Singapore

We present the Hierarchical AI-Meteorologist, an LLM-agent system that generates explainable weather reports using a hierarchical forecast reasoning and weather keyword generation. Unlike standard approaches that treat forecasts as flat time series, our framework performs multi-scale reasoning across hourly, 6-hour, and daily aggregations to capture both short-term dynamics and long-term trends. Its core reasoning agent converts structured meteorological inputs into coherent narratives while simultaneously extracting a few keywords effectively summarizing the dominant meteorological events. These keywords serve as semantic anchors for validating consistency, temporal coherence and factual alignment of the generated reports. Using OpenWeather and Meteostat data, we demonstrate that hierarchical context and keyword-based validation substantially improve interpretability and robustness of LLM-generated weather narratives, offering a reproducible framework for semantic evaluation of automated meteorological reporting and advancing agent-based scientific reasoning.

AAAI 2026

Hierarchical AI-Meteorologist: LLM-Agent System for Multi-Scale and Explainable Weather Forecast Reporting

hierarchical weather reasoning

weather forecast reporting

agent-based systems

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.

Rising sea levels pose an imminent hazard to coastal communities that depend on the sea for their livelihood. This paper presents a robust computational technique for forecasting tidal surge and resulting shoreline changes in the vulnerable Krishna-Godavari (KG) Delta, India. To ensure the reliability of decades-long projections and directly address the methodological shortcomings of prior work, this study is grounded in an multi-decadal historical dataset spanning 39 years (1977--2015), thus providing a scientifically justifiable temporal basis. The core methodology employs a deep-learning architecture: the Long Short-Term Memory (LSTM)network. To maximize its predictive efficiency, the LSTM's critical hyperparameters are rigorously optimized using the nature-inspired meta-heuristic : the K-means Whale Optimization Algorithm (WOA). This novel hybrid KWOA-LSTM approach achieves superior predictive performance over conventional models, demonstrated by a R2 score of 0.916 and a Mean Absolute Error (MAE) of 0.012. The predicted and forecasted surge values, computed by the optimized model, are visualized via the Google Earth Engine elevation map. Forecasts indicate significant saline intrusion and inundation will affect the KG Delta by 2060, posing a grave threat to the region's food security. The utilization of WOA for hyperparameter tuning, combined with the multi-decadal dataset, establishes a highly accurate and scientifically rigorous framework for long-term coastal hazard assessment and proactive planning.

Coastal Shoreline Dynamics in the Krishna - Godavari Delta Using K-Means Whale Optimized Long Short Term Memory Networks

Reliable detection of methane plumes in vegetated regions remains challenging due to the spectral similarity between vegetation reflectance and methane absorption features, as well as the coarse spectral resolution of freely available multispectral satellites such as Sentinel-2. To address this problem, we propose VegiSR, a vegetation-aware spectral super-resolution framework that reconstructs EMIT-like hyperspectral representations from Sentinel-2 imagery. VegiSR leverages cross-sensor spectral learning using a large-scale Sentinel-2 and EMIT paired dataset we constructed, and incorporates a vegetation-aware spectral regularization and spectral attention mechanism to suppress vegetation-induced confusion while enhancing methane-sensitive wavelengths. The reconstructed hyperspectral data is then integrated with matched filter to detect methane plumes in vegetated area where direct Sentinel-2 detection fails. Experimental results demonstrate that VegiSR achieves superior spectral reconstruction fidelity and significantly improves methane detection accuracy over
vegetated surfaces compared to U-Net and direct Sentinel-2 methods. This work highlights the potential of cross-sensor spectral super resolution for global methane monitoring using free multispectral observations.

Learning Cross-Sensor Spectral Super-Resolution for Methane Plume Detection over Vegetated Areas Using Sentinel-2 and EMIT Data

Processing wildlife imagery for conservation and management poses significant challenges, especially when limited data hinders the ability of classification models to extract sufficient features from each wildlife class. In this study, we propose the use of a 3D free-viewpoint image-to-video generative model to augment the dataset by synthesizing new images for fine-tuning classification models. Our results demonstrate a notable improvement in the F1-score, increasing from a baseline of 0.338 to 0.525. While the quality of the synthesized images can be further enhanced, particularly in terms of incorporating wildlife-specific semantics, the study highlights the potential of generative AI not only for media creation but also for advancing environmental monitoring applications, despite challenges such as high computational cost.

Diffusion-based free-viewpoint synthesis for dataset expansion and wildlife classification

Recent progress in vision language models (VLMs) has enabled remarkable perception and reasoning capabilities, yet their potential for scientific regression in Earth Observation (EO) remains largely unexplored. Existing EO datasets mainly emphasize semantic understanding tasks such as captioning or classification, lacking benchmarks that align multimodal perception with measurable biophysical variables. To fill this gap, we present REO-Instruct, the first unified benchmark designed for both descriptive and regression tasks in EO. REO-Instruct establishes a cognitively interpretable logic chain in forest ecological scenario: human activity → land-cover classification → ecological patch counting → above-ground biomass (AGB) regression, bridging qualitative understanding and quantitative prediction. The dataset integrates co-registered Sentinel-2 and ALOS-2 imagery with structured textual annotations generated and validated through a hybrid human–AI pipeline. Comprehensive evaluation protocols and baseline results across generic VLMs reveal that current models struggle with numeric reasoning, highlighting an essential challenge for scientific VLMs. REO-Instruct offers a standardized foundation for developing and assessing next-generation geospatial models capable of both description and scientific inference. The dataset will be publicly available.

Towards Unified Vision Language Models for Forest Ecological Analysis in Earth Observation

Typical domestic Immersion water heater systems are always turned on during the winter, it heats quickly rather than efficiently and ignores predictable demand windows and ambient losses. We study deadline-aware control, where the aim is to reach a target temperature at a specified time while minimising energy. We introduce an efficient Gymnasium environment that models an immersion hot-water heater with first-order thermal losses and discrete on and off actions ${0, 6000}$ W applied every 120 s. Methods include a time-optimal bang-bang baseline, a zero-shot Monte Carlo Tree Search planner, and a Proximal Policy Optimization policy. We report total energy (Wh) under identical physics. Across sweeps of initial temperature (10–30 °C), deadline (30–90 steps), and target temperature (40–80 °C), PPO achieves the most energy-efficient performance at a 60-step horizon (2 h) it uses 3.23 kWh, versus bang-bang’s 4.37–10.45 kWh and MCTS’s 4.18–6.46 kWh, yielding savings of 26\% at 30 steps and 69\% at 90 steps. In a representative trajectory (50 kg, 20 °C ambient, 60 °C target), PPO consumes 54\% less energy than bang-bang and 33\% less than MCTS. These results show that learned, deadline-aware control reduces energy under identical physics where planners provide partial savings without training, while policies offer near-zero-cost inference once trained.

Deadline-Aware, Energy-Efficient Control of Domestic Immersion Hot Water Heater

Infrared and visible object detection (IVOD) is essential for numerous around-the-clock applications. Despite notable advancements, current IVOD models exhibit notable performance declines when confronted with incomplete modality data, particularly if the dominant modality is missing. In this paper, we take a thorough investigation on modality incomplete IVOD problem from an architecture compatibility perspective. Specifically, we propose a plug-and-play Scarf Neck module for DETR variants, which introduces a modalityagnostic deformable attention mechanism to enable the IVOD detector to flexibly adapt to any single or double modalities during training and inference. When training Scarf-DETR, we design a pseudo modality dropout strategy to fully utilize the multi-modality information, making the detector compatible and robust to both working modes of single and double modalities. Moreover, we introduce a comprehensive benchmark for the modality-incomplete IVOD task aimed at thoroughly assessing situations where the absent modality is either dominant or secondary. Our proposed Scarf-DETR not only performs excellently in missing modality scenarios but also achieves superior performances on the standard IVOD modality complete benchmarks. We will release our code.

On Modality Incomplete Infrared-Visible Object Detection: An Architecture Compatibility Perspective

Semi-supervised learning (SSL) has emerged as a critical paradigm for medical image segmentation, mitigating the immense cost of dense annotations. However, prevailing SSL frameworks are fundamentally “inward-looking”, recycling information and biases solely from within the target dataset. This design triggers a vicious cycle of confirmation bias under class imbalance, leading to the catastrophic failure to recognize minority classes. To dismantle this systemic issue, we propose a paradigm shift to a multi-level “outward-looking” framework. Our primary innovation is Foundational Knowledge Distillation (FKD), which looks outward beyond the confines of medical imaging by introducing a pre-trained visual foundation model, DINOv3, as an unbiased external semantic teacher. Instead of trusting the student’s biased high confidence, our method distills knowledge from DINOv3’s robust understanding of high semantic uniqueness, providing a stable, cross-domain supervisory signal that anchors the learning of minority classes. To complement this core strategy, we further look outward within the data by proposing Progressive Imbalance-aware CutMix (PIC), which creates a dynamic curriculum that adaptively forces the model to focus on minority classes in both labeled and unlabeled subsets. This layered strategy forms our framework, DINO-Mix, which breaks the vicious cycle of bias and achieves remarkable performance on challenging semi-supervised class-imbalanced medical image segmentation benchmarks Synapse and AMOS.

DINO-Mix: Distilling Foundational Knowledge with Cross-Domain CutMix for Semi-supervised Class-imbalanced Medical Image Segmentation

Membership inference attacks (MIAs) test whether a data point was part of a model's training set, posing serious privacy risks. Existing methods often depend on shadow models or heavy query access, which limits their practicality. We propose GP-MIA, an efficient and interpretable approach based on Gaussian process (GP) meta-modeling. Using post-hoc metrics such as accuracy, entropy, dataset statistics, and optional sensitivity features (e.g. gradients, NTK measures) from a single trained model, GP-MIA trains a GP classifier to distinguish members from non-members while providing calibrated uncertainty estimates. Experiments on synthetic data, real-world fraud detection data, CIFAR-10, and WikiText-2 show that GP-MIA achieves high accuracy and generalizability, offering a practical alternative to existing MIAs.

Training data membership inference via Gaussian process meta-modeling: a post-hoc analysis approach

Recursive Self-Improvement (RSI) is an effective approach to alleviating data scarcity. This paper explores the application of RSI in text-to-image diffusion models, addressing the challenge of training collapse caused by biased synthetic data. We identify two key factors contributing to this collapse: the lack of perceptual alignment and the accumulation of generative hallucinations. To mitigate these issues, we propose three strategies: (1) a prompt construction and filtering pipeline designed to facilitate the generation of perceptual aligned data, (2) a preference sampling method to identify human-preferred samples and filter out generative hallucinations, and (3) a distribution-based weighting scheme to penalize selected samples with hallucinatory errors. Our extensive experiments validate the effectiveness of these approaches.

Generating on Generated: An Approach Towards Self-Evolving Diffusion Models

Machine learning holds great promise for advancing the field of medicine, with electronic health records (EHRs) serving as a primary data source. However, EHRs are often sparse and contain missing data due to various challenges and limitations in data collection and sharing between healthcare providers. Existing techniques for imputing missing values predominantly focus on known unknowns, such as missing or unavailable values of lab test results; most do not explicitly address situations where it is difficult to distinguish what is missing. For instance, a missing diagnosis code in an EHR could signify either that the patient has not been diagnosed with the condition or that a diagnosis was made, but not shared by a provider. Such situations fall into the paradigm of unknown unknowns. To address this challenge, we develop a general purpose algorithm for denoising data to recover unknown missing values in binary EHRs. We design a transformer-based denoising neural network where the output is thresholded adaptively to recover values in cases where we predict data are missing. Our method demonstrates improved accuracy in denoising medical codes within a real EHR dataset compared to existing imputation approaches and leads to increased performance on downstream tasks using the denoised data. In particular, when applying our method to a real world application, predicting hospital readmission from EHRs, our method achieves statistically significant improvement over all existing baselines.

Downloads

Next from AAAI 2026

Coastal Shoreline Dynamics in the Krishna - Godavari Delta Using K-Means Whale Optimized Long Short Term Memory Networks

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES