This year, RoboCup will be held in Salvador, Brazil, from 15–21 July. The event will see around 3000 participants, from 300-400 different teams take part in competitions, training sessions, and a symposium. You can see the schedule for the week here. The league competitions will take place on 17-20 July. This year will see a new demo at the event – the Flying Robots Demo.

We often attribute human characteristics to large language models (LLMs) and claim that they "know" certain things. LLMs have an internal probabilistic knowledge that represents information retained during training. How can we assess the veracity of this knowledge? We examine two common methods for probing the veracity of LLMs and discover several assumptions that are flawed. To address these flawed assumptions, we introduce sAwMIL (short for Sparse Aware Multiple-Instance Learning), a probing method that utilizes the internal activations of LLMs to separate statements into true, false, and neither. sAwMIL is based on multiple-instance learning and conformal prediction. We evaluate sAwMIL on 5 validity criteria across 16 open-source LLMs, including both default and chat-based variants, as well as on 3 new datasets. Among the insights we provide are: (1) the veracity signal is often concentrated in the third quarter of an LLM's depth; (2) truth and falsehood signals are not always symmetric; (3) linear probes perform better on chat models than on default models; (4) nonlinear probes may be required to capture veracity signals for some LLMs with reinforcement learning from human feedback or knowledge distillation; and (5) LLMs capture a third type of signal that is distinct from true and false and is neither true nor false. These findings provide a reliable method for verifying what LLMs "know" and how certain they are of their probabilistic internal knowledge.

Differentially private (DP) mechanisms are difficult to interpret and calibrate because existing methods for mapping standard privacy parameters to concrete privacy risks -- re-identification, attribute inference, and data reconstruction -- are both overly pessimistic and inconsistent. In this work, we use the hypothesis-testing interpretation of DP ($f$-DP), and determine that bounds on attack success can take the same unified form across re-identification, attribute inference, and data reconstruction risks. Our unified bounds are (1) consistent across a multitude of attack settings, and (2) tunable, enabling practitioners to evaluate risk with respect to arbitrary (including worst-case) levels of baseline risk. Empirically, our results are tighter than prior methods using $\varepsilon$-DP, Rényi DP, and concentrated DP. As a result, calibrating noise using our bounds can reduce the required noise by 20% at the same risk level, which yields, e.g., more than 15pp accuracy increase in a text classification task. Overall, this unifying perspective provides a principled framework for interpreting and calibrating the degree of protection in DP against specific levels of re-identification, attribute inference, or data reconstruction risk.

Creating a cast of characters by attending to their relational dynamics is a critical aspect of most long-form storywriting. However, our formative study (N=14) reveals that writers struggle to envision new characters that could influence existing ones, to balance similarities and differences among characters, and to intricately flesh out their relationships. Based on these observations, we designed Constella, an LLM-based multi-agent tool that supports storywriters' interconnected character creation process. Constella suggests related characters (FRIENDS DISCOVERY feature), reveals the inner mindscapes of several characters simultaneously (JOURNALS feature), and manifests relationships through inter-character responses (COMMENTS feature). Our 7-8 day deployment study with storywriters (N=11) shows that Constella enabled the creation of expansive communities composed of related characters, facilitated the comparison of characters' thoughts and emotions, and deepened writers' understanding of character relationships. We conclude by discussing how multi-agent interactions can help distribute writers' attention and effort across the character cast.

In this paper, we propose PLACE (Prompt Learning for Attributed Community Search), an innovative graph prompt learning framework for ACS. Enlightened by prompt-tuning in Natural Language Processing (NLP), where learnable prompt tokens are inserted to contextualize NLP queries, PLACE integrates structural and learnable prompt tokens into the graph as a query-dependent refinement mechanism, forming a prompt-augmented graph. Within this prompt-augmented graph structure, the learned prompt tokens serve as a bridge that strengthens connections between graph nodes for the query, enabling the GNN to more effectively identify patterns of structural cohesiveness and attribute similarity related to the specific query. We employ an alternating training paradigm to optimize both the prompt parameters and the GNN jointly. Moreover, we design a divide-and-conquer strategy to enhance scalability, supporting the model to handle million-scale graphs. Extensive experiments on 9 real-world graphs demonstrate the effectiveness of PLACE for three types of ACS queries, where PLACE achieves higher F1 scores by 22% compared to the state-of-the-arts on average.

AI-powered search systems are emerging as new information gatekeepers, fundamentally transforming how users access news and information. Despite their growing influence, the citation patterns of these systems remain poorly understood. We address this gap by analyzing data from the AI Search Arena, a head-to-head evaluation platform for AI search systems. The dataset comprises over 24,000 conversations and 65,000 responses from models across three major providers: OpenAI, Perplexity, and Google. Among the over 366,000 citations embedded in these responses, 9% reference news sources. We find that while models from different providers cite distinct news sources, they exhibit shared patterns in citation behavior. News citations concentrate heavily among a small number of outlets and display a pronounced liberal bias, though low-credibility sources are rarely cited. User preference analysis reveals that neither the political leaning nor the quality of cited news sources significantly influences user satisfaction. These findings reveal significant challenges in current AI search systems and have important implications for their design and governance.

User growth is a major strategy for consumer internet companies. To optimize costly marketing campaigns and maximize user engagement, we propose a novel treatment effect optimization methodology to enhance user growth marketing. By leveraging deep learning, our algorithm learns from past experiments to optimize user selection and reward allocation, maximizing campaign impact while minimizing costs. Unlike traditional prediction methods, our model directly models uplifts in key business metrics. Further, our deep learning model can jointly optimize parameters for an aggregated loss function using softmax gating. Our approach surpasses traditional methods by directly targeting desired business metrics and demonstrates superior algorithmic flexibility in handling complex business constraints. Comprehensive evaluations, including comparisons with state-of-the-art techniques such as R-learner and Causal Forest, validate the effectiveness of our model. We experimentally demonstrate that our proposed constrained and direct optimization algorithms significantly outperform state-of-the-art methods by over $20\%$, proving their cost-efficiency and real-world impact. The versatile methods can be applied to various product scenarios, including optimal treatment allocation. Its effectiveness has also been validated through successful worldwide production deployments.

We consider a method for conformalizing a stacked ensemble of predictive models, showing that the potentially simple form of the meta-learner at the top of the stack enables a procedure with manageable computational cost that achieves approximate marginal validity without requiring the use of a separate calibration sample. Empirical results indicate that the method compares favorably to a standard inductive alternative.

In scientific domains -- from biology to the social sciences -- many questions boil down to \textit{What effect will we observe if we intervene on a particular variable?} If the causal relationships (e.g.~a causal graph) are known, it is possible to estimate the intervention distributions. In the absence of this domain knowledge, the causal structure must be discovered from the available observational data. However, observational data are often compatible with multiple causal graphs, making methods that commit to a single structure prone to overconfidence. A principled way to manage this structural uncertainty is via Bayesian inference, which averages over a posterior distribution on possible causal structures and functional mechanisms. Unfortunately, the number of causal structures grows super-exponentially with the number of nodes in the graph, making computations intractable. We propose to circumvent these challenges by using meta-learning to create an end-to-end model: the Model-Averaged Causal Estimation Transformer Neural Process (MACE-TNP). The model is trained to predict the Bayesian model-averaged interventional posterior distribution, and its end-to-end nature bypasses the need for expensive calculations. Empirically, we demonstrate that MACE-TNP outperforms strong Bayesian baselines. Our work establishes meta-learning as a flexible and scalable paradigm for approximating complex Bayesian causal inference, that can be scaled to increasingly challenging settings in the future.

We propose Bayesian Hierarchical Invariant Prediction (BHIP) reframing Invariant Causal Prediction (ICP) through the lens of Hierarchical Bayes. We leverage the hierarchical structure to explicitly test invariance of causal mechanisms under heterogeneous data, resulting in improved computational scalability for a larger number of predictors compared to ICP. Moreover, given its Bayesian nature BHIP enables the use of prior information. In this paper, we test two sparsity inducing priors: horseshoe and spike-and-slab, both of which allow us a more reliable identification of causal features. We test BHIP in synthetic and real-world data showing its potential as an alternative inference method to ICP.