The Herculaneum Papyri and the Vesuvius Challenge

The library came from the Villa dei Papiri, a large villa on the shore at Herculaneum that was tunnelled between 1750 and 1765⁠[1]. The rolls were taken to Naples, where most are still kept in the Officina dei Papiri of the Biblioteca Nazionale; a few were given away as diplomatic gifts and are now in Paris and Oxford⁠[5]⁠[9]. Each roll or fragment carries an inventory number prefixed PHerc. A complete roll was about 19 to 24 cm high and 4 to 6 cm across when carbonised⁠[2].

Four rolls carry the recent results. PHerc. Paris. 4, at the Institut de France, gave the 2023 Grand Prize text⁠[5]. PHerc. 172, at the Bodleian Libraries in Oxford, was scanned at Diamond Light Source in July 2024 and gave up its title in 2025⁠[9]. PHerc. 139 and PHerc. 1667 are in Naples⁠[2]. PHerc. 1667 is itself only the inner core of a roll: an attempt to open it in the nineteenth century, another by the conservator A. Fackelmann in November 1969 and a mechanical opening in the 1980s reduced it from 4.9 to 2 cm across and left only the lower 8 cm of its height⁠[2].

The opened rolls show that the core of the library was Epicurean philosophy, much of it by Philodemus of Gadara, who worked in Italy in the first century BC, together with older Epicurean texts and some works in Latin⁠[1]. Who owned the villa is not settled; a long-standing view links it to Lucius Calpurnius Piso, Philodemus's patron⁠[1].

The new results add specific texts. The 2023 Grand Prize passages in PHerc. Paris. 4 discuss pleasure and whether scarcity of goods such as food affects the pleasure they give; Richard Janko judged it very likely to be by Philodemus⁠[5]. PHerc. 172 carries the title Philodemus, On Vices, Book 1, read independently in 2025 by Sean Johnson of the Vesuvius Challenge and by Marcel Roth and Micha Nowak of the University of Würzburg⁠[9]. PHerc. 139 is Philodemus, On Gods, Book 8⁠[2]. PHerc. 1667 is the end of an ethical treatise. Its title is lost with the upper part of the roll, so the author cannot be named, but its vocabulary and the mention of Aristocreon, Chrysippus's nephew and pupil, point to a Stoic work of the second century BC, perhaps the very late third⁠[2].

X-ray phase-contrast tomography at the European Synchrotron Radiation Facility first showed letter-like signals inside a closed roll in 2015⁠[10]. A 2016 study found lead in the ink of some Herculaneum papyri, which gives X-ray contrast where carbon ink alone does not⁠[11]. The EduceLab-Scrolls dataset of 2023 trained ink-detection models on opened fragments, using infrared photographs of the exposed surface as ground truth, and reported moderate character recall (about 42 per cent) with about 18 per cent false positives when checked by a papyrologist⁠[4].

The 2026 paper tests three things on three rolls⁠[2]. For PHerc. 1667 it claims complete coverage of the preserved writing surface, about 860 square centimetres in 22 columns, with only 33 square centimetres in the first three columns left untranscribed. For PHerc. Paris. 4 a scan at 2.4 µm makes ink deposits 10 to 20 µm thick directly visible in the volume, and the segmented ink matches the 2023 machine-learning reading, which is an independent check that the models were reading real ink. For PHerc. 139 it recovers a title and author line. The ink models were trained iteratively on labelled columns of the same roll while a held-out column was used for validation⁠[2].

Three questions remain open. The first is whether the method scales. PHerc. 1667 was a small, already reduced roll; PHerc. 172 was about 70 per cent unwrapped at the time of the 2026 paper; and the 2027 Grand Prize asks for one complete scroll from a set of 13 eligible rolls, which shows that a full intact roll has not yet been read⁠[2]⁠[12]. Heavily compressed or damaged rolls may defeat surface tracing even where ink is visible.

The second is what the sealed rolls contain. Every title recovered so far from a closed roll is by Philodemus except PHerc. 1667, which looks Stoic⁠[9]⁠[2]. One reading is that the sealed rolls will mirror the opened ones: mostly Philodemus and the Epicurean school. Another is that they hold a wider philosophical collection, including works by rival schools that survive nowhere else. A third, argued from the villa's size and its unexcavated levels, is that a larger library, including Latin literature, lies unexcavated⁠[1]. The third depends on excavation as much as on imaging.

The third question is attribution. With the title of PHerc. 1667 lost, its author can be argued from content and style only; press accounts mention Chrysippus as a working hypothesis, which the paper itself does not claim⁠[13]⁠[2].

A complete, intact roll unwrapped from outside to core, with its title, would show that the pipeline scales; the 2027 Grand Prize is designed to produce one. Reading titles across many sealed rolls would answer what the library contains far faster than full transcriptions, because Herculaneum rolls usually repeat their title at the end. Attribution of PHerc. 1667 would be settled by a parallel quotation in another source or by a stylistic comparison against the surviving fragments of Chrysippus and other Stoics. Whether more of the library survives underground can only be settled by excavating the villa's lower levels.