Publications

Rapid De Novo Antibody Design with GeoFlow-V3
BioGeometry Team
Jian Tang
Recent years have witnessed striking advances in miniprotein design, yet de novo antibody discovery remains challenging, marked by low bindi… (voir plus)ng rates and the need for extensive, labor-intensive experimental screening of millions of candidates. This technical report introduces GeoFlow-V3, a unified atomic generative model for structure prediction and protein design. GeoFlow-V3 delivers improved accuracy on antibody-antigen complex structure prediction relative to our previous version, and its performance is further enhanced when experimental constraints or prior knowledge are provided, enabling precise control over both folding and design. The model also demonstrates reliable ability to discriminate binders from non-binders based on its confidence scores. Leveraging this capability, we build a GeoFlow-V3 in silico pipeline to design no more than 50 nanobodies per therapeutically relevant target de novo, completing a single round of wet-lab characterization in under three weeks. GeoFlow-V3 identifies at least one binder for 8 tested epitopes and achieves an average hit rate of 15.5%, representing a two-orders-of-magnitude improvement over prior computational pipelines. These results position GeoFlow-V3 as an appealing platform for rapid, AI-driven therapeutic antibody discovery, significantly reducing experimental screening demands and offering a powerful avenue to tackle previously undruggable targets. A demo of GeoFlow-V3 can be accessed via prot.design for non-commercial use.
2025-10-20
bioRxiv (prépublication)
doi.org
Representing Positional Information in Generative World Models for Object Manipulation
Stefano Ferraro
Pietro Mazzaglia
Tim Verbelen
Bart Dhoedt
Sai Rajeswar
Object manipulation capabilities are essential skills that set apart embodied agents engaging with the world, especially in the realm of rob… (voir plus)otics. The ability to predict outcomes of interactions with objects is paramount in this setting. While model-based control methods have started to be employed for tackling manipulation tasks, they have faced challenges in accurately manipulating objects. As we analyze the causes of this limitation, we identify the cause of underperformance in the way current world models represent crucial positional information, especially about the target's goal specification for object positioning tasks. We introduce a general approach that empowers world model-based agents to effectively solve object-positioning tasks. We propose two declinations of this approach for generative world models: position-conditioned (PCP) and latent-conditioned (LCP) policy learning. In particular, LCP employs object-centric latent representations that explicitly capture object positional information for goal specification. This naturally leads to the emergence of multimodal capabilities, enabling the specification of goals through spatial coordinates or a visual goal. Our methods are rigorously evaluated across several manipulation environments, showing favorable performance compared to current model-based control approaches.
2025-10-20
Frontiers in Artificial Intelligence and Applications (publié)
doi.org
arxiv.org
Improved Localized Machine Unlearning Through the Lens of Memorization
Reihaneh Torkzadehmahani
Reza Nasirigerdeh
Georgios Kaissis
Daniel Rueckert
Gintare Karolina Dziugaite
Eleni Triantafillou
Machine unlearning refers to removing the influence of a specified subset of training data from a machine learning model, efficiently, after… (voir plus) it has already been trained. This is important for key applications, including making the model more accurate by removing outdated, mislabeled, or poisoned data. In this work, we study localized unlearning, where the unlearning algorithm operates on a (small) identified subset of parameters. Drawing inspiration from the memorization literature, we propose an improved localization strategy that yields strong results when paired with existing unlearning algorithms. We also propose a new unlearning algorithm, Deletion by Example Localization (DEL), that resets the parameters deemed-to-be most critical according to our localization strategy, and then finetunes them. Our extensive experiments on different datasets, forget sets and metrics reveal that DEL sets a new state-of-the-art for unlearning metrics, against both localized and full-parameter methods, while modifying a small subset of parameters, and outperforms the state-of-the-art localized unlearning in terms of test accuracy too.
2025-10-19
TMLR (accepté)
doi.org
openreview.net
Toward the Decarbonization of Maritime Supply Chains: A Ship Emissions Prediction Framework
Abdelhak El Aissi
Ismail Bourzak
Loubna Benabbou
Abdelaziz Berrado
Maritime transport is a vital component of international trade, yet the industry contributes substantially to greenhouse gas (GHG) emissions… (voir plus), with carbon dioxide
2025-10-19
2025 International Conference on Intelligent Systems: Theories and Applications (SITA) (publié)
doi.org
High-Dimensional Privacy-Utility Dynamics of Noisy Stochastic Gradient Descent on Least Squares
Shurong Lin
Eric D. Kolaczyk
Adam Smith
Elliot Paquette
2025-10-18
ArXiv (prépublication)
doi.org
arxiv.org
PEACE: Prompt Engineering Automation for CLIPSeg Enhancement for Safe-Landing Zone Segmentation
Haechan Mark Bong
Rongge Zhang
Antoine Robillard
Giovanni Beltrame
Safe landing is essential in robotics applications, from industrial settings to space exploration. As artificial intelligence advances, we h… (voir plus)ave developed PEACE (Prompt Engineering Automation for CLIPSeg Enhancement), a system that automatically generates and refines prompts for identifying landing zones in changing environments. Traditional approaches using fixed prompts for open-vocabulary models struggle with environmental changes and can lead to dangerous outcomes when conditions are not represented in the predefined prompts. PEACE addresses this limitation by dynamically adapting to shifting data distributions. Our key innovation is the dual segmentation of safe and unsafe landing zones, allowing the system to refine the results by removing unsafe areas from potential landing sites. Using only monocular cameras and image segmentation, PEACE can safely guide descent operations from 100 meters to altitudes as low as 20 meters. The testing shows that PEACE significantly outperforms the standard CLIP and CLIPSeg prompting methods, improving the successful identification of safe landing zones from 57% to 92%. We have also demonstrated enhanced performance when replacing CLIPSeg with FastSAM. The complete source code is available as an open-source software 1.
2025-10-18
2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (publié)
doi.org
arxiv.org
Perpetua: Multi-Hypothesis Persistence Modeling for Semi-Static Environments
Miguel Saavedra-Ruiz
Samer B. Nashed
Charlie Gauthier
Liam Paull
Many robotic systems require extended deployments in complex, dynamic environments. In such deployments, parts of the environment may change… (voir plus) between subsequent robot observations. Most robotic mapping or environment modeling algorithms are incapable of representing dynamic features in a way that enables predicting their future state. Instead, they opt to filter certain state observations, either by removing them or some form of weighted averaging. This paper introduces Perpetua, a method for modeling the dynamics of semi-static features. Perpetua is able to: incorporate prior knowledge about the dynamics of the feature if it exists, track multiple hypotheses, and adapt over time to enable predicting of future feature states. Specifically, we chain together mixtures of"persistence"and"emergence"filters to model the probability that features will disappear or reappear in a formal Bayesian framework. The approach is an efficient, scalable, general, and robust method for estimating the states of features in an environment, both in the present as well as at arbitrary future times. Through experiments on simulated and real-world data, we find that Perpetua yields better accuracy than similar approaches while also being online adaptable and robust to missing observations.
2025-10-18
2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (publié)
doi.org
arxiv.org
Prompt4Trust: A Reinforcement Learning Prompt Augmentation Framework for Clinically-Aligned Confidence Calibration in Multimodal Large Language Models
Anita Kriz
Elizabeth Laura Janes
Xing Shen
Tal Arbel
2025-10-18
ICCVW @ IEEE/CVF International Conference on Computer Vision (publié)
doi.org
arxiv.org
Continuously Learning Bug Locations
Paulina Stevia Nouwou Mindom
Leuson Da Silva
Amin Nikanjam
Foutse Khomh
Automatically locating buggy changesets associated with bug reports is crucial in the software development process. Deep Learning (DL)-based… (voir plus) techniques show promising results by leveraging structural information from the code and learning links between changesets and bug reports. However, since source code associated with changesets evolves, the performance of such models tends to degrade over time due to concept drift. Aiming to address this challenge, in this paper, we evaluate the potential of using Continual Learning (CL) techniques in multiple sub-tasks setting for bug localization (each of which operates on either stationary or non-stationary data), comparing it against a bug localization technique that leverages the BERT model, a deep reinforcement learning-based technique that leverages the A2C algorithm, and a DL-based function-level interaction model for semantic bug localization. Additionally, we enhanced the CL techniques by using logistic regression to identify and integrate the most significant bug-inducing factors. Our empirical evaluation across seven widely used software projects shows that CL techniques perform better than DL-based techniques by up to 61% in terms of Mean Reciprocal Rank (MRR), 44% in terms of Mean Average Precision (MAP), 83% in terms of top@1, 56% in terms of top@5, and 66% in terms of top@10 metrics in non-stationary setting. Further, we show that the CL techniques we studied are effective at localizing changesets relevant to a bug report while being able to mitigate catastrophic forgetting across the studied tasks and require up to 5x less computational effort during training. Our findings demonstrate the potential of adopting CL for bug localization in non-stationary settings, and we hope it helps to improve bug localization activities in Software Engineering using CL techniques.
2025-10-17
ACM Transactions on Software Engineering and Methodology (publié)
doi.org
arxiv.org
Hierarchical Differentiable Fluid Simulation
Xiangyu Kong
Arnaud Schoentgen
Damien Rioux‐Lavoie
Paul G. Kry
Derek Nowrouzezahrai
Differentiable simulation is an emerging field that offers a powerful and flexible route to fluid control. In grid‐based settings, high me… (voir plus)mory consumption is a long‐standing bottleneck that constrains optimization resolution. We introduce a two‐step algorithm that significantly reduces memory usage: our method first optimizes for bulk forces at reduced resolution, then refines local details over sub‐domains while maintaining differentiability. In trading runtime for memory, it enables optimization at previously unattainable resolutions. We validate its effectiveness and memory savings on a series of fluid control problems.
2025-10-16
Computer Graphics Forum (publié)
doi.org
Improving autoformalization via cycle consistency and incremental type-checking using language-model probabilistic programs
Mauricio Barba da Costa
Fabian Zaiser
Katherine M. Collins
Romir Patel
Timothy J. O'Donnell
Alexander K. Lew
Joshua B. Tenenbaum
Vikash Mansinghka
Cameron Freer
2025-10-16
NeurIPS.cc/2025/Workshop/MATH-AI (poster)
openreview.net
openreview.net
Learning Heuristics for Transit Network Design and Improvement with Deep Reinforcement Learning
Andrew Holliday
Ahmed El-Geneidy
Gregory Dudek
2025-10-16
Transportmetrica B: Transport Dynamics (publié)
doi.org
arxiv.org