Multimodal conversational recommendation has emerged as a promising paradigm for delivering personalized experiences through natural dialogue enriched by visual and contextual grounding. Yet, current multimodal conversational recommendation datasets remain limited: existing resources either simulate conversations, omit user history, or fail to collect sufficiently detailed feedback, all of which constrain the types of research and evaluation they support. To address these gaps, we introduce VOGUE, a novel dataset of 60 humanhuman dialogues in realistic fashion shopping scenarios. Each dialogue is paired with a shared visual catalogue, item metadata, user fashion profiles and histories, and post-conversation ratings from both Seekers and Assistants. This design enables rigorous evaluation of conversational inference, including not only alignment between predicted and ground-truth preferences, but also calibration against full rating distributions and comparison with explicit and implicit user satisfaction signals. Our initial analyses of VOGUE reveal distinctive dynamics of visually grounded dialogue. For example, recommenders frequently suggest items simultaneously in feature-based groups, which creates distinct conversational phases bridged by Seeker critiques and refinements. Benchmarking multimodal large language models against human recommenders shows that while MLLMs approach human-level alignment in aggregate, they exhibit systematic distribution errors in reproducing human ratings and struggle to generalize preference inference beyond explicitly discussed items. These findings establish VOGUE as both a unique resource for studying multimodal conversational systems and as a challenge dataset beyond the current recommendation capabilities of existing top-tier multimodal foundation models such as GPT-4o-mini, GPT-5-mini, and Gemini-2.5-Flash.

Sets represent a fundamental abstraction across many types of data. To handle the unordered nature of set-structured data, models such as DeepSets and PointNet rely on fixed, non-learnable pooling operations (e.g., sum or max) -- a design choice that can hinder the transferability of learned embeddings and limits model expressivity. More recently, learnable aggregation functions have been proposed as more expressive alternatives. In this work, we advance this line of research by introducing the Neuralized Kolmogorov Mean (NKM) -- a novel, trainable framework for learning a generalized measure of central tendency through an invertible neural function. We further propose quasi-arithmetic neural networks (QUANNs), which incorporate the NKM as a learnable aggregation function. We provide a theoretical analysis showing that, QUANNs are universal approximators for a broad class of common set-function decompositions and, thanks to their invertible neural components, learn more structured latent representations. Empirically, QUANNs outperform state-of-the-art baselines across diverse benchmarks, while learning embeddings that transfer effectively even to tasks that do not involve sets.

To effectively interact with the real world, Large Language Models (LLMs) require entity-based commonsense reasoning, a challenging task that necessitates integrating factual knowledge about specific entities with commonsense inference. Existing datasets for evaluating LLM entity-based commonsense reasoning have largely focused on True/False or multiple-choice questions, leaving the explicit assessment of the model's ability in abductive reasoning about causes and effects and generating explanations largely unexamined. In this work, we introduce CommonWhy, a dataset of 15,000 why questions designed to evaluate entity-based commonsense reasoning about causal relationships in LLMs. CommonWhy also serves as a Knowledge Graph Question Answering (KGQA) benchmark, as all supporting knowledge required to answer its queries is available in the Wikidata knowledge graph. Unlike existing KGQA datasets, which primarily test fact retrieval, CommonWhy targets causal commonsense reasoning, establishing a new paradigm for KGQA evaluation. Experiments with state-of-the-art LLMs and LLM-based KGQA methods reveal their significant shortcomings, including frequent factual hallucinations and failures in causal reasoning.

Analyzing social change requires detecting patterns of continuity and difference over time. While time-series clustering offers a valuable approach, existing techniques are often limited by assuming fixed cluster definitions and static assignments of entities to clusters. To address these limitations, we introduce a unified framework of temporal clustering methods that allows for both dynamic cluster definitions and the transition of entities between clusters, generalizing and extending previous work. We also provide new algorithms for this dynamic clustering that optimize global objectives, with optional constraints on the transitions of entities across clusters. This framework expands the methodological toolkit for analyzing social change, and we provide guidelines for its application. We illustrate our approach with three case studies: polarization of social and political attitudes across U.S. states; cross-national cultural change; and the evolution of neighborhood business patterns. We conclude with directions for further research.

Planning in open-world environments, where agents must act with partially observed states and incomplete knowledge, is a central challenge in embodied AI. Open-world planning involves not only sequencing actions but also determining what information the agent needs to sense to enable those actions. Existing approaches using Large Language Models (LLM) and Vision-Language Models (VLM) cannot reliably plan over long horizons and complex goals, where they often hallucinate and fail to reason causally over agent-environment interactions. Alternatively, classical PDDL planners offer correct and principled reasoning, but fail in open-world settings: they presuppose complete models and depend on exhaustive grounding over all objects, states, and actions; they cannot address misalignment between goal specifications (e.g., “heat the bread”) and action specifications (e.g., “toast the bread”); and they do not generalize across modalities (e.g., text, vision). To address these core challenges: (i) we extend symbolic PDDL into a flexible natural language representation that we term NL-PDDL, improving accessibility for non-expert users as well as generalization over modalities; (ii) we generalize regression-style planning to NL-PDDL with commonsense entailment reasoning to determine what needs to be observed for goal achievement in partially-observed environments with potential goal–action specification misalignment; and (iii) we leverage the lifted specification of NL-PDDL to facilitate open-world planning that avoids exhaustive grounding and yields a time and space complexity independent of the number of ground objects, states, and actions. Our experiments in three diverse domains — classical Blocksworld and the embodied ALFWorld environment with both textual and visual states — show that NL-PDDL substantially outperforms existing baselines, is more robust to longer horizons and more complex goals, and generalizes across modalities.

Background The integration of retrieval-augmented generation (RAG) systems into the domain of medical question-answering (QA) presents a significant opportunity to enhance the effectiveness and accuracy of clinical support systems. Objective This study aimed to explore the design choices within the RAG framework and the use of large language model (LLM) classifiers to optimize medical QA systems, enhancing response quality for patient and caregiver queries of varying risk levels. Methods In total, we curated a dataset of 109 patient and caregiver questions related to heart failure (HF)—categorized into answerable (direct, fact-based queries), helpful deferral (general guidance or lifestyle advisory queries), and nonanswerable (out-of-scope or high-risk and medical intervention queries) types—along with relevant documents and a target answer for each question from the website The Heart Hub. Applying a system architecture leveraging RAG with a structured query taxonomy and robust classification mechanisms, this paper provided an empirical assessment for medical QA on a HF dataset and introduced a QA system pipeline design, providing a foundation for extended application across various medical fields. Specifically, we evaluated design choices in the initial retrieval stage of RAG and their impact on performance. We assessed final answer quality from the generation stage using popular passage scoring methods for QA, such as Recall-Oriented Understudy for Gisting Evaluation (ROUGE), BERTScore, and Intersection over Union score. Results The pipeline first uses an LLM-based classifier, achieving 65% accuracy for answerable and helpful deferral queries and 100% accuracy for identifying nonanswerable queries. In information retrieval, the BioMedical Contrastive Pre-trained Transformers (MedCPT) cross encoder performed best as a dense retrieval method, delivering an average of 93% recall @ 7 through ranked relevance scores to obtain the top documents with recall @ k denoting recall computed over the top-k retrieved items. For further retrieving snippets from such documents, its average performance was 72.5% for sentence-level snippets and 83% for paragraph-level snippets. A second LLM-based classifier, used to refine the generated responses, resulted in an overall reduction in ROUGE-1 recall by 13% and Bidirectional Encoder Representations from Transformers (BERT) precision by 11%. However, Intersection over Union scores, or the overlap between “gold answers” and system answers, increased by 24%, demonstrating enhanced alignment with ground truth responses. This also indicates the system’s improved ability to generate concise and accurate medical responses. Conclusions The implementation of a structured RAG framework paired with LLM classifiers for medical QA introduces a promising avenue for enhancing clinical decision support systems. By systematically analyzing the impact of query taxonomy, retrieval configurations, and response strategies, this approach clarifies the relative importance of each component within the medical RAG system using a HF dataset. Our findings provide actionable guidance on optimal design choices for maximizing retrieval and response accuracy; thus, informing the development of robust, scalable medical QA systems.

Medical diagnostic reasoning is a high-impact use case for LLMs that carries significant implications for the health and wellbeing of users. When OpenAI (2026) reports that more than 5

Diffusion models have shown promise in learning to solve constraint optimization problems. However, they are mostly restricted to problems with binary variables and rely on graph neural networks, hindering their application to a broader range of problems such as those with general discrete variables or constraint structures that necessitate global rather than local reasoning. We investigate the use of Diffusion Transformers to address the aforementioned limitations. A naive implementation performs poorly due to a fundamental mismatch between the standard diffusion process and constraint solving: while the former applies small, incremental denoising across all variables, the latter requires substantially altering specific subsets of variables to attain feasibility or optimality. Our method, Blocked Gibbs Diffusion Transformer (BloGDiT), is the first to address this limitation by replacing standard joint Gaussian denoising with blocked Gaussian denoising. BloGDiT uses iterative block resampling and anneals the block size over time to facilitate large, targeted edits within a block of variables. Across Sudoku, Graph Coloring, Maximum Independent Set, and MaxCut, BloGDiT matches or outperforms existing methods, demonstrating that blocked Gibbs-style diffusion provides a highly effective inductive bias for Transformer-based constraint satisfaction and optimization.

We study the neural program synthesis of $\textit{parameterized}$ function families through the lens of meta-learning with hypernetworks. Given a user intent $U$, a meta-learner $M_{\phi}$ produces a full weight set $\hat{\theta}=M_{\phi}(U)$ for a target neural network with fixed architecture $S$, and the instantiated network $m_{S,\hat{\theta}}(X)\to Y$ realizes the behavior intended for $U$. Classical hypernetworks typically $\textit{ignore the target network’s structure}$ and emit a flat list of weights; as a consequence, they fail to account for $\textit{neuron-permutation symmetry}$—many distinct parameterizations of $S$ implement the same function—so equivalent solutions are treated as different targets, fragmenting supervision and hurting out-of-distribution generalization. To address this, we propose $\textit{Meta-GNN}$, a hypernetwork that constructs a $\textit{neural graph}$ from the target architecture $S$ and applies $\textbf{structure-aware}$ message passing with parameter-tied encoders and decoders. This design reduces the search space during learning by collapsing equivalent classes of target networks, without loss of expressivity. Empirically, across modular arithmetic ($\textit{AddMod}$-$p$), array operations ($\textit{SumFirst}$-$n$), and inverse-rule tasks from 1D-ARC, $\textit{Meta-GNN}$ substantially improves learning and $\textbf{out-of-distribution generalization}$ compared to classic hypernetworks and direct $(U,X)\to Y$ baselines. Mechanistic analyses reveal $\textit{what is learned}$: on $\textit{AddMod}$-$p$ the synthesized Transformers recover the canonical clock representation and admit a compact closed-form map $U\mapsto\theta$. These results demonstrate that structure-aware Meta-GNNs enable reliable generalization to $\textit{unseen program parameterizations}$, providing a critical advance for the nascent field of neural program synthesis.

The central challenge in multimodal generative modeling lies in accurately approximating the joint data distribution, even when some modalities are missing. Existing multimodal VAEs solve this by designing increasingly complex encoding architectures, relying on modality-specific encoders, factorized posteriors, and custom inference procedures. This restricts their ability to capture relations among modalities by amortizing the encoding parameters. We challenge this paradigm by introducing a model trained for arbitrary conditioning, i.e., generating any modality given a subset of observed modalities and a logical index indicating which modalities are present or missing. This enables a single unified encoder to handle any subset of modalities while capturing inter-modal relationships via a compact, shared posterior. We find that to work efficiently in the multimodal setup, arbitrary conditioning requires replacing the KL divergence with Wasserstein regularization, which allows more dispersed latent embeddings to support learning over diverse data and modality subsets. This key insight exposes a critical deficiency in existing methods, which rely on KL regularization that tends to concentrate individual embeddings near the standard Gaussian prior, despite coming from very diverse subsets of multimodal inputs. We prove that Wasserstein regularization ensures that the aggregate latent distribution -- spanning all conditioning subsets -- aligns with the prior without requiring mixture models or auxiliary inference tricks. Empirically, the proposed model improves cross-modal generation and yields better reconstructions than state-of-the-art multimodal VAEs.

Cooperative tasks in Multi-Agent Reinforcement Learning (MARL) require agents to collectively maximize a shared return. Under the Centralized Training with Decentralized Execution (CTDE) paradigm, policy gradients have remained difficult to compute directly. Prior methods largely follow two approaches: independent factorized updates with centralized critics, which lack general joint-improvement guarantees without value decomposition assumptions, or alternating best-response updates, which can converge to suboptimal Nash Equilibria. In this paper, we show the joint policy gradient admits an exact decentralized decomposition of per-agent terms, each formed from per-agent score functions and decentralized critics. Based on this decomposition, we develop Agent-Chained Policy Optimization (ACPO), where actors are trained independently, with their updates together constituting a single step on the joint policy gradient. Central to this result is a serialized view of the simultaneous joint decision in which agents commit actions one at a time, each conditioning on a belief over preceding actions. The belief acts as the coordination mechanism which ties the independent per-agent updates into a joint gradient step. We evaluate ACPO on Multi-Robot Warehouse, SMACv2, and MA-MuJoCo, where it outperforms strong baselines, with the gap widening as the number of agents grows.

Long-horizon conversational agents need to interact with users through evolving events, tasks, and goals. Such histories are naturally temporal, yet many existing memory systems organize information primarily by topical similarity and may ignore the order in which events occur. We introduce Segment Tree Memory, or SegTreeMem, a memory architecture that represents conversation history as a temporally ordered Segment Tree over utterances. SegTreeMem incrementally inserts new utterances through an online rightmost-frontier update rule, preserving chronological order while forming hierarchical memory segments. For retrieval, SegTreeMem propagates relevance scores through the tree to combine local semantic matching with hierarchical temporal context. Across three long-horizon memory benchmarks and two LLM backbones, SegTreeMem improves answer quality over flat retrieval, graph-structured memory, and tree-structured memory baselines. Additional temporal-order permutation analysis shows that the performance gain depends on preserving temporal order during memory construction, supporting the claim that temporal order is a key structure for agentic memory.

LLM-based conversational AI agents struggle to maintain coherent behavior over long horizons due to limited context. While RAG-based approaches are increasingly adopted to overcome this limitation by storing interactions in external memory modules and performing retrieval from them, their effectiveness in answering challenging questions (e.g., multi-hop, commonsense) ultimately depends on the agent's ability to reason over the retrieved information. However, existing methods typically retrieve memory based on semantic similarity to the raw user utterance, which lacks explicit reasoning about missing intermediate facts and often returns evidence that is irrelevant or insufficient for grounded reasoning. In this work, we introduce Goal-Mem, a goal-oriented reasoning framework for RAG-based agentic memory that performs explicit backward chaining from the user's utterance as a goal. Rather than progressively expanding from retrieved context, Goal-Mem decomposes each goal into atomic subgoals, performs targeted memory retrieval to satisfy each subgoal, and iteratively identifies what information from memory should be retrieved when intermediate goals cannot be resolved. We formalize this process in Natural Language Logic, a logical system that combines the verifiability of reasoning provided by FOL with the expressivity of natural language. Through extensive experiments on two datasets and comparing to nine strong memory baselines, we show that Goal-Mem consistently improves performance, particularly on tasks requiring multi-hop reasoning and implicit inference.

The growing ubiquity of Extended Reality (XR) is driving Conversational Recommendation Systems (CRS) toward visually immersive experiences. We formalize this paradigm as Immersive CRS (ICRS), where recommended items are highlighted directly in the user's scene-based visual environment and augmented with in-situ labels. While item recommendation has been widely studied, the problem of how to select and evaluate which information to present as immersive labels remains an open problem. To this end, we introduce a principled categorization of information needs into explicit intent satisfaction and proactive information needs and use these to define novel evaluation metrics for item label selection. We benchmark IR-, LLM-, and VLM-based methods across three datasets and ICRS scenarios: fashion, movie recommendation, and retail shopping. Our evaluation reveals three important limitations of existing methods: (1) they fail to leverage scenario-specific information modalities (e.g., visual cues for fashion, meta-data for retail), (2) they present redundant information that is visually inferable, and (3) they poorly anticipate users' proactive information needs from explicit dialogue alone. In summary, this work provides both a novel evaluation paradigm for in-situ item labeling in ICRS and highlights key challenges for future work.

While Large Language Models (LLMs) exhibit exceptional zero-shot relevance modeling, their high computational cost necessitates framing passage retrieval as a budget-constrained global optimization problem. Existing approaches passively rely on first-stage dense retrievers, which leads to two limitations: (1) failing to retrieve relevant passages in semantically distinct clusters, and (2) failing to propagate relevance signals to the broader corpus. To address these limitations, we propose Bayesian Active Learning with Gaussian Processes guided by LLM relevance scoring (BAGEL), a novel framework that propagates sparse LLM relevance signals across the embedding space to guide global exploration. BAGEL models the multimodal relevance distribution across the entire embedding space with a query-specific Gaussian Process (GP) based on LLM relevance scores. Subsequently, it iteratively selects passages for scoring by strategically balancing the exploitation of high-confidence regions with the exploration of uncertain areas. Extensive experiments across four benchmark datasets and two LLM backbones demonstrate that BAGEL effectively explores and captures complex relevance distributions and outperforms LLM reranking methods under the same LLM budget on all four datasets.

LLM-reranking is limited by the top-k documents retrieved by vector similarity, which neither enables contextual query-document token interactions nor captures multimodal relevance distributions. While LLM query reformulation attempts to improve recall by generating improved or additional queries, it is still followed by vector similarity retrieval. We thus propose to address these top-k retrieval stage failures by introducing ReBOL, which 1) uses LLM query reformulations to initialize a multimodal Bayesian Optimization (BO) posterior over document relevance, and 2) iteratively acquires document batches for LLM query-document relevance scoring followed by posterior updates to optimize relevance. After exploring query reformulation and document batch diversification techniques, we evaluate ReBOL against LLM reranker baselines on five BEIR datasets and using two LLMs (Gemini-2.5-Flash-Lite, GPT-5.2). ReBOL consistently achieves higher recall and competitive rankings, for example compared to the best LLM reranker on the Robust04 dataset with 46.5

Linear Predictive Clustering (LPC) partitions samples based on shared linear relationships between feature and target variables, with numerous applications including marketing, medicine, and education. Greedy optimization methods, commonly used for LPC, alternate between clustering and linear regression but lack global optimality. While effective for separable clusters, they struggle in non-separable settings where clusters overlap in feature space. In an alternative constrained optimization paradigm, previous works formulated LPC as a Mixed-Integer Program (MIP), ensuring global optimality regardless of separability but at the cost of poor scalability. This work builds on the constrained optimization paradigm to introduce two novel approaches that improve the efficiency of global optimization for LPC. By leveraging key theoretical properties of separability, we derive near-optimal approximations with provable error bounds, significantly reducing the MIP formulation’s complexity and improving scalability. Additionally, we can further approximate LPC as a Quadratic Pseudo-Boolean Optimization (QPBO) problem, achieving additional computational gains in the special case of two clusters. Comparative analyses on synthetic and real-world datasets demonstrate that our methods consistently achieve near-optimal solutions with substantially lower regression errors than greedy optimization while exhibiting superior scalability over existing MIP formulations.

Learning how an environment behaves from interaction is central to building agents that adapt to unfamiliar tasks. World models learned with deep networks are flexible but data-hungry and transfer poorly beyond their training distribution. Program-synthesized world models, written as source code by LLMs and refined through counterexample-guided inductive synthesis (CEGIS), are instead data-efficient and reusable, yet they have been demonstrated mainly on structured-state worlds with a given object vocabulary, and a single program search does not scale to pixel-rendered environments whose object structure must be hypothesized flexibly. We introduce OPINE-World, an LLM agent that learns an object-centric programmatic world model online from interaction. OPINE-World couples two cooperating agents in a loop of hypothesis and test, one acting in the environment and one synthesizing the model in code with replay verification and model-based planning, and it steers exploration with a Bayesian measure of object-type adequacy we call ontology error. We evaluate OPINE-World on ARC-AGI-3, a benchmark for skill-acquisition efficiency in which the object vocabulary, the goal, and the action semantics are withheld. OPINE-World solves 20 of 25 games without per-game training and reaches an action-efficiency score of 78.4 against the human baseline.

Conversational AI (ConvAI) agents increasingly maintain structured memory to support long-term, task-oriented interactions. In-context memory approaches append the growing history to the model input, which scales poorly under context-window limits. RAG-based methods retrieve request-relevant information, but most assume flat memory collections and ignore structure. We propose Semantic XPath, a tree-structured memory module to access and update structured conversational memory. Semantic XPath improves performance over flat-RAG baselines by 176.7% while using only 9.1% of the tokens required by in-context memory. We also introduce SemanticXPath Chat, an end-to-end ConvAI demo system that visualizes the structured memory and query execution details. Overall, this paper demonstrates a candidate for the next generation of long-term, task-oriented ConvAI systems built on structured memory.

Natural Language Recommendation (NLRec) generates item suggestions based on the relevance between user-issued NL requests and NL item description passages. Existing NLRec approaches often use Dense Retrieval (DR) to compute item relevance scores from aggregation of inner products between user request embeddings and relevant passage embeddings. However, DR views the request as the sole relevance label, thus leading to a unimodal scoring function centered on the query embedding that is often a weak proxy for query relevance. To better capture the potential multimodal distribution of the relevance scoring function that may arise from complex NLRec data, we propose GPR-LLM that uses Gaussian Process Regression (GPR) with LLM relevance judgments for a subset of candidate passages. Experiments on four NLRec datasets and two LLM backbones demonstrate that GPR-LLM with an RBF kernel, capable of modeling multimodal relevance scoring functions, consistently outperforms simpler unimodal kernels (dot product, cosine similarity), as well as baseline methods including DR, cross-encoder, and pointwise LLM-based relevance scoring by up to 65