Publications

Comparing Bottom-Up and Top-Down Steering Approaches on In-Context Learning Tasks
Madeline Brumley
Joe Kwon
David Scott Krueger
Dmitrii Krasheninnikov
Usman Anwar
A key objective of interpretability research on large language models (LLMs) is to develop methods for robustly steering models toward desir… (voir plus)ed behaviors. To this end, two distinct approaches to interpretability -- ``bottom-up"and ``top-down"-- have been presented, but there has been little quantitative comparison between them. We present a case study comparing the effectiveness of representative vector steering methods from each branch: function vectors (FV; arXiv:2310.15213), as a bottom-up method, and in-context vectors (ICV; arXiv:2311.06668) as a top-down method. While both aim to capture compact representations of broad in-context learning tasks, we find they are effective only on specific types of tasks: ICVs outperform FVs in behavioral shifting, whereas FVs excel in tasks requiring more precision. We discuss the implications for future evaluations of steering methods and for further research into top-down and bottom-up steering given these findings.
2024-11-11
ArXiv (prépublication)
arxiv.org
arxiv.org
Feature learning as alignment: a structural property of gradient descent in non-linear neural networks
Daniel Beaglehole
Ioannis Mitliagkas
Atish Agarwala
Understanding the mechanisms through which neural networks extract statistics from input-label pairs through feature learning is one of the … (voir plus)most important unsolved problems in supervised learning. Prior works demonstrated that the gram matrices of the weights (the neural feature matrices, NFM) and the average gradient outer products (AGOP) become correlated during training, in a statement known as the neural feature ansatz (NFA). Through the NFA, the authors introduce mapping with the AGOP as a general mechanism for neural feature learning. However, these works do not provide a theoretical explanation for this correlation or its origins. In this work, we further clarify the nature of this correlation, and explain its emergence. We show that this correlation is equivalent to alignment between the left singular structure of the weight matrices and the newly defined pre-activation tangent features at each layer. We further establish that the alignment is driven by the interaction of weight changes induced by SGD with the pre-activation features, and analyze the resulting dynamics analytically at early times in terms of simple statistics of the inputs and labels. We prove the derivative alignment occurs with high probability in specific high dimensional settings. Finally, motivated by the observation that the NFA is driven by this centered correlation, we introduce a simple optimization rule that dramatically increases the NFA correlations at any given layer and improves the quality of features learned.
2024-11-11
TMLR (accepté)
openreview.net
openreview.net
Impact of LLM-based Review Comment Generation in Practice: A Mixed Open-/Closed-source User Study
Doriane Olewicki
Léuson M. P. Da Silva
Suhaib Mujahid
Arezou Amini
Benjamin Mah
Marco Castelluccio
Sarra Habchi
Foutse Khomh
Bram Adams
2024-11-11
ArXiv (prépublication)
arxiv.org
arxiv.org
Non-Adversarial Inverse Reinforcement Learning via Successor Feature Matching
Arnav Kumar Jain
Harley Wiltzer
Jesse Farebrother
Irina Rish
Glen Berseth
Sanjiban Choudhury
2024-11-11
ArXiv (prépublication)
arxiv.org
arxiv.org
Specific inhibition and disinhibition in the higher-order structure of a cortical connectome
Michael W. Reimann
Daniela Egas Santander
András Ecker
Eilif Benjamin Muller
Neuronal network activity is thought to be structured around the activation of assemblies, or low-dimensional manifolds describing states of… (voir plus) activity. Both views describe neurons acting not independently, but in concert, likely facilitated by strong recurrent excitation between them. The role of inhibition in these frameworks – if considered at all – is often reduced to blanket inhibition with no specificity with respect to which excitatory neurons are targeted. We analyzed the structure of excitation and inhibition in the MICrONS 1mm3 dataset, an electron microscopic reconstruction of a piece of cortical tissue. We found that excitation was structured around a feed-forward flow in non-random motifs of seven or more neurons. This revealed a structure of information flow from a small number of sources to a larger number of potential targets that became only visible when larger motifs were considered instead of individual pairs. Inhibitory neurons targeted and were targeted by neurons in specific sequential positions of these motifs. Additionally, disynaptic inhibition was strongest between target motifs excited by the same group of source neurons, implying competition between them. The structure of this inhibition was also highly specific and symmetrical, contradicting the idea of non-specific blanket inhibition. None of these trends are detectable in only pairwise connectivity, demonstrating that inhibition is specifically structured by these large motifs. Further, we found that these motifs represent higher order connectivity patterns which are present, but to a lesser extent in a recently released, detailed computational model, and not at all in a distance-dependent control. These findings have important implications for how synaptic plasticity reorganizes neocortical connectivity to implement learning and for the specific role of inhibition in this process.
2024-11-11
Cerebral Cortex (New York, NY) (publié)
doi.org
Reaction-conditioned De Novo Enzyme Design with GENzyme
Chenqing Hua
Jiarui Lu
Yong Liu
Odin Zhang
Jian Tang
Rex Ying
Wengong Jin
Guy Wolf
Doina Precup
Shuangjia Zheng
The introduction of models like RFDiffusionAA, AlphaFold3, AlphaProteo, and Chai1 has revolutionized protein structure modeling and interact… (voir plus)ion prediction, primarily from a binding perspective, focusing on creating ideal lock-and-key models. However, these methods can fall short for enzyme-substrate interactions, where perfect binding models are rare, and induced fit states are more common. To address this, we shift to a functional perspective for enzyme design, where the enzyme function is defined by the reaction it catalyzes. Here, we introduce \textsc{GENzyme}, a \textit{de novo} enzyme design model that takes a catalytic reaction as input and generates the catalytic pocket, full enzyme structure, and enzyme-substrate binding complex. \textsc{GENzyme} is an end-to-end, three-staged model that integrates (1) a catalytic pocket generation and sequence co-design module, (2) a pocket inpainting and enzyme inverse folding module, and (3) a binding and screening module to optimize and predict enzyme-substrate complexes. The entire design process is driven by the catalytic reaction being targeted. This reaction-first approach allows for more accurate and biologically relevant enzyme design, potentially surpassing structure-based and binding-focused models in creating enzymes capable of catalyzing specific reactions. We provide \textsc{GENzyme} code at https://github.com/WillHua127/GENzyme.
2024-11-10
ArXiv (prépublication)
arxiv.org
arxiv.org
Optimal Approximate Minimization of One-Letter Weighted Finite Automata
Clara Lacroce
Borja Balle
Prakash Panangaden
Guillaume Rabusseau
2024-11-08
Mathematical Structures in Computer Science (publié)
doi.org
arxiv.org
Robustness of Neural Ratio and Posterior Estimators to Distributional Shifts for Population-Level Dark Matter Analysis in Strong Gravitational Lensing
Andreas Filipp
Yashar Hezaveh
Laurence Perreault-Levasseur
2024-11-08
ArXiv (prépublication)
arxiv.org
arxiv.org
Solving Hidden Monotone Variational Inequalities with Surrogate Losses
Ryan D'Orazio
Danilo Vucetic
Zichu Liu
Junhyung Lyle Kim
Ioannis Mitliagkas
Gauthier Gidel
Deep learning has proven to be effective in a wide variety of loss minimization problems. However, many applications of interest, like minim… (voir plus)izing projected Bellman error and min-max optimization, cannot be modelled as minimizing a scalar loss function but instead correspond to solving a variational inequality (VI) problem. This difference in setting has caused many practical challenges as naive gradient-based approaches from supervised learning tend to diverge and cycle in the VI case. In this work, we propose a principled surrogate-based approach compatible with deep learning to solve VIs. We show that our surrogate-based approach has three main benefits: (1) under assumptions that are realistic in practice (when hidden monotone structure is present, interpolation, and sufficient optimization of the surrogates), it guarantees convergence, (2) it provides a unifying perspective of existing methods, and (3) is amenable to existing deep learning optimizers like ADAM. Experimentally, we demonstrate our surrogate-based approach is effective in min-max optimization and minimizing projected Bellman error. Furthermore, in the deep reinforcement learning case, we propose a novel variant of TD(0) which is more compute and sample efficient.
2024-11-07
ArXiv (prépublication)
arxiv.org
arxiv.org
Unlearning in- vs. out-of-distribution data in LLMs under gradient-based method
Teodora Băluță
Pascal Lamblin
Daniel Tarlow
Fabian Pedregosa
Gintare Karolina Dziugaite
Machine unlearning aims to solve the problem of removing the influence of selected training examples from a learned model. Despite the incre… (voir plus)asing attention to this problem, it remains an open research question how to evaluate unlearning in large language models (LLMs), and what are the critical properties of the data to be unlearned that affect the quality and efficiency of unlearning. This work formalizes a metric to evaluate unlearning quality in generative models, and uses it to assess the trade-offs between unlearning quality and performance. We demonstrate that unlearning out-of-distribution examples requires more unlearning steps but overall presents a better trade-off overall. For in-distribution examples, however, we observe a rapid decay in performance as unlearning progresses. We further evaluate how example's memorization and difficulty affect unlearning under a classical gradient ascent-based approach.
2024-11-07
ArXiv (prépublication)
arxiv.org
arxiv.org
Boosting Latent Diffusion with Perceptual Objectives
Tariq Berrada
Pietro Astolfi
Jakob Verbeek
Melissa Hall
Marton Havasi
Michal Drozdzal
Yohann Benchetrit
Adriana Romero Soriano
Karteek Alahari
2024-11-06
ArXiv (prépublication)
arxiv.org
arxiv.org
A Capacitated Collection-and-Delivery-Point Location Problem with Random Utility Maximizing Customers
David Pinzon Ulloa
Ammar Metnan
Emma Frejinger
2024-11-06
ArXiv (prépublication)
arxiv.org
arxiv.org