five-fold cross-validation, the ConvNeXt1D model achieved an impressive average classification accuracy of 93%, demonstrating its efficacy in capturing geospatial patterns inherent in Raman spectra.