Adrià Salvador Palau

#story Understanding neural networks through sparse circuits (OpenAI Blog, November 2025) Abstract: "Finding human-understandable circuits in language models is a central goal of the field of mechanistic interpretability. We train models to have more understandable circuits by constraining most of their weights to be zeros, so that each neuron only has a few connections. To recover fine-grained circuits underlying each of several hand-crafted tasks, we prune the models to isolate the part responsible for the task. These circuits often contain neurons and residual channels that correspond to natural concepts, with a small number of straightforwardly interpretable connections between them. We study how these models scale and find that making weights sparser trades off capability for interpretability, and scaling model size improves the capability-interpretability frontier. However, scaling sparse models beyond tens of millions of nonzero parameters while preserving interpretability remains a challenge. In addition to training weight-sparse models de novo, we show preliminary results suggesting our method can also be adapted to explain existing dense models. Our work produces circuits that achieve an unprecedented level of human understandability and validates them with considerable rigor." https://lnkd.in/gqWJyw_b Paper: Weight-sparse transformers have interpretable circuits https://lnkd.in/gcjjRMYw Article: Sparse Transformers to Expose Interpretable Circuits https://lnkd.in/gDmfwTch

2

We have a new paper on the arXiv that provides a widely applicable graphical language for photonically interlocking modular components for various kinds of quantum hardware. With Giovanni de Felice Boldizsár Poór Cole Comfort Lia Yeh Mateusz Kupper Will Cashman https://lnkd.in/ePxpJGsv

130

7 komentarzy

📢 Highly Cited in #Forecasting 📖 Searching for Promisingly Trained Artificial Neural Networks ✍️ Juan M. Lujano-Rojas, Rodolfo Dufo-López, Jesús Sergio Artal-Sevil, and Eduardo García-Paricio This study explores innovative approaches for training artificial neural networks, enhancing their performance in forecasting applications. A valuable reference for researchers and practitioners in AI, machine learning, and predictive analytics. 🔗 Read the full study: https://brnw.ch/21wYBTn #ArtificialIntelligence #MachineLearning #NeuralNetworks #PredictiveAnalytics

4

To accelerate convergence for strongly coupled scenarios, preCICE offers Quasi-Newton methods. But these methods disregard bound constraints arising from the physics of the problem. Jun C. works on modifying these methods using different modular methods. To know more about the methods, take a look at Jun's talk at last years workshop: https://lnkd.in/djQR4T8i

21

1 komentarz

McCulloch-Pitts networks with two nodes, two state values, excitatory/inhibitor/absent links, are simple enough that we can study all of them in a model/rule space: https://lnkd.in/exYz3s2b It is a seed project towards an understanding of "simplest complex systems". An accessible explanation of the paper (excellently done by the people at gist.science) can be found here: https://lnkd.in/exnwXhQa

2

1 komentarz

A new publication in Quantum: Variational quantum algorithms for permutation-based combinatorial problems: Optimal ansatz generation with applications to quadratic assignment problems and beyond by Dylan Laplace Mermoud, Andrea Simonetto, a... https://lnkd.in/egry32ex

47

2 komentarze

Research Paper Highlights: "Evaluating Large Language Models in Scientific Discovery", By Zhangde Song et al. Large language models are increasingly used in scientific research, but existing benchmarks fail to capture how real scientific discovery unfolds. Most evaluations focus on isolated factual questions, ignoring hypothesis formation, experimentation, and iterative reasoning. This work introduces a scenario-grounded Scientific Discovery Evaluation (SDE) framework to assess LLMs on realistic, expert-defined research workflows. Challenges: - Decontextualized evaluation: Traditional benchmarks overlook the project-level reasoning central to scientific discovery. - Misleading scaling signals: Gains from larger models and stronger reasoning do not translate proportionally to discovery tasks. - Hidden weaknesses: Existing benchmarks mask systematic reasoning gaps that emerge only in end-to-end research scenarios. Key Takeaways: - Scenario-grounded design: Expert-defined research projects are decomposed into modular scenarios across biology, chemistry, materials, and physics. - Two-level evaluation: Models are assessed on both question-level accuracy and project-level scientific reasoning, including hypothesis generation and interpretation. - Revealing insights: Top-tier LLMs show consistent performance gaps, large variability across scenarios, and no clear winner across discovery projects. This research emphasizes that true scientific intelligence cannot be measured by isolated QA performance alone. The SDE framework exposes fundamental limitations of current LLMs in discovery-driven reasoning, showing they remain far from scientific superintelligence. At the same time, it highlights LLMs’ emerging potential in guided exploration and provides a reproducible path forward for advancing AI-assisted scientific discovery. Further reading-https://lnkd.in/gYEpMaDB

5