Singapore

As LLM-based agents grow more autonomous and multi- modal, ensuring they remain controllable, auditable, and faithful to deployer intent becomes critical. Prior benchmarks measured propensity for misaligned behavior and showed that agent personalities and tool access significantly influ- ence misalignment. Building on those insights, we propose a Verifiability-First architecture that (1) integrates run-time attestations of agent actions (cryptographic &amp; symbolic), (2) embeds lightweight Audit Agents that continuously verify in- tent vs. behavior using constrained reasoning, and (3) en- forces challenge–response attestation protocols for high-risk operations. We introduce OPERA (Observability, Provable Execution, Red-team, Attestation), a benchmark suite and evaluation protocol designed to measure (i) detectability of misalignment, (ii) time-to-detection under stealthy strategies, and (iii) resilience of verifiability mechanisms to adversar- ial prompt/persona injection. Our approach aims to shift the evaluation focus from ”how likely misalignment is” to ”how quickly and reliably misalignment can be detected and reme- diated.”

AAAI 2026

Verifiability-First Agents: Provable Observability and Lightweight Audit Agents for Controlling Autonomous LLM Systems

As LLM-based agents grow more autonomous and multi- modal, ensuring they remain controllable, auditable, and faithful to deployer intent becomes critical. Prior benchmarks measured propensity for misaligned behavior and showed that agent personalities and tool access significantly influ- ence misalignment. Building on those insights, we propose a Verifiability-First architecture that (1) integrates run-time attestations of agent actions (cryptographic & symbolic), (2) embeds lightweight Audit Agents that continuously verify in- tent vs. behavior using constrained reasoning, and (3) en- forces challenge–response attestation protocols for high-risk operations. We introduce OPERA (Observability, Provable Execution, Red-team, Attestation), a benchmark suite and evaluation protocol designed to measure (i) detectability of misalignment, (ii) time-to-detection under stealthy strategies, and (iii) resilience of verifiability mechanisms to adversar- ial prompt/persona injection. Our approach aims to shift the evaluation focus from ”how likely misalignment is” to ”how quickly and reliably misalignment can be detected and reme- diated.”

workshop paper

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-26 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.<br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to AAAI 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://aaai.org/conference/aaai/aaai-26/registration/) first.

**Online Registration Form**: https://aaai.getregistered.net/conference-2026 

Registration Required

We are pleased to announce the Fortieth AAAI Conference on Artificial Intelligence (AAAI-26), which will be held in Singapore EXPO from January 20 to January 27, 2026.

Text-to-image diffusion models help visualize urban futures but can amplify group-level harms. We propose collective recourse: structured community "visual bug reports" that trigger fixes to models and planning workflows. We (1) formalize collective recourse and a practical pipeline (report, triage, fix, verify, closure); (2) situate four recourse primitives—counter-prompts, negative prompts, dataset edits, reward-model tweaks—within the diffusion stack; (3) define mandate thresholds via a mandate score combining severity, volume saturation, representativeness, and evidence; and (4) evaluate a synthetic program of 240 reports. Prompt-level fixes were fastest (median $2.1$-$3.4$ days) but less durable (21-38\% recurrence); dataset edits and reward tweaks were slower (13.5 and 21.9 days) yet more durable (12-18\% recurrence) with higher planner uptake (30-36\%). A threshold of $0.12$ yielded 93\% precision and 75\% recall; increasing representativeness raised recall to 81\% with little precision loss. We discuss integration with participatory governance, risks (overfitting to vocal groups), and safeguards (dashboards, rotating juries).

Collective Recourse for Generative Urban Visualizations

Urban digital twins represent a paradigm shift in city modeling, yet current data-driven approaches operate as black boxes, fundamentally incapable of reasoning about the regulatory constraints and policy frameworks that govern urban development. We introduce Neuro-Symbolic Urban Twins (NS-UT), a novel architecture that integrates neural perception with symbolic reasoning to enable compliant, interpretable urban simulation. Our framework features: (1) multimodal encoders for satellite imagery, IoT sensor data, and demographic information; (2) a symbolic knowledge base encoding zoning codes, building regulations, and sustainability policies; (3) an LLM-based semantic parser that automatically translates natural language regulations into executable logical constraints; and (4) a differentiable neuro-symbolic interface enabling joint optimization. We evaluate NS-UT on a newly curated Urban Regulatory Compliance (URC) benchmark comprising 15,000 development scenarios across five major U.S. cities. NS-UT achieves state-of-the-art performance with 94.3% compliance prediction accuracy (p < 0.001 vs. baselines) while providing human-interpretable explanations for violations and recommendations. This work establishes a new foundation for policy-aware urban AI that respects the complex rule systems shaping our cities.

Neuro-Symbolic Urban Twins: Embedding Reasoning in Data-Driven City Models

Ensuring safe and inclusive mobility for vulnerable older adults is an emerging priority in urban planning. However, existing data sources such as surveys or GIS-based audits provide limited insight into how micro-scale built environment (BE) features influence real-world behavior and perception. This study presents a novel multimodal data-fusion approach that integrates wearable and environmental sensing to dynamically represent human-environment interactions and quantify the BE impacts on mobility among vulnerable older adults, specifically, those with knee osteoarthritis or a history of falls. Data collected during naturalistic walking sessions in Singapore, are used to demonstrate this framework of synchronized streams from eye tracking, kinematic sensors, physiological monitors, GPS, and video recordings. Preliminary results show how AI-driven data fusion can uncover behaviorally and perceptually significant urban segments, providing a basis for actionable insights in inclusive design. This human-centered analytical approach advances the representation of urban environments from the perspective of vulnerable pedestrians, establishing a foundation for evidence-based, age-friendly city planning.

Multimodal Data Fusion to Capture Dynamic Interactions between Built Environment and Vulnerable Older Adults

AI has proven highly successful at urban planning analysis---learning patterns from data to predict future conditions. The next frontier is AI-assisted decision-making: agents that recommend sites, allocate resources, and evaluate trade-offs while reasoning transparently about constraints and stakeholder values. Recent breakthroughs in reasoning AI---CoT prompting, ReAct, and multi-agent collaboration frameworks---now make this vision achievable. This position paper presents the Agentic Urban Planning AI Framework for reasoning-capable planning agents that integrates three cognitive layers (Perception, Foundation, Reasoning) with six logic components (Analysis, Generation, Verification, Evaluation, Collaboration, Decision) through a multi-agents collaboration framework. We demonstrate why planning decisions require explicit reasoning capabilities that are value-based (applying normative principles), rule-grounded (guaranteeing constraint satisfaction), and explainable (generating transparent justifications)---requirements that statistical learning alone cannot fulfill. We compare reasoning agents with statistical learning, present a comprehensive architecture with benchmark evaluation metrics, and outline critical research challenges. This framework shows how AI agents can augment human planners by systematically exploring solution spaces, verifying regulatory compliance, and deliberating over trade-offs transparently---not replacing human judgment but amplifying it with computational reasoning capabilities.

Reasoning Is All You Need for Urban Planning AI

Project Sidewalk is a web-based platform that enables crowdsourcing accessibility of sidewalks at city-scale by virtually walking through city streets using Google Street View. The tool has been used in 40 cities across the world, including the US, Mexico, Chile, and Europe. In this paper, we describe adaptation efforts to enable deployment in Chandigarh, India, including modifying annotation types, provided examples, and integrating VLM-based mission guidance, which adapts instructions based on a street scene and metadata analysis. Our evaluation with 3 annotators indicates the utility of AI-mission guidance with an average score of 4.66. Using this adapted Project Sidewalk tool, we conduct a Points of Interest (POI)-centric accessibility analysis for three sectors in Chandigarh with very different land uses---residential, commercial and institutional covering about 40 km of sidewalks. Across ~40 km of roads audited in three sectors and around 230 POIs, we identified 1,644 of 2,913 locations where infrastructure improvements could enhance accessibility.

Towards Human-AI Accessibility Mapping in India: VLM-Guided Annotations and POI-Centric Analysis in Chandigarh

In many real-world sequential predictions, such as forecasting taxi zone profitability, uncertainty quantification—which is often overlooked in traditional prediction models—is essential for trustworthy decision-making by taxi drivers. Traditional conformal prediction (CP) offers a distribution-free, model-agnostic approach to construct prediction intervals, but its assumption that data points are identically and independently distributed (i.i.d.) fails in temporal data. This work applies sequential CP to quantify uncertainty in time-series profitability predictions for taxi zones, marking the first such application in this context. Feedback from prediction residuals is utilized to ensure user-specified statistical coverage while minimizing interval widths. Based on the empirical Hong Kong taxi trip and trajectory data from February to July 2025, we evaluate state-of-the-art time-series models under split and sequential CP variants. Results demonstrate superior empirical coverage compared to baselines, highlighting the advantage of utilizing residuals for sequential uncertainty quantification. This advances uncertainty-aware forecasting and bridges theoretical guarantees with practical mobility applications.

Enhancing Trust in Taxi Zone Profit Predictions with Sequential Conformal Inference

Human mobility prediction is a critical task but remains challenging due to its complexity and variability across populations and regions. Recently, large language models (LLMs) have made progress in zero-shot prediction, but existing methods suffer from limited interpretability (due to black-box reasoning), lack of iterative learning from new data, and poor transferability. In this paper, we introduce \textbf{ARMove}, a fully transferable framework for predicting human mobility through agentic reasoning. To address these limitations, ARMove employs standardized feature management with iterative optimization and user-specific customization: four major feature pools for foundational knowledge, user profiles for segmentation, and an automated generation mechanism integrating LLM knowledge. Robust generalization is achieved via agentic decision-making that adjusts feature weights to maximize accuracy while providing interpretable decision paths. Finally, large-small model synergy distills strategies from large LLMs (e.g., 72B) to smaller ones (e.g., 7B), reducing costs and enhancing performance ceilings. Extensive experiments on four global datasets show ARMove outperforms state-of-the-art baselines on 6 out of 12 metrics (gains of 0.78\% to 10.47\%), with transferability tests confirming robustness across regions, users, and scales. The other 4 items also achieved suboptimal results. Transferability tests confirm its 19 robustness across regions, user groups, and model scales, while interpretability 20 analysis highlights its transparency in decision-making. Our codes are available at: \url{https://anonymous.4open.science/r/ARMove-F847}.

ARMove: Learning to Predict Human Mobility through Agentic Reasoning

Deep learning models have shown promise in reservoir inflow prediction, yet their performance often deteriorates when applied to different reservoirs due to distributional differences, referred to as the domain shift problem. Domain generalization (DG) solutions aim to address this issue by extracting domain-invariant representations that mitigate errors in unseen domains. However, in hydrological settings, each reservoir exhibits unique inflow patterns, while some metadata beyond observations like spatial information exerts indirect but significant influence. This mismatch limits the applicability of conventional DG techniques to many-domain hydrological systems. To overcome these challenges, we propose HydroDCM, a scalable DG framework for cross-reservoir inflow forecasting. Spatial metadata of reservoirs is used to construct pseudo-domain labels that guide adversarial learning of invariant temporal features. During inference, HydroDCM adapts these features through light-weight conditioning layers informed by the target reservoir’s metadata, reconciling DG’s invariance with location-specific adaptation. Experiment results on 30 real-world reservoirs in the Upper Colorado River Basin demonstrate that our method substantially outperforms state-of-the-art DG baselines under many-domain conditions and remains computationally efficient.

HydroDCM: Hydrological Domain-Conditioned Modulation for Cross-Reservoir Inflow Prediction

Landmine contamination poses a significant threat in post-conflict regions, endangering civilian populations even decades after hostilities end. Demining efforts are resource-intensive, and prioritizing areas for clearance is particularly challenging in regions with widespread contamination. This paper presents a country-wide Digital Twin for landmine risk assessment, addressing key challenges: (a) modeling of a country-wide Digital Twin, (b) data integration, enrichment and augmentation, and (c) efficient machine learning (ML) training. The system is deployed in Cambodia premises, where it streamlines the development of ML workflows and achieves up to 91% accuracy in assessing risks of landmine areas.

Country-Wide Digital Twin for Humanitarian AI: The Case Study of Cambodian Mine Action

Financial risk assessment plays a vital role in accelerating the global energy transition and achieving the 1.5°C climate target. This study develops a three-stage machine learning framework to analyze the impact of fossil fuel price volatility on the weighted average cost of capital (WACC) for energy transition projects. First, a hybrid Long Short-Term Memory-Gaussian Regression Process (LSTM-GRP) model forecasts fossil fuel prices, followed by an LSTM-based inflation predictor. In the final stage, an AutoML (AutoGluon) regression model evaluates how variations in fossil fuel prices and inflation influence WACC across nine energy sectors and multiple global regions. 

The results show that the WACC in energy-importing regions like Europe (R²=0.67) and China(R²=0.64) is highly sensitive to price shocks, making their decarbonization pathways vulnerable to external market instability. Conversely, the energy-exporting countries in Middle East (R²=0.54) exhibit lower sensitivity, indicating a fiscal cushion that may reduce the economic impetus for transition. The findings demonstrate how energy price volatility shapes financing conditions and investment risks for renewable energy deployment. This approach integrates AI-driven forecasting with climate finance modeling, providing an interpretable and data-driven tool to support policy design, green investment strategies, and climate-resilient energy planning.

Premium content

Next from AAAI 2026

Collective Recourse for Generative Urban Visualizations

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES