PolymathicIntro

Generative Solutions for Cosmic Problems

Flatiron Institute

Institute for Advanced Studies

Carol(ina) Cuesta-Lazaro

smsharma/awesome-neural-sbi

Astrophysics proliferates in Simulation-based Inference

on Simulations

Carolina Cuesta-Lazaro Flatiron/IAS

Scaling Cosmological Simulations

In Volume

In Parameter Space

Carolina Cuesta-Lazaro Flatiron/IAS

Cheap Gravity Only sims

p(\mathrm{Baryons}|\mathrm{Cheap})

(Test suite)

Astrid

EAGLE

\alpha = 0

\alpha = 0.25

\alpha = 0.5

\alpha = 0.75

\alpha = 1

Disentangling Physics and Instrument via Counterfactuals

Carolina Cuesta-Lazaro Flatiron/IAS

Physics

Systematics

[arXiv:2503.15312]

Object 1

Object 2

Object 1

z_\mathrm{instrument}

Orientation + Scale

Number

z_\mathrm{instrument},

z_\mathrm{object}

)

Instrument 1

Instrument 2

Instrument Encoder

z_\mathrm{object}

Object Encoder

Instrument Pair

Object Pair

Instrument Pair

Object Pair

Carolina Cuesta-Lazaro Flatiron/IAS

z_\mathrm{instrument}

Instrument Encoder

z_\mathrm{instrument},

z_\mathrm{object}

)

z_\mathrm{object}

Object Encoder

Orientation + Scale

Number

Instrument Pair(s)

Object Pair

Selected by proximity

Nearby approx similar systematics

Target

x^\mathcal{O}

x^\mathcal{S}

Simulated Data

Observed Data

z^\mathcal{O}_p

z^\mathcal{O}_s

z^\mathcal{S}_s

z^\mathcal{S}_p

Alignment Loss

\mathcal{L} = \sum_{\mathcal{D} \in (\mathcal{S}, \mathcal{O})} p(x^\mathcal{D}|z^\mathcal{D}_s, z^\mathcal{D}_p) + \lambda d(z^\mathcal{O}_s,z^\mathcal{S}_s)

Reconstruction

Statistical Alignment

50\%

(OT / Adversarial)

Encoder

Obs

Encoder

Sims

Private Domain Information

Shared Information

\hat{x}^\mathcal{O}

\hat{x}^\mathcal{S}

Observed Reconstructed

Simulated Reconstructed

Shared Decoder

Carolina Cuesta-Lazaro Flatiron/IAS

A Toy Model Example

Idealized Simulations

Observations

+ Scale Dependent Noise

+ Bump

x^\mathcal{O}

x^\mathcal{S}

Carolina Cuesta-Lazaro Flatiron/IAS

Amplitude

Tilt

p(\theta|z^\mathcal{O}_s)

p(\theta|z^\mathcal{O}_p)

p(\theta|z^\mathcal{O}_p,z^\mathcal{O}_s)

p(\theta|z^\mathcal{O}_p)

Robust SBI from Shared

p(x^\mathcal{O}|z^\mathcal{O}_p,z^\mathcal{O}_s)

p(x^\mathcal{O}|z^\mathcal{O}_s)

Visualizing Information Split

Carolina Cuesta-Lazaro Flatiron/IAS

Trained Gaussian - Tested Zero Column

\mathcal{L}(q_\mathrm{obs},q_\mathrm{latent},p_\mathrm{obs},p_\mathrm{latent})

1. Design next Experiment

2. Hypothesize Equation of motion

3. Simulate and Compare

p(\mathrm{World})

p(\mathrm{Prompt}|\mathrm{World})

Carolina Cuesta-Lazaro Flatiron/IAS

Learning to Play Scientist

Pavel Izmailov

["An LLM-driven framework for cosmological
model-building and exploration" Mudur, Cuesta-Lazaro, Toomey 
ML4PS NeurIPs Workshop 2025]

Pablo Mercader

Daniel Muthukrishna

Jeroen Audenaert

Legacy Survey

HSC

DESI

SDSS

Same Object / Different Instrument

Different Object / Same Instrument

Carolina Cuesta-Lazaro Flatiron/IAS

Ground Truth

Instrument Pair

Object Pair

Recon