DRAFT v5.4 · iips.institute · Homepage · Card-grid reorganisation · Two neuro-symbolic methods · Co-Founding Director: F. Rinaudo (pending formal confirmation).
An independent research institute

Methods for the recovery of latent structure from observed data.

A collective of senior researchers working on inverse problems through two neuro-symbolic methods — classical scientific machine learning, and LLM-orchestrated RAG-mediated calibration of complex simulators.

Founded 2026 · Cambridge UK + Brussels BE · No physical premises · Distributed across Europe
IIPS — horse emerging from a polychromatic cloud of points
· · ·
I.

About — the inversion act, today.

The Institute reads the inverse problem as a precise operational act. In every instance of work, two functional roles are distinguished and brought into relation: a material to be inverted — the heterogeneous evidence on which the conversion operates — and a functional receiver — the structured model that accepts the inverted configuration and uses it. The two roles are not properties of the entities involved; they are states that materialise only at the moment of the inversion.

The inversion act — two functional roles distributed across Line A and Line B Diagram showing for each Line how the "material to be inverted" is converted into a configuration received by the "functional receiver". The grey zone represents the instance-specific nature of the role assignment. LINE A · FRONTIER MATERIAL TO BE INVERTED RAG + LLM corpus + orchestration inversion act parametric profiles FUNCTIONAL RECEIVER Physical / process model simulator that accepts the configuration re-activate for next instance LINE B · CONSOLIDATED MATERIAL TO BE INVERTED Data + system knowledge measurements + structural priors inversion act parameters / closures FUNCTIONAL RECEIVER Physical model ODE / PDE / state-space + neural closure re-activate for next instance The grey zone — role assignments materialise per-instance; nothing is permanent.

The two functional roles are not properties of the entities involved. They are states that the entities take on for the duration of a single inversion act. A new instance — a new lake, a new regulatory question, a new time interval — requires the roles to be re-activated.

Line B · consolidated SciML

Material: data + system knowledge. Receiver: physical model.

Measurements, structural priors, and accumulated knowledge of the system are converted into parameter values and closure conditions. The receiver is the deterministic model — ODE compartmental, PDE distributed-parameter, state-space — augmented where needed by PINNs, Universal Differential Equations, or graph neural networks.

Lineage: 1973 IFAC → 1976 Brebbia → 1987 Memorie 45 → 2026 Limnology.
Line A · frontier method

Material: RAG + LLM. Receiver: physical / process model.

A hybrid syntactic-vector RAG indexing a domain corpus, queried under LLM orchestration, is the material from which the inverted configuration is extracted. The receiver is the same category of mechanistic model as in Line B — a simulator that accepts parametric profiles, initial conditions, and boundary structures the RAG-LLM has assembled.

Prototype: TRAG, operational on the EMA corpus (~3,200 documents).

Read the full Founding Manifest →

II.

Work — two lines, six sub-areas.

Six sub-areas in total. Line A is articulated in a methodological foundation plus two applied projects on Italian sub-alpine lakes. Line B is articulated in three reformulations of foundational papers from the Institute's lineage.

Line A Frontier inversion through RAG-LLM

Foundation

TRAG

Targeted Retrieval-Augmented Generation — hybrid syntactic-vector RAG on the EMA EPAR + Product Information corpus. Mature operational prototype.

A·1 · Applied

Lake Maggiore Daphnia

RAG on Manca's 30+ years of zooplankton data generates the egg age-structures required by a compartmental ODE population model for parametric inversion.

Partner: Marina Manca (Founding Fellow).
A·2 · Applied

Lake Varese GLM-AED2

RAG on Lake Varese documentation generates the parametric profiles required to calibrate the General Lake Model + AED2 biogeochemistry module.

Partner: to be defined.

Line B Consolidated SciML neuro-symbolic

B·1 · Reformulation

Cladocera Lake Maggiore

McKendrick-Von Foerster age-transport + PINN-based birth-rate inversion + deep image classifiers on plankton microscopy.

Lineage: 1987 Memorie IIDr 45 → 2026 J. Limnology.
B·2 · Reformulation

Material accountancy

NUMSAS lineage (Brussels 1979) and Loss-Pattern Identification (INMM 1982 with LANL), reformulated as state-space neural inversion.

Target: ESARDA Bulletin, MBE Working Group.
B·3 · Reformulation

Radioecology Zn-65

Anatomically-informed graph neural network for compartmental tracer kinetics. Foundation: Merlini 1971 on Lepomis gibbosus.

Target: J. Environmental Radioactivity.

Project portfolio in detail →

III.

People — a nucleus with two co-Directors.

Founding Director

Flavio Argentesi

Cambridge UK. Coordinator of the JRC AI Laboratory (1980s); Head of Computational Nuclear Safeguards (1990s); Head of EMEA London start-up (1995-2003). Author of TRAG.

Co-Founding Director

Franco Rinaudo

Brussels BE. JRC AI Lab (1980s) on Case-Based Reasoning and Hypertext projects; first system engineer of the European Medicines Agency (1995-). Retired EU Commission officer. Pending formal acceptance.

Founding Fellow · confirmed

Marina Manca

Verbania IT. Formerly CNR-IRSA Verbania. Scientific lead CIPAIS 1981-2023. Operative on B·1 and A·1.

See the full nucleus →

IV.

Library — five decades of inverse-problem work.

Twenty-one entries from the founding nucleus, 1973-2026. Filterable by line, by recovery status, and by free-text. Each entry carries its original scanned PDF and, where appropriate, an English regenerated edition.

Browse the catalogue →