Emiliano Penaloza

Elena Tiukhova, Emiliano Penaloza, María Óskarsdóttir et al.

Leveraging network information for prediction tasks has become a common practice in many domains. Being an important part of targeted marketing, influencer detection can potentially benefit from incorporating dynamic network representation. In this work, we investigate different dynamic Graph Neural Networks (GNNs) configurations for influencer detection and evaluate their prediction performance using a unique corporate data set. We show that using deep multi-head attention in GNN and encoding temporal attributes significantly improves performance. Furthermore, our empirical evaluation illustrates that capturing neighborhood representation is more beneficial that using network centrality measures.

Elena Tiukhova, Emiliano Penaloza, María Óskarsdóttir et al.

Leveraging network information for predictive modeling has become widespread in many domains. Within the realm of referral and targeted marketing, influencer detection stands out as an area that could greatly benefit from the incorporation of dynamic network representation due to the continuous evolution of customer-brand relationships. In this paper, we present INFLECT-DGNN, a new method for profit-driven INFLuencer prEdiCTion with Dynamic Graph Neural Networks that innovatively combines Graph Neural Networks (GNNs) and Recurrent Neural Networks (RNNs) with weighted loss functions, synthetic minority oversampling adapted to graph data, and a carefully crafted rolling-window strategy. We introduce a novel profit-driven framework that supports decision-making based on model predictions. To test the framework, we use a unique corporate dataset with diverse networks, capturing the customer interactions across three cities with different socioeconomic and demographic characteristics. Our results show how using RNNs to encode temporal attributes alongside GNNs significantly improves predictive performance, while the profit-driven framework determines the optimal classification threshold for profit maximization. We compare the results of different models to demonstrate the importance of capturing network representation, temporal dependencies, and using a profit-driven evaluation. Our research has significant implications for the fields of referral and targeted marketing, expanding the technical use of deep graph learning within corporate environments.

Linfeng Du, Ye Yuan, Zichen Zhao et al.

Large language models (LLMs) excel at general-purpose tasks, yet adapting their responses to individual users remains challenging. Retrieval augmentation provides a lightweight alternative to fine-tuning by conditioning LLMs on user history records, and existing approaches typically select these records based on semantic relevance. We argue that relevance serves as an unreliable proxy for utility: a record may be semantically similar to a query yet fail to improve generation quality or even degrade it due to redundancy or conflicting information. To bridge this gap, we propose PURPLE, a contextual bandit framework that oPtimizes UseR Profiles for LLM pErsonalization. In contrast to a greedy selection of the most relevant records, PURPLE treats profile construction as an order-sensitive generation process and utilizes a Plackett-Luce ranking model to capture complex inter-record dependencies. By training with semantically rich feedback provided by the likelihood of the reference response, our method aligns retrieval directly with generation quality. Extensive experiments on nine personalization tasks demonstrate that PURPLE consistently outperforms strong heuristic and retrieval-augmented baselines in both effectiveness and efficiency, establishing a principled and scalable solution for optimizing user profiles.

Dheeraj Vattikonda, Santhoshi Ravichandran, Emiliano Penaloza et al. · mila

LLM-based web agents have recently made significant progress, but much of it has occurred in closed-source systems, widening the gap with open-source alternatives. Progress has been held back by two key challenges: first, a narrow focus on single-step tasks that overlooks the complexity of multi-step web interactions; and second, the high compute costs required to post-train LLM-based web agents. To address this, we present the first statistically grounded study on compute allocation for LLM web-agent post-training. Our approach uses a two-stage pipeline, training a Llama 3.1 8B student to imitate a Llama 3.3 70B teacher via supervised fine-tuning (SFT), followed by on-policy reinforcement learning. We find this process highly sensitive to hyperparameter choices, making exhaustive sweeps impractical. To spare others from expensive trial-and-error, we sample 1,370 configurations and use bootstrapping to estimate effective hyperparameters. Our results show that combining SFT with on-policy RL consistently outperforms either approach alone on both WorkArena and MiniWob++. Further, this strategy requires only 55% of the compute to match the peak performance of pure SFT on MiniWob++, effectively pushing the compute-performance Pareto frontier, and is the only strategy that can close the gap with closed-source models.

Emiliano Penaloza, Tianyue H. Zhang, Laurent Charlin et al.

Concept Bottleneck Models (CBMs) propose to enhance the trustworthiness of AI systems by constraining their decisions on a set of human-understandable concepts. However, CBMs typically assume that datasets contain accurate concept labels-an assumption often violated in practice, which we show can significantly degrade performance (by 25% in some cases). To address this, we introduce the Concept Preference Optimization (CPO) objective, a new loss function based on Direct Preference Optimization, which effectively mitigates the negative impact of concept mislabeling on CBM performance. We provide an analysis of key properties of the CPO objective, showing it directly optimizes for the concept's posterior distribution, and contrast it against Binary Cross Entropy (BCE), demonstrating that CPO is inherently less sensitive to concept noise. We empirically confirm our analysis by finding that CPO consistently outperforms BCE on three real-world datasets, both with and without added label noise. We make our code available on Github.

Emiliano Penaloza, Nathaniel Stevens

Detecting anomalous behavior in dynamic networks remains a constant challenge. This problem is further exacerbated when the underlying topology of these networks is affected by individual highly-dimensional node attributes. We address this issue by tracking a network's modularity as a proxy of its community structure. We leverage Graph Neural Networks (GNNs) to estimate each snapshot's modularity. GNNs can account for both network structure and high-dimensional node attributes, providing a comprehensive approach for estimating network statistics. Our method is validated through simulations that demonstrate its ability to detect changes in highly-attributed networks by analyzing shifts in modularity. Moreover, we find our method is able to detect a real-world event within the \#Iran Twitter reply network, where each node has high-dimensional textual attributes.

Emiliano Penaloza, Dheeraj Vattikonda, Nicolas Gontier et al.

Training-time privileged information (PI) can enable language models to succeed on tasks they would otherwise fail, making it a powerful tool for reinforcement learning in hard, long-horizon settings. However, transferring capabilities learned with PI to policies that must act without it at inference time remains a fundamental challenge. We study this problem in the context of distilling frontier models for multi-turn agentic environments, where closed-source systems typically hide their internal reasoning and expose only action trajectories. This breaks standard distillation pipelines, since successful behavior is observable but the reasoning process is not. For this, we introduce π-Distill, a joint teacher-student objective that trains a PI-conditioned teacher and an unconditioned student simultaneously using the same model. Additionally, we also introduce On-Policy Self-Distillation (OPSD), an alternative approach that trains using Reinforcement Learning (RL) with a reverse KL-penalty between the student and the PI-conditioned teacher. We show that both of these algorithms effectively distill frontier agents using action-only PI. Specifically we find that π-Distill and in some cases OPSD, outperform industry standard practices (Supervised finetuning followed by RL) that assume access to full Chain-of-Thought supervision across multiple agentic benchmarks, models, and forms of PI. We complement our results with extensive analysis that characterizes the factors enabling effective learning with PI, focusing primarily on π-Distill and characterizing when OPSD is competitive.

Emiliano Penaloza, Olivier Gouvert, Haolun Wu et al.

Traditional recommender systems rely on high-dimensional (latent) embeddings for modeling user-item interactions, often resulting in opaque representations that lack interpretability. Moreover, these systems offer limited control to users over their recommendations. Inspired by recent work, we introduce TExtuAl Representations for Scrutable recommendations (TEARS) to address these challenges. Instead of representing a user's interests through a latent embedding, TEARS encodes them in natural text, providing transparency and allowing users to edit them. To do so, TEARS uses a modern LLM to generate user summaries based on user preferences. We find the summaries capture user preferences uniquely. Using these summaries, we take a hybrid approach where we use an optimal transport procedure to align the summaries' representation with the learned representation of a standard VAE for collaborative filtering. We find this approach can surpass the performance of three popular VAE models while providing user-controllable recommendations. We also analyze the controllability of TEARS through three simulated user tasks to evaluate the effectiveness of a user editing its summary.