@imperialcollegeldn.bsky.social @imperialeee.bsky.social

Excited to host Prof. @iaugenstein.bsky.social at the
@imperial-ix.bsky.social & Digital Foundry Seminar Series!

Title: Understanding The Interplay Between LLMs' Utilisation of Parametric and Contextual
Knowledge

📅 30 April, 14:00–15:00
📍 White City & MS Teams

Register via poster!

@imperialeee.bsky.social @imperial-ix.bsky.social

📣 PhD opening – Fall 2026

The DUCK Lab @imperialcollegeldn.bsky.social is looking for a PhD student to join us!

Why 🦆? We work on foundational aspects of #neurosymbolicAI and #SafeAI.
👉 DUCK = Data, Uncertainty, Constraints & Knowledge

📩 Apply by emailing: e.giunchiglia@imperial.ac.uk

I will presenting this work at #EurIPS tomorrow!

📍Come to poster #30!

@emilevankrieken.com @nolovedeeplearning.bsky.social @ema-ridopoco.bsky.social @lucderaedt.bsky.social @luislamb.bsky.social @arturgarcez.bsky.social @imperialcollegeldn.bsky.social @imperial-ix.bsky.social @imperialeee.bsky.social

GitHub - Giotto-maker/r4rr: Official code for the paper 'Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototype-Augmented Neurosymbolic AI' Official code for the paper 'Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototype-Augmented Neurosymbolic AI' - Giotto-maker/r4rr

Link to paper: arxiv.org/pdf/2510.25497
Link to code: github.com/Giotto-maker...

In the paper, we also show that with just 1 labelled datapoint per concept, our prototypical NeSy model achieves the same number of symbolic optima as full dense supervision.

👉 Prototypes cut annotation cost without losing expressivity.

With just 2 labeled points, we can anchor each class to a single prototype—so the embedding updates reflect both similarity to the few labels and constraint satisfaction.

This helps in making the NeSy models #robust wrt the number of unsupervised datapoints!

🔍 Watch how the decision boundaries evolve!

This GIF shows PCA projections of the Semantic Loss model’s logits during training on MNIST-EvenOdd.

Colors = model predictions, points = true digit concepts

Above: standard semantic loss
Below: prototypical semantic loss

Want to use your favourite #NeSy model but afraid of the reasoning shortcuts?🫣

Fear not💪🏻In our #NeurIPS2025 paper we show that you just need to equip your favourite NeSy model with prototypical networks and the reasoning shortcuts will be a problem of the past!

Research Associate positions in Computational Chemistry, AI in Chemistry, and in Neurosymbolic Learning for Chemistry at Imperial College London Apply now for the Research Associate positions in Computational Chemistry, AI in Chemistry, and in Neurosymbolic Learning for Chemistry role on jobs.ac.uk - the leading job board for higher education ...

🤝 Collaborate with us to push the boundaries of neurosymbolic AI in scientific discovery.

🔗Link to the position: jobs.ac.uk/job/DPB378

#NeuroSymbolicAI #AI4Science #ChemAI #Postdoc #Hiring #AcademicJobs

🚀 We're hiring!

@imperialcollegeldn.bsky.social is recruiting a Postdoc in Neurosymbolic AI for Chemistry in the amazing Prof. Kim Jelfs’ group!🎯

Work at the intersection of symbolic reasoning, ML, and molecular design — incl. logic-guided generative models & GNNs. ⚛️🤖

Organising NeSy 2025 together with some of my favourite people (@e-giunchiglia.bsky.social @lgilpin.bsky.social @pascalhitzler.bsky.social) really was a dream. Super proud of what we've achieved! See you next year 😘

If you're around London (or online!) tomorrow, don’t miss this — @nolovedeeplearning.bsky.social will be talking about #neurosymbolic #AI, #reliability, and #reasoning at @imperialcollegeldn.bsky.social and @imperial-ix.bsky.social ⚡️

Come along, tune in, or ping him for a coffee ☕👇

@emilevankrieken.com @pascalhitzler.bsky.social @lgilpin.bsky.social

⏳⏳⏳ One more day to the NeSy abstract submission! Don't miss it!

Full timeline at: 2025.nesyconf.org/call-for-pap...

🚨🚨 Last I-X Seminar of the Year!

Don’t miss Dr Antonio Vergari (@nolovedeeplearning.bsky.social) on Logically-Consistent Deep Learning — a crucial topic at the intersection of logic & neural networks! 🌟

@andreapasserini.bsky.social @emilevankrieken.com @neuralnoise.com @pascalhitzler.bsky.social @arturgarcez.bsky.social @imperial-ix.bsky.social
@imperialeee.bsky.social

If you made it this far, you're probably really into constraint-aware models 😄

Don’t miss the new paper by @nolovedeeplearning.bsky.social et al!
👉https://arxiv.org/abs/2503.19466

It proposes a probabilistic layer ensuring satisfaction of algebraic constraints over continuous variables—cool stuff!

🗓️ 𝐂𝐮𝐫𝐢𝐨𝐮𝐬? Come chat with us at 𝐇𝐚𝐥𝐥 𝟑 + 𝐇𝐚𝐥𝐥 𝟐𝐁, 𝐏𝐨𝐬𝐭𝐞𝐫 #𝟏𝟕𝟐 on 𝐀𝐩𝐫𝐢𝐥 𝟐𝟔𝐭𝐡!

Link to paper: arxiv.org/pdf/2502.18237

🔍
We applied this to tabular data generation, showing our models not only satisfy the constraints, but also produce more realistic datapoints and 𝐨𝐮𝐭𝐩𝐞𝐫𝐟𝐨𝐫𝐦 𝐮𝐧𝐜𝐨𝐧𝐬𝐭𝐫𝐚𝐢𝐧𝐞𝐝 𝐛𝐚𝐬𝐞𝐥𝐢𝐧𝐞𝐬.

🚨New at #ICLR: we introduce the first ever 𝐥𝐚𝐲𝐞𝐫 that makes 𝐚𝐧𝐲 neural network 𝐜𝐨𝐦𝐩𝐥𝐢𝐚𝐧𝐭 𝐛𝐲 𝐝𝐞𝐬𝐢𝐠𝐧 with constraints expressed as 𝐝𝐢𝐬𝐣𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐬 𝐨𝐟 𝐥𝐢𝐧𝐞𝐚𝐫 𝐢𝐧𝐞𝐪𝐮𝐚𝐥𝐢𝐭𝐢𝐞𝐬—even if they define 𝐧𝐨𝐧-𝐜𝐨𝐧𝐯𝐞𝐱 𝐬𝐩𝐚𝐜𝐞𝐬!

🌟absolutely unmissable event! 🌟

Don’t forget to register!!

Join us for our next I-X seminar with Professor Tim Rocktäschel (UCL & Google DeepMind) titled "Open-Endedness and General Intelligence".

🕓 13:30 - 14:30 (GMT)
📅 Thursday, 3 April
📍 Hybrid (White City Campus / MS Teams)

🔗Register now: ix.imperial.ac.uk/event/i-x-se...

@e-giunchiglia.bsky.social

🙌 Don’t forget to register!!!

🙌🏻 NeSy Confence = Amazing Speakers!!! 🙌🏻

⏳Do not forget to submit your abstract before the deadline!

📩 Interested? Apply now: e.giunchiglia@imperial.ac.uk

📌 Use [PhD Application] in the email subject

@imperial-ix.bsky.social @imperialeee.bsky.social @imperialcollegeldn.bsky.social @nesyconf.org

[3/3]

📢 What’s included?

✅ Full scholarship for UK/EU citizens with (pre-)settled status
✅ Industry collaboration: 3-month internship with our partner company
✅ Be part of a world-class research environment 🌍

[2/3]