Pokémon Go was the world’s first augmented-reality megahit. Released in 2016 by the Google spinout Niantic, the AR twist on the juggernaut Pokémon franchise fast became a global phenomenon. From Chicago to Oslo to Enoshima, players hit the streets in the urgent hope of catching a Jigglypuff or a Squirtle or (with a huge amount of luck) an ultra-rare Galarian Zapdos hovering just out of reach, superimposed on the everyday world.

In short, we’re talking about a huge number of people pointing their phones at a huge number of buildings. “Five hundred million people installed that app in 60 days,” says Brian McClendon, CTO at Niantic Spatial, an AI company that Niantic spun out in May last year. According to the video-game firm Scopely, which bought Pokémon Go from Niantic at the same time, the game still drew more than 100 million players in 2024, eight years after it launched. 

Now Niantic Spatial is using that vast and unparalleled trove of crowdsourced data—images of urban landmarks tagged with super-accurate location markers taken from the phones of hundreds of millions of Pokémon Go players around the world—to build a kind of world model, a buzzy new technology that grounds the smarts of LLMs in real environments. 

The company’s latest product is a model that it says can pinpoint your location on a map to within a few centimeters, based on a handful of snapshots of the buildings or other landmarks in view. The firm wants to use it to help robots navigate with greater precision in places where GPS is unreliable.

In the first big test of its technology, Niantic Spatial has just teamed up with Coco Robotics, a startup that deploys last-mile delivery robots in a number of cities across the US and Europe. “Everybody thought that AR was the future, that AR glasses were coming,” says McClendon. “And then robots became the audience.”

From Pikachu to pizza delivery

Coco Robotics deploys around 1,000 flight-case-size robots—built to carry up to eight extra-large pizzas or four grocery bags—in Los Angeles, Chicago, Jersey City, Miami, and Helsinki. According to CEO Zach Rash, the robots have made more than half a million deliveries to date, covering a few million miles in all weather conditions.

But to compete with human couriers, Coco’s robots, which trundle along sidewalks at around five miles per hour, must be as reliable as possible. “The best way we can do our job is by arriving exactly when we told you we were going to arrive,” says Rash. And that means not getting lost.

The problem Coco faces is that it cannot rely on GPS, which can be weak in cities because radio signals bounce off buildings and interfere with each other. “We do deliveries in a lot of dense areas with high-rises and underpasses and freeways, and those are the areas where GPS just never really works,” says Rash. 

“The urban canyon is the worst place in the world for GPS,” says McClendon. “If you look at that blue dot on your phone, you’ll often see it drift 50 meters, which puts you on a different block going a different direction on the wrong side of the street.” That’s where Niantic Spatial comes in. 

For the last few years, Niantic Spatial has been taking the data collected from players of Pokémon Go and Ingress (Niantic’s previous phone-based AR game, launched in 2013) and building a visual positioning system, technology that tells you where you are based on what you can see. “It turns out that getting Pikachu to realistically run around and getting Coco’s robot to safely and accurately move through the world is actually the same problem,” says John Hanke, CEO of Niantic Spatial.

“Visual positioning is not a very new technology,” says Konrad Wenzel at ESRI, a company that develops digital mapping and geospatial analysis software. “But it’s obvious that the more cameras we have out there, the better it becomes.” 

Niantic Spatial has trained its model on 30 billion images captured in urban environments. In particular, the images are clustered around hot spots—places that served as important locations in Niantic’s games that players were encouraged to visit, such as Pokémon battle arenas. “We had a million-plus locations around the world where we can locate you precisely,” says McClendon. “We know where you’re standing within several centimeters of accuracy and, most importantly, where you’re looking.”

The upshot is that for each of those million locations, Niantic Spatial has many thousands of images taken in more or less the same place but from different angles, at different times of day, and in different weather conditions. Each of those images comes with detailed metadata that pinpoints where in space the phone was at the time it captured the image, including which way the phone was facing, which way up it was, whether or not it was moving, how fast and in which direction, and more.   

The firm has used this data set to train a model to predict exactly where it is by taking into account what it is looking at—even for locations other than those million hot spots, where good sources of image and location data are scarcer.

In addition to GPS, Coco’s robots, which are fitted with four cameras, will now use this model to try to figure out where they are and where they are headed. The robots’ cameras are hip-height and point in all directions at once, so their viewpoint is a little different from a Pokémon Go player’s, but adapting the data was straightforward, says Rash. 

Rival companies use visual positioning systems too. For example, Starship Technologies, a robot delivery firm founded in Estonia in 2014, says its robots use their sensors to build a 3D map of their surroundings, plotting the edges of buildings and the position of streetlights. 

But Rash is betting that Niantic Spatial’s tech will give Coco an edge. He claims it will allow his robots to position themselves in the correct pickup spots outside restaurants, making sure they don’t get in anybody’s way, and stop just outside the customer’s door instead of a few steps away, which might have happened in the past.  

A Cambrian explosion in robotics 

When Niantic Spatial started work on its visual positioning system, the idea was to apply it to augmented reality, says Hanke. “If you are wearing AR glasses and you want the world to lock in to where you’re looking, then you need some method for doing that,” he says. “But now we’re seeing a Cambrian explosion in robotics.”

Some of those robots may need to share spaces with humans—spaces such as construction sites and sidewalks. “If robots are ever going to assimilate into that environment in a way that’s not disruptive for human beings, they’re going to have to have a similar level of spatial understanding,” says Hanke. “We can help robots find exactly where they are when they’ve been jostled and bumped.”

The Coco Robotics partnership is the start. What Niantic Spatial is putting in place, says Hanke, are the first pieces of what he calls a living map: a hyper-detailed virtual simulation of the world that changes as the world changes. As robots from Coco and other firms move about the world, they will provide new sources of map data, feeding into more and more detailed digital replicas of the world. 

But the way Hanke and McClendon see it, maps are not only becoming more detailed; they are being used more and more by machines. That shifts what maps are for. Maps have long been used to help people locate themselves in the world. As they moved from 2D to 3D to 4D (think of real-time simulations, such as digital twins), the basic principle hasn’t changed: Points on the map correspond to points in space or time.

And yet maps for machines may need to become more like guidebooks, full of information that humans take for granted. Companies like Niantic Spatial and ESRI want to add descriptions that tell machines what they’re actually looking at, with every object tagged with a list of its properties. “This era is about building useful descriptions of the world for machines to comprehend,” says Hanke. “The data that we have is a great starting point in terms of building up an understanding of how the connective tissue of the world works.”

There is a lot of buzz about world models right now—and Niantic Spatial knows it. LLMs may seem like know-it-alls, but they have very little common sense when it comes to interpreting and interacting with everyday environments. World models aim to fix that. Some firms, such as Google DeepMind and World Labs, are developing models that generate virtual fantasy worlds on the fly, which can then be used as training dojos for AI agents. 

Niantic Spatial says it is coming at the problem from a different angle. Push map-making far enough and you’ll end up capturing everything, says McClendon: “I’m very focused on trying to re-create the real world. We’re not there yet, but we want to be there.”

arXiv:2603.08575v1 Announce Type: new Abstract: The question of \emph{knowledge}, \emph{truth} and \emph{trust} is explored via a mathematical formulation of claims and sources. We define truth as the reproducibly perceived subset of knowledge, formalize sources as having both generative and discriminative roles, and develop a framework for reputation grounded in the \emph{conviction} -- the likelihood that a source's stance is vindicated by independent consensus. We argue that conviction, rather than correctness or faithfulness, is the principled basis for trust: it is regime-independent, rewards genuine contribution, and demands the transparent and self-sufficient perceptions that make external verification possible. We formalize reputation as the expected weighted signed conviction over a realm of claims, characterize its behavior across source-claim regimes, and identify continuous verification as both a theoretical necessity and a practical mechanism through which reputation accrues. The framework is applied to AI agents, which are identified as capable but error-prone sources for whom verifiable conviction and continuously accrued reputation constitute the only robust foundation for trust.

arXiv:2603.08655v1 Announce Type: new Abstract: We introduce OfficeQA Pro, a benchmark for evaluating AI agents on grounded, multi-document reasoning over a large and heterogeneous document corpus. The corpus consists of U.S. Treasury Bulletins spanning nearly 100 years, comprising 89,000 pages and over 26 million numerical values. OfficeQA Pro consists of 133 questions that require precise document parsing, retrieval, and analytical reasoning across both unstructured text and tabular data. Frontier LLMs including Claude Opus 4.6, GPT-5.4, and Gemini 3.1 Pro Preview achieve less than 5% accuracy on OfficeQA Pro when relying on parametric knowledge, and less than 12% with additional access to the web. When provided directly with the document corpus, frontier agents still struggle on over half of questions, scoring 34.1% on average. We find that providing agents with a structured document representation produced by Databricks' ai_parse_document yields a 16.1% average relative performance gain across agents. We conduct additional ablations to study the effects of model selection, table representation, retrieval strategy, and test-time scaling on performance. Despite these improvements, significant headroom remains before agents can be considered reliable at enterprise-grade grounded reasoning.

arXiv:2603.08704v1 Announce Type: new Abstract: Large language models are increasingly used for financial analysis and investment research, yet systematic evaluation of their financial reasoning capabilities remains limited. In this work, we introduce the AI Financial Intelligence Benchmark (AFIB), a multi-dimensional evaluation framework designed to assess financial analysis capabilities across five dimensions: factual accuracy, analytical completeness, data recency, model consistency, and failure patterns. We evaluate five AI systems: GPT, Gemini, Perplexity, Claude, and SuperInvesting, using a dataset of 95+ structured financial analysis questions derived from real-world equity research tasks. The results reveal substantial differences in performance across models. Within this benchmark setting, SuperInvesting achieves the highest aggregate performance, with an average factual accuracy score of 8.96/10 and the highest completeness score of 56.65/70, while also demonstrating the lowest hallucination rate among evaluated systems. Retrieval-oriented systems such as Perplexity perform strongly on data recency tasks due to live information access but exhibit weaker analytical synthesis and consistency. Overall, the results highlight that financial intelligence in large language models is inherently multi-dimensional, and systems that combine structured financial data access with analytical reasoning capabilities provide the most reliable performance for complex investment research workflows.

arXiv:2603.06603v1 Announce Type: cross Abstract: Data duplication during pretraining can degrade generalization and lead to memorization, motivating aggressive deduplication pipelines. However, at web scale, it is unclear what constitutes a ``duplicate'': beyond surface-form matches, semantically equivalent documents (e.g. translations) may induce redundant training signals once models become sufficiently capable. Practically, this means that semantic duplicates operate increasingly like exact duplicates during training. We present evidence that duplication is scale-dependent in two ways. First, as model capability increases, cross-entropy loss gradients for semantically equivalent documents become more aligned. Smaller models, by contrast, produce gradients that reflect surface similarity (e.g., shared tokens) rather than semantic similarity. Second, we embedded all 192 million FineWeb-Edu-Dedup documents using EmbeddingGemma-300m. For moderate corpus sizes, the cosine similarity between nearest-neighbors follows an isotropic power law baseline. However, as corpus size grows to hundreds of billions of tokens, the nearest-neighbor similarities deviate sharply, indicating accelerated semantic collisions. Finally, controlled pretraining on data sampled with replacement from pools of finite unique documents shows that limited uniqueness yields mild degradation for small models, but rapidly increasing loss penalties for larger models, breaking naive scaling extrapolation. We derive explicit scaling laws that allow practitioners to estimate deviation from expected scaling due to limited semantic uniqueness of the pretraining corpus. Our results identify and resolve an unstudied source of scale-dependence, allowing for more accurate prediction at scale.

arXiv:2603.06607v1 Announce Type: cross Abstract: Multi-agent deep reinforcement learning (DRL) has emerged as a promising approach for radio resource allocation (RRA) in cellular vehicle-to-everything (C-V2X) networks. However, the multifaceted challenges inherent to multi-agent reinforcement learning (MARL) - including non-stationarity, coordination difficulty, large action spaces, partial observability, and limited robustness and generalization - are often intertwined, making it difficult to understand their individual impact on performance in vehicular environments. Moreover, existing studies typically rely on different baseline MARL algorithms, and a systematic comparison of their capabilities in addressing specific challenges in C-V2X RRA remains lacking. In this paper, we bridge this gap by formulating C-V2X RRA as a sequence of multi-agent interference games with progressively increasing complexity, each designed to isolate a key MARL challenge. Based on these formulations, we construct a suite of learning tasks that enable controlled evaluation of performance degradation attributable to each challenge. We further develop large-scale, diverse training and testing datasets using SUMO-generated highway traces to capture a wide range of vehicular topologies and corresponding interference patterns. Through extensive benchmarking of representative MARL algorithms, we identify policy robustness and generalization across diverse vehicular topologies as the dominant challenge in C-V2X RRA. We further show that, on the most challenging task, the best-performing actor-critic method outperforms the value-based approach by 42%. By emphasizing the need for zero-shot policy transfer to both seen and unseen topologies at runtime, and by open-sourcing the code, datasets, and interference-game benchmark suite, this work provides a systematic and reproducible foundation for evaluating and advancing MARL algorithms in vehicular networks.

arXiv:2603.06676v1 Announce Type: cross Abstract: Performing a timely and accurate identification of crop diseases is vital to maintain agricultural productivity and food security. The current work presents a hybrid few-shot learning model that integrates Explainable Artificial Intelligence (XAI) and Few-Shot Learning (FSL) to address the challenge of identifying and classifying the stages of disease of the diseases of maize, rice, and wheat leaves under limited annotated data conditions. The proposed model integrates Siamese and Prototypical Networks within an episodic training paradigm to effectively learn discriminative disease features from a few examples. To ensure model transparency and trustworthiness, Gradient-weighted Class Activation Mapping (Grad-CAM) is employed for visualizing key decision regions in the leaf images, offering interpretable insights into the classification process. Experimental evaluations on custom few-shot datasets developed in the study prove that the model consistently achieves high accuracy, precision, recall, and F1-scores, frequently exceeding 92% across various disease stages. Comparative analyses against baseline FSL models further confirm the superior performance and explainability of the proposed approach. The framework offers a promising solution for real-world, data-constrained agricultural disease monitoring applications.

arXiv:2603.06713v1 Announce Type: cross Abstract: Agentic systems operating over large tool ecosystems must plan and execute long-horizon workflows under weak or non-verifiable supervision. While frontier models mitigate these challenges through scale and large context budgets, small language models (SLMs) remain brittle: eager tool loading saturates context, execution errors compound over time, and sparse rewards limit learning. We introduce ATLAS, a reinforcement finetuning framework that enables SLMs to operate effectively in large-scale toolspace environments by learning how to acquire context and how to execute actions. Our approach makes two key contributions. First, we treat context control and execution structure as learnable decisions, combining iterative tool loading with programmatic tool orchestration to bound context growth and stabilize long-horizon trajectories. Second, we propose rubric-based reinforcement finetuning, which decomposes task success into structured, task-aligned criteria and enables scalable training using small judge models. Across MCP benchmarks, these design choices yield large and consistent gains over generic RL baselines, allowing a 4B SLM to approach frontier-agent performance under far tighter parameter and context budgets.

arXiv:2603.06722v1 Announce Type: cross Abstract: Protein language models often take into consideration the alignment between a protein sequence and its textual description. However, they do not take structural information into consideration. Traditional methods treat sequence and structure separately, limiting the ability to exploit the alignment between the structure and protein sequence embeddings. In this paper, we introduce a sequence structure contrastive alignment framework, which learns a shared embedding space where proteins are represented consistently across modalities. By training on large-scale pairs of sequences and experimentally resolved or predicted structures, the model maximizes agreement between matched sequence structure pairs while pushing apart unrelated pairs. This alignment enables cross-modal retrieval (e.g., finding structural neighbors given a sequence), improves downstream prediction tasks such as function annotation and stability estimation, and provides interpretable links between sequence variation and structural organization. Our results demonstrate that contrastive learning can serve as a powerful bridge between protein sequences and structures, offering a unified representation for understanding and engineering proteins.

arXiv:2603.06749v1 Announce Type: cross Abstract: Robotic foundation models (RFMs) are emerging as a promising route towards flexible, instruction- and demonstration-driven robot control, however, a critical investigation of their industrial applicability is still lacking. This survey gives an extensive overview over the RFM-landscape and analyses, driven by concrete implications, how industrial domains and use cases shape the requirements of RFMs, with particular focus on collaborative robot platforms, heterogeneous sensing and actuation, edge-computing constraints, and safety-critical operation. We synthesise industrial deployment perspectives into eleven interdependent implications and operationalise them into an assessment framework comprising a catalogue of 149 concrete criteria, spanning both model capabilities and ecosystem requirements. Using this framework, we evaluate 324 manipulation-capable RFMs via 48,276 criterion-level decisions obtained via a conservative LLM-assisted evaluation pipeline, validated against expert judgements. The results indicate that industrial maturity is limited and uneven: even the highest-rated models satisfy only a fraction of criteria and typically exhibit narrow implication-specific peaks rather than integrated coverage. We conclude that progress towards industry-grade RFMs depends less on isolated benchmark successes than on systematic incorporation of safety, real-time feasibility, robust perception, interaction, and cost-effective system integration into auditable deployment stacks.

arXiv:2603.06761v1 Announce Type: cross Abstract: Physics-Informed Neural Networks (PINNs) enforce governing equations by penalizing PDE residuals at interior collocation points, but standard collocation strategies - uniform sampling and residual-based adaptive refinement - can oversample smooth regions, produce highly correlated point sets, and incur unnecessary training cost. We reinterpret collocation selection as a coreset construction problem: from a large candidate pool, select a fixed-size subset that is simultaneously informative (high expected impact on reducing PDE error) and diverse (low redundancy under a space-time similarity notion). We formulate this as a QUBO/BQM objective with linear terms encoding residual-based importance and quadratic terms discouraging redundant selections. To avoid the scalability issues of dense k-hot QUBOs, we propose a sparse graph-based BQM built on a kNN similarity graph and an efficient repair procedure that enforces an exact collocation budget. We further introduce hybrid coverage anchors to guarantee global PDE enforcement. We evaluate the method on the 1D time-dependent viscous Burgers equation with shock formation and report both accuracy and end-to-end time-to-accuracy, including a timing breakdown of selection overhead. Results demonstrate that sparse and hybrid formulations reduce selection overhead relative to dense QUBOs while matching or improving accuracy at fixed collocation budgets.

arXiv:2603.06862v1 Announce Type: cross Abstract: Artifact Evaluation (AE) is essential for ensuring the transparency and reliability of research, closing the gap between exploratory work and real-world deployment is particularly important in cybersecurity, particularly in IoT and CPSs, where large-scale, heterogeneous, and privacy-sensitive data meet safety-critical actuation. Yet, manual reproducibility checks are time-consuming and do not scale with growing submission volumes. In this work, we demonstrate that Large Language Models (LLMs) can provide powerful support for AE tasks: (i) text-based reproducibility rating, (ii) autonomous sandboxed execution environment preparation, and (iii) assessment of methodological pitfalls. Our reproducibility-assessment toolkit yields an accuracy of over 72% and autonomously sets up execution environments for 28% of runnable cybersecurity artifacts. Our automated pitfall assessment detects seven prevalent pitfalls with high accuracy ($F_1$ > 92%). Hence, the toolkit significantly reduces reviewer effort and, when integrated into established AE processes, could incentivize authors to submit higher-quality and more reproducible artifacts. IoT, CPS, and cybersecurity conferences and workshops may integrate the toolkit into their peer-review processes to support reviewers' decisions on awarding artifact badges, improving the overall sustainability of the process.

arXiv:2603.06926v1 Announce Type: cross Abstract: Mindfulness meditation is a widely accessible and evidence-based method for supporting mental health. Despite the proliferation of mindfulness meditation apps, sustaining user engagement remains a persistent challenge. Personalizing the meditation experience is a promising strategy to improve engagement, but it often requires costly and unscalable manual effort. We present MindfulAgents, a multi-agent system powered by large language models that (1) generates guided meditation scripts based on an expert-established mindfulness framework, (2) encourages users' reflection on emotional states and mindfulness skills, and (3) enables real-time personalization of the mindfulness meditation experience for each user. In a formative lab study (N=13), MindfulAgents significantly improved in-session engagement (p = 0.011) and self-awareness (p = 0.014), and reduced momentary stress (p = 0.020). Furthermore, a four-week deployment study (N=62) demonstrated a notable increase in long-term engagement (p = 0.002) and level of mindfulness (p = 0.023). Participants reported that MindfulAgents offered more relevant meditation sessions personalized to individual needs in various contexts, supporting sustained practice. Our findings highlight the potential of LLM-driven personalization for enhancing user engagement in digital mindfulness meditation interventions.

arXiv:2603.06957v1 Announce Type: cross Abstract: We study post-training linear autoregressive models with outcome and process rewards. Given a context $\boldsymbol{x}$, the model must predict the response $\boldsymbol{y} \in Y^N$, a sequence of length $N$ that satisfies a $\gamma$ margin condition, an extension of the standard separability to sequences. We prove that on test samples where the base model achieves a non-trivial likelihood $\alpha$, a variant of policy gradient (PG) can achieve likelihood $1 - \varepsilon$ with an essentially minimax optimal number of reward queries $\tilde{O}((\alpha^{-1} + \varepsilon^{-1})/\gamma^2)$. However, a barrier arises for going beyond the support of the base model. We prove that the overall expected error after post-training with outcome rewards is governed by a property of the base model called the Likelihood Quantile (LQ), and that variants of PG, while minimax optimal, may require a number of reward queries exponential in $N$ to go beyond this support, regardless of the pre-training algorithm. To overcome this barrier, we study post-training with a process reward model, and demonstrate how PG variants in this setting avoid the curse of dimensionality in $N$ via dependence on a token-level LQ. Along the way, we prove that under the margin condition, SGD with adaptive learning rate (LR) achieves a near optimal test error for statistical learning, and PG with adaptive LR achieves a near optimal number of mistakes for online learning while being computationally efficient whenever possible, both of which may be of independent interest.

arXiv:2603.07246v1 Announce Type: cross Abstract: Geospatial foundation models provide precomputed embeddings that serve as compact feature vectors for large-scale satellite remote sensing data. While these embeddings can reduce data-transfer bottlenecks and computational costs, Earth observation (EO) applications can still face geometric mismatches between user-defined areas of interest and the fixed precomputed embedding grid. Standard latent-space interpolation is unreliable in this setting because the embedding manifold is highly non-convex, yielding representations that do not correspond to realistic inputs. We verify this using Prithvi-EO-2.0 to understand the shortcomings of interpolation applied to patch embeddings. As a substitute, we propose a Learned Equivariance-Predicting Architecture (LEPA). Instead of averaging vectors, LEPA conditions a predictor on geometric augmentations to directly predict the transformed embedding. We evaluate LEPA on NASA/USGS Harmonized Landsat-Sentinel (HLS) imagery and ImageNet-1k. Experiments show that standard interpolation achieves a mean reciprocal rank (MRR) below 0.2, whereas LEPA increases MRR to over 0.8, enabling accurate geometric adjustment without re-encoding.

arXiv:2603.07253v1 Announce Type: cross Abstract: A significant element of human cooperative intelligence lies in our ability to identify opportunities for fruitful collaboration; and conversely to recognise when the task at hand is better pursued alone. Research on flexible cooperation in machines has left this meta-level problem largely unexplored, despite its importance for successful collaboration in heterogeneous open-ended environments. Here, we extend the typical Ad Hoc Teamwork (AHT) setting to incorporate the idea of agents having heterogeneous goals that in any given scenario may or may not overlap. We introduce a novel approach to learning policies in this setting, based on a hierarchical combination of imitation and reinforcement learning, and show that it outperforms baseline methods across extended versions of two cooperative environments. We also investigate the contribution of an auxiliary component that learns to model teammates by predicting their actions, finding that its effect on performance is inversely related to the amount of observable information about teammate goals.

arXiv:2603.07299v1 Announce Type: cross Abstract: Continuous symmetries are fundamental to many scientific and learning problems, yet they are often unknown a priori. Existing symmetry discovery approaches typically search directly in the space of transformation generators or rely on learned augmentation schemes. We propose a fundamentally different perspective based on spectral structure. We introduce a framework for discovering continuous one-parameter subgroups using the Generalized Fourier Transform (GFT). Our central observation is that invariance to a subgroup induces structured sparsity in the spectral decomposition of a function across irreducible representations. Instead of optimizing over generators, we detect symmetries by identifying this induced sparsity pattern in the spectral domain. We develop symmetry detection procedures on maximal tori, where the GFT reduces to multi-dimensional Fourier analysis through their irreducible representations. Across structured tasks, including the double pendulum and top quark tagging, we demonstrate that spectral sparsity reliably reveals one-parameter symmetries. These results position spectral analysis as a principled and interpretable alternative to generator-based symmetry discovery.

arXiv:2603.07339v1 Announce Type: cross Abstract: Deliberative democratic theory suggests that civic competence: the capacity to navigate disagreement, weigh competing values, and arrive at collective decisions is not innate but developed through practice. Yet opportunities to cultivate these skills remain limited, as traditional deliberative processes like citizens' assemblies reach only a small fraction of the population. We present Agora, an early-stage AI-powered platform that uses LLMs to organize authentic human voices on policy issues, helping users build consensus-finding skills by proposing and revising policy recommendations, hearing supporting and opposing perspectives, and receiving feedback on how policy changes affect predicted support. In a preliminary study with 44 university students, participants using the full interface (with access to voice explanations) reported higher levels of problem-solving skills, internal deliberation, and produced higher quality consensus statements compared to a control condition showing only aggregate support distributions. These initial findings point toward a promising direction for scaling civic education.

arXiv:2603.07368v1 Announce Type: cross Abstract: Biases in large language models (LLMs) often manifest as systematic distortions in associations between demographic attributes and professional or social roles, reinforcing harmful stereotypes across gender, ethnicity, and geography. This position paper advocates for addressing demographic and gender biases in LLMs through a dual-pronged methodology, integrating category-theoretic transformations and retrieval-augmented generation (RAG). Category theory provides a rigorous, structure-preserving mathematical framework that maps biased semantic domains to unbiased canonical forms via functors, ensuring bias elimination while preserving semantic integrity. Complementing this, RAG dynamically injects diverse, up-to-date external knowledge during inference, directly countering ingrained biases within model parameters. By combining structural debiasing through functor-based mappings and contextual grounding via RAG, we outline a comprehensive framework capable of delivering equitable and fair model outputs. Our synthesis of the current literature validates the efficacy of each approach individually, while addressing potential critiques demonstrates the robustness of this integrated strategy. Ensuring fairness in LLMs, therefore, demands both the mathematical rigor of category-theoretic transformations and the adaptability of retrieval augmentation.

arXiv:2603.07372v1 Announce Type: cross Abstract: Quality Estimation (QE) is essential for assessing machine translation quality in reference-less settings, particularly for domain-specific and low-resource language scenarios. In this paper, we investigate sentence-level QE for English to Indic machine translation across four domains (Healthcare, Legal, Tourism, and General) and five language pairs. We systematically compare zero-shot, few-shot, and guideline-anchored prompting across selected closed-weight and open-weight LLMs. Findings indicate that while closed-weight models achieve strong performance via prompting alone, prompt-only approaches remain fragile for open-weight models, especially in high-risk domains. To address this, we adopt ALOPE, a framework for LLM-based QE that uses Low-Rank Adaptation with regression heads attached to selected intermediate Transformer layers. We also extend ALOPE with recently proposed Low-Rank Multiplicative Adaptation (LoRMA). Our results show that intermediate-layer adaptation consistently improves QE performance, with gains in semantically complex domains, indicating a path toward more robust QE in practical scenarios. We release code and domain-specific QE datasets publicly to support further research.