Publications

Accelerated Inorganic Materials Design with Generative AI Agents
Izumi Takahara
Teruyasu Mizoguchi
Bang Liu
2025-09-19
NeurIPS.cc/2025/Workshop/AI4Mat (poster)
openreview.net
openreview.net
Concept-based Steering of Large Language Models for Conditional Molecular Generation
Jeremy Qin
Rushil Gupta
Boris Knyazev
Yang Zhang
Glen Berseth
Bang Liu
Modern LLMs, with their internet-scale pretraining and advanced human-level capabilities across specialized tasks, have demonstrated promisi… (voir plus)ng performance in molecular discovery using existing text-based molecular representations, such as SMILES and SELFIES. However, generating valid, unique, and high-fidelity molecules while precisely controlling for multiple properties simultaneously remains challenging. While prior works demonstrated success by fine-tuning language models on a novel corpus of molecules with property-conditioned tags, real-world applications require generating molecules from diverse property distributions, previously unseen in the training data. To this end, we present Concept-based Activation STeering (CAST), the first approach to apply activation steering to directly edit a model's internal representation for conditional molecular generation. CAST offers a lightweight, flexible alternative to fine-tuning by computing property-conditioned steering vectors via a concept network that does not require retraining the LLM. Through extensive experiments on datasets such as Therapeutics Data Commons, we show that CAST consistently outperforms existing methods on both in-distribution and out-of-distribution conditional generation tasks. We also conduct comprehensive ablation studies to highlight the extent of control our concept-guided steering provides on the molecules generated by the LLM.
2025-09-19
NeurIPS.cc/2025/Workshop/AI4Mat (poster)
openreview.net
openreview.net
Exposing and Mitigating Calibration Biases and Demographic Unfairness in MLLM Few-Shot In-Context Learning for Medical Image Classification
Xing Shen
Justin Szeto
Mingyang Li
Hengguan Huang
Tal Arbel
2025-09-19
Lecture Notes in Computer Science (publié)
doi.org
arxiv.org
The curriculum effect in visual learning: the role of readout dimensionality
Charlotte Volk
Christopher C. Pack
Shahab Bakhtiari
2025-09-19
bioRxiv (prépublication)
doi.org
Variational Visible Layers: A Practical Framework for Uncertainty Estimation
Zeinab Abboud
Hervé Lombaert
Samuel Kadoury
2025-09-19
Lecture Notes in Computer Science (publié)
doi.org
FocalCodec-Stream: Streaming Low-Bitrate Speech Coding via Causal Distillation
Luca Della Libera
Yusuf Cem Sübakan
Mirco Ravanaelli
Neural audio codecs are a fundamental component of modern generative audio pipelines. Although recent codecs achieve strong low-bitrate reco… (voir plus)nstruction and provide powerful representations for downstream tasks, most are non-streamable, limiting their use in real-time applications. We present FocalCodec-Stream, a hybrid codec based on focal modulation that compresses speech into a single binary codebook at 0.55 - 0.80 kbps with a theoretical latency of 80 ms. Our approach combines multi-stage causal distillation of WavLM with targeted architectural improvements, including a lightweight refiner module that enhances quality under latency constraints. Experiments show that FocalCodec-Stream outperforms existing streamable codecs at comparable bitrates, while preserving both semantic and acoustic information. The result is a favorable trade-off between reconstruction quality, downstream task performance, latency, and efficiency. Code and checkpoints will be released at https://github.com/lucadellalib/focalcodec.
2025-09-18
ArXiv (prépublication)
doi.org
arxiv.org
Spherical Harmonic Exponentials for Efficient Glossy Reflections
Ari Silvennoinen
Peter‐Pike Sloan
Michaƚ Iwanicki
Derek Nowrouzezahrai
Abstract We propose a high‐performance and compact method for computing glossy specular reflections. Commonly‐used prefiltered environme… (voir plus)nt maps have large storage requirements and high error due to constrained treatment of view‐dependence. We propose a factorized spherical harmonic exponential representation that exploits new observations of the benefits of log‐space reconstruction for reflectance. Our method is compact, properly accounts for view‐dependent reflections, and is more accurate than the state‐of‐the‐industry solutions. We achieve higher quality results with an order of magnitude less memory, all with efficient and alias‐free reconstruction of glossy reflections from environment lights and continuously‐varying material roughness.
2025-09-18
Computer Graphics Forum (publié)
doi.org
Capturing Individual Human Preferences with Reward Features
Andre Barreto
Vincent Dumoulin
Yiran Mao
Nicolas Perez-Nieves
Mark Rowland
Bobak Shahriari
Yann Dauphin
Doina Precup
Hugo Larochelle
Reinforcement learning from human feedback usually models preferences using a reward model that does not distinguish between people. We argu… (voir plus)e that this is unlikely to be a good design choice in contexts with high potential for disagreement, like in the training of large language models. We propose a method to specialise a reward model to a person or group of people. Our approach builds on the observation that individual preferences can be captured as a linear combination of a set of general reward features. We show how to learn such features and subsequently use them to quickly adapt the reward model to a specific individual, even if their preferences are not reflected in the training data. We present experiments with large language models comparing the proposed architecture with a non-adaptive reward model and also adaptive counterparts, including models that do in-context personalisation. Depending on how much disagreement there is in the training data, our model either significantly outperforms the baselines or matches their performance with a simpler architecture and more stable training.
2025-09-17
NeurIPS.cc/2025/Conference (poster)
doi.org
openreview.net
Detecting High-Stakes Interactions with Activation Probes
Alex McKenzie
Urja Pawar
Phil Blandfort
William Bankes
David M. Krueger
Ekdeep Singh Lubana
Dmitrii Krasheninnikov
Monitoring is an important aspect of safely deploying Large Language Models (LLMs). This paper examines activation probes for detecting "hig… (voir plus)h-stakes" interactions -- where the text indicates that the interaction might lead to significant harm -- as a critical, yet underexplored, target for such monitoring. We evaluate several probe architectures trained on synthetic data, and find them to exhibit robust generalization to diverse, out-of-distribution, real-world data. Probes' performance is comparable to that of prompted or finetuned medium-sized LLM monitors, while offering computational savings of six orders-of-magnitude. Our experiments also highlight the potential of building resource-aware hierarchical monitoring systems, where probes serve as an efficient initial filter and flag cases for more expensive downstream analysis. We release our novel synthetic dataset and codebase to encourage further study.
2025-09-17
NeurIPS.cc/2025/Conference (poster)
doi.org
openreview.net
Distributional Training Data Attribution: What do Influence Functions Sample?
Bruno Mlodozeniec
Isaac Reid
Samuel Power
David M. Krueger
Murat A Erdogdu
Richard E. Turner
Roger Baker Grosse
2025-09-17
NeurIPS.cc/2025/Conference (spotlight)
openreview.net
openreview.net
FreqPolicy: Efficient Flow-based Visuomotor Policy via Frequency Consistency
Yifei Su
Ning Liu
Dong Chen
Zhen Zhao
Kun Wu
Meng Li
Zhiyuan Xu
Zhengping Che
Jian Tang
2025-09-17
NeurIPS.cc/2025/Conference (poster)
openreview.net
openreview.net
From Dormant to Deleted: Tamper-Resistant Unlearning Through Weight-Space Regularization
Shoaib Ahmed Siddiqui
Adrian Weller
David M. Krueger
Gintare Karolina Dziugaite
Michael Curtis Mozer
Eleni Triantafillou
Recent unlearning methods for LLMs are vulnerable to relearning attacks: knowledge believed-to-be-unlearned re-emerges by fine-tuning on a s… (voir plus)mall set of (even seemingly-unrelated) examples. We study this phenomenon in a controlled setting for example-level unlearning in vision classifiers. We make the surprising discovery that forget-set accuracy can recover from around 50% post-unlearning to nearly 100% with fine-tuning on just the retain set -- i.e., zero examples of the forget set. We observe this effect across a wide variety of unlearning methods, whereas for a model retrained from scratch excluding the forget set (gold standard), the accuracy remains at 50%. We observe that resistance to relearning attacks can be predicted by weight-space properties, specifically,
2025-09-17
NeurIPS.cc/2025/Conference (poster)
doi.org
openreview.net