arXiv subjects

Davide Romano

Publications and source records attributed to Davide Romano.

9 recordsLinked to original sources

arXiv2026

Thomson: Continual Learning of Frontier Models for SovereignAI

The development of frontier models is commonly perceived to be the exclusive remit of a small number of heavily funded players, creating an information, economic and power asymmetry between developers and the diverse user base of modern AI. Recent public discourse acknowledges this concern, calling for SovereignAI (an organisation's capability to independently build, deploy and govern AI use), but offers little concrete advice on how this can be achieved in the short term under a diversity of funding settings. We argue that frontier performance is achievable by a wide range of institutions through Continual Learning on readily available open-weight models. Unlike limited approaches such as small-scale fine-tuning, prompt engineering, or tool-augmentation of a frozen model, our approach exploits a modern mid- & post-training stack while introducing safeguards that preserve both plasticity and stability at each stage, making the minimal number of high-impact interventions on the parameters. This yields gains comparable to those typically seen across multiple successive model generations, at compute and personnel budgets substantially lower than commonly thought, making ownership of large parts of the SovereignAI stack (model, tool infrastructure, values & data privacy) viable for far more actors. We demonstrate this with Thomson, a general-purpose frontier model trained with an enhanced focus on high-stakes professional work. Thomson performs competitively with recent frontier models across agentic tasks, safety, legal, tax & multilingualism, and large-scale Deep Research. Evaluations show a distinctive $\pi$-shaped pattern: distinct improvements across a wide range of capabilities, including those not explicitly targeted, while almost completely eliminating the forgetting problem common to narrow domain adaptation.

cs.AI↗

arXiv2026

Test-Time Scaling in the Wild: Why Exploitation, Not Exploration, Is the Bottleneck

Test-time scaling (TTS) improves language model outputs by spending additional inference compute - generating multiple candidates, searching over partial sequences, or iteratively refining drafts. These techniques yield large gains on mathematics and code, but have been developed and stress-tested almost exclusively on tasks where verification is straightforward. We conduct the first compute-normalised comparison of five TTS families across five open-ended generation benchmarks spanning medicine, law, finance, general chat, and creative writing - grounded in a unified framework that decomposes the effectiveness of each method's token budget into exploration and exploitation. The answer depends on which side of that decomposition you examine. Scaling exploration works: the best candidate in the pool improves steadily with compute across all settings. What breaks is exploitation - the step that converts a rich candidate pool into a final output. With state-of-the-art generators, reward models correlate at only $\rho_v \approx 0.12$ with true quality, rendering selection near-random regardless of budget. Tree search amplifies this failure through diversity collapse. Refinement helps on one of five benchmarks; its apparent gains elsewhere are confounded. Only synthesis across candidates (Fusion) consistently improves over single-sample baselines, yet still recovers only ~40% of available quality. The candidate pool is not the bottleneck - choosing from it is.

cs.CL↗

arXiv2025

Evaluating the Role of Verifiers in Test-Time Scaling for Legal Reasoning Tasks

Test-time scaling (TTS) techniques can improve the performance of large language models (LLMs) at the expense of additional computation and latency. While TTS has proven effective in formal domains such as mathematics and programming, its value in argumentative domains such as law remains underexplored. We present an empirical study of verifier-based TTS methods for legal multiple-choice QA (MCQA) across five benchmarks. Using a family of 7 reward models, we evaluate both outcome-level (Best-of-$N$) and process-level (tree search) verification under realistic low-$N$ budgets. Our analysis systematically investigates how verifier utility is affected by key properties such as domain specialization, model size, and supervision type (process-supervised PRMs vs. outcome-only ORMs), even when applied across different roles.

cs.CL↗

arXiv2025

Multi-field as a determinable

The paper advances the hypothesis that the multi-field is a determinable, that is, a physical object characterized by indeterminate values with respect to some properties. The multi-field is a realist interpretation of the wave function in quantum mechanics, specifically it interprets the wave function as a new physical entity in 3D space: a multi-field (Hubert & Romano 2018; Romano 2021). The multi-field is similar to a field as it assigns determinate values to N-tuples of points, but is also different from a field as it does not assign pre-existing values at each point of 3D space. In particular, the multi-field values corresponding to the empty points (points where no particles are located) have indeterminate values until a particle is located at those points. The paper suggests that the multi-field so defined can be precisely characterized in terms of determinable-based, object-level, account of metaphysical indeterminacy. Under this view, the multi-field as novel physical entity is, in fact, a metaphysically indeterminate quantum object, that is, a determinable.

physics.hist-ph↗

arXiv2024

iLLuMinaTE: An LLM-XAI Framework Leveraging Social Science Explanation Theories Towards Actionable Student Performance Feedback

Recent advances in eXplainable AI (XAI) for education have highlighted a critical challenge: ensuring that explanations for state-of-the-art AI models are understandable for non-technical users such as educators and students. In response, we introduce iLLuMinaTE, a zero-shot, chain-of-prompts LLM-XAI pipeline inspired by Miller's cognitive model of explanation. iLLuMinaTE is designed to deliver theory-driven, actionable feedback to students in online courses. iLLuMinaTE navigates three main stages - causal connection, explanation selection, and explanation presentation - with variations drawing from eight social science theories (e.g. Abnormal Conditions, Pearl's Model of Explanation, Necessity and Robustness Selection, Contrastive Explanation). We extensively evaluate 21,915 natural language explanations of iLLuMinaTE extracted from three LLMs (GPT-4o, Gemma2-9B, Llama3-70B), with three different underlying XAI methods (LIME, Counterfactuals, MC-LIME), across students from three diverse online courses. Our evaluation involves analyses of explanation alignment to the social science theory, understandability of the explanation, and a real-world user preference study with 114 university students containing a novel actionability simulation. We find that students prefer iLLuMinaTE explanations over traditional explainers 89.52% of the time. Our work provides a robust, ready-to-use framework for effectively communicating hybrid XAI-driven insights in education, with significant generalization potential for other human-centric fields.

cs.CY↗

arXiv2023

A decoherence-based approach to the classical limit in Bohm's theory

The paper explains why the de Broglie-Bohm theory reduces to Newtonian mechanics in the macroscopic classical limit. The quantum-to-classical transition is based on three steps: (i) interaction with the environment produces effectively factorized states, leading to the formation of effective wave functions and hence decoherence; (ii) the effective wave functions selected by the environment--the pointer states of decoherence theory--will be well-localized wave packets, typically Gaussian states; (iii) the quantum potential of a Gaussian state becomes negligible under standard classicality conditions; therefore, the effective wave function will move according to Newtonian mechanics in the correct classical limit. As a result, a Bohmian system in interaction with the environment will be described by an effective Gaussian state and--when the system is macroscopic--it will move according to Newtonian mechanics.

physics.hist-ph↗

arXiv2021

The metaphysics of decoherence

The paper investigates the type of realism that best suits the framework of decoherence taken at face value without postulating a plurality of worlds, or additional hidden variables, or non-unitary dynamical mechanisms. It is argued that this reading of decoherence leads to an extremely radical type of perspectival realism, especially when cosmological decoherence is considered.

quant-ph↗

arXiv2018

The Wave-Function as a Multi-Field

It is generally argued that if the wave-function in the de Broglie--Bohm theory is a physical field, it must be a field in configuration space. Nevertheless, it is possible to interpret the wave-function as a multi-field in three-dimensional space. This approach hasn't received the attention yet it really deserves. The aim of this paper is threefold: first, we show that the wave-function is naturally and straightforwardly construed as a multi-field; second, we show why this interpretation is superior to other interpretations discussed in the literature; third, we clarify common misconceptions.

physics.hist-ph↗

arXiv2016

Bohmian Classical Limit in Bounded Regions

Bohmian mechanics is a realistic interpretation of quantum theory. It shares the same ontology of classical mechanics: particles following continuous trajectories in space through time. For this ontological continuity, it seems to be a good candidate for recovering the classical limit of quantum theory. Indeed, in a Bohmian framework, the issue of the classical limit reduces to showing how classical trajectories can emerge from Bohmian ones, under specific classicality assumptions. In this paper, we shall focus on a technical problem that arises from the dynamics of a Bohmian system in bounded regions; and we suggest that a possible solution is supplied by the action of environmental decoherence. However, we shall show that, in order to implement decoherence in a Bohmian framework, a stronger condition is required (disjointness of supports) rather than the usual one (orthogonality of states).

quant-ph↗