Unlearning as Ablation: Toward a Falsifiable Benchmark for Generative Scientific Discovery

TL;DR

This paper proposes a systematic unlearning-as-ablation method to test whether large language models genuinely generate new scientific knowledge or simply remix memorized information, aiming to improve AI-for-Science evaluation.

Contribution

It introduces a conceptual and methodological framework for using ablation as a falsifiable probe to distinguish genuine knowledge generation from recall in AI models.

Findings

Proposes unlearning-as-ablation as a scientific discovery probe

Outlines a pilot in mathematics and algorithms

Suggests extension to physics and chemistry domains

Abstract

Bold claims about AI's role in science-from "AGI will cure all diseases" to promises of radically accelerated discovery-raise a central epistemic question: do large language models (LLMs) truly generate new knowledge, or do they merely remix memorized fragments? We propose unlearning-as-ablation as a falsifiable probe of constructive scientific discovery. The idea is to systematically remove a target result together with its forget-closure (supporting lemmas, paraphrases, and multi-hop entailments) and then evaluate whether the model can re-derive the result from only permitted axioms and tools. Success would indicate generative capability beyond recall; failure would expose current limits. Unlike prevailing motivations for unlearning-privacy, copyright, or safety-our framing repositions it as an epistemic probe for AI-for-Science. We outline a minimal pilot in mathematics and algorithms to illustrate feasibility, and sketch how the same approach could later be extended to domains such as physics or chemistry. This is a position paper: our contribution is conceptual and methodological, not empirical. We aim to stimulate discussion on how principled ablation tests could help distinguish models that reconstruct knowledge from those that merely retrieve it, and how such probes might guide the next generation of AI-for-Science benchmarks.