Singapore

Financial AI systems suffer from a critical blind spot: while Retrieval-Augmented Generation (RAG) excels at finding relevant documents, language models still generate calculation errors and regulatory violations during reasoning, even with perfect retrieval. This paper introduces VERAFI (Verified Agentic Financial Intelligence), an agentic framework with neurosymbolic policy generation for verified financial intelligence. VERAFI combines state-of-the-art dense retrieval and cross-encoder reranking with financial tool-enabled agents and automated reasoning policies covering GAAP compliance, SEC requirements, and mathematical validation. Our comprehensive evaluation on FinanceBench demonstrates remarkable improvements: while traditional dense retrieval with reranking achieves only 52.4% factual correctness, VERAFI&#39;s integrated approach reaches 94.7%, an 81% relative improvement. The neurosymbolic policy layer alone contributes a 4.3 percentage point gain over pure agentic processing, specifically targeting persistent mathematical and logical errors. By embedding financial domain expertise directly into the reasoning process, VERAFI offers a practical pathway toward trustworthy financial AI that meets the stringent accuracy demands of regulatory compliance, investment decisions, and risk management.

AAAI 2026

VERAFI: Verified Agentic Financial Intelligence through Neurosymbolic Policy Generation

nlp: question answering

nlp: (large) language models

ml: neuro-symbolic learning

### Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Welcome to the 2025 meeting of the Nations of the Americas Chapter of the Association for Computational Linguistics! I am proud to help organize the first NAACL conference to carry the new name of our organization, one that emphasizes inclusion for all of the Americas. I am also pleased to welcome you to Albuquerque, New Mexico, a state whose unique blend of cultural influences will make for an excellent backdrop for NAACL 2025, especially with this year’s special theme on NLP in a Multicultural World. 
**[Continue reading...](https://drive.google.com/file/d/1jX-qGhqVSZZCIrAnJaz798pflu5Irrdn/view?usp=sharing)**

*- Colin Cherry, Google, NAACL 2025 General Chair* 

[![](https://assets.underline.io/markdown_image/1/image/b087f8a4dc5816d6e1a6514e59c59ac3.png)](https://drive.google.com/file/d/1T96GzPqObXrMln2BMByCSXTSizjTg69P/view?usp=sharing)

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

To access NAACL 2025 event page you are required to register. Please follow [**this link**](https://2025.naacl.org/registration/registration/) to register. Access will depend on your registration type.

Please register!

NAACL 2025

Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Financial AI systems suffer from a critical blind spot: while Retrieval-Augmented Generation (RAG) excels at finding relevant documents, language models still generate calculation errors and regulatory violations during reasoning, even with perfect retrieval. This paper introduces VERAFI (Verified Agentic Financial Intelligence), an agentic framework with neurosymbolic policy generation for verified financial intelligence. VERAFI combines state-of-the-art dense retrieval and cross-encoder reranking with financial tool-enabled agents and automated reasoning policies covering GAAP compliance, SEC requirements, and mathematical validation. Our comprehensive evaluation on FinanceBench demonstrates remarkable improvements: while traditional dense retrieval with reranking achieves only 52.4% factual correctness, VERAFI's integrated approach reaches 94.7%, an 81% relative improvement. The neurosymbolic policy layer alone contributes a 4.3 percentage point gain over pure agentic processing, specifically targeting persistent mathematical and logical errors. By embedding financial domain expertise directly into the reasoning process, VERAFI offers a practical pathway toward trustworthy financial AI that meets the stringent accuracy demands of regulatory compliance, investment decisions, and risk management.

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.

To access this event page, you need to log in with the **email address you registered with**. 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.

This paper presents a first empirical study of agentic AI as autonomous decision-makers in decentralized governance. Using more than 3K proposals from major protocols, we build an agentic AI voter that interprets proposal contexts, retrieves historical deliberation data, and independently determines its voting position. The agent operates within a realistic financial simulation environment grounded in verifiable blockchain data, implemented through a modular composable program (MCP) workflow that defines data flow and tool usage via Agentics framework. We evaluate how closely the agent’s decisions align with the human and token-weighted outcomes, uncovering strong alignments measured by carefully designed evaluation metrics. Our findings demonstrate that agentic AI can augment collective decision-making by producing interpretable, auditable, and empirically grounded signals in realistic DAO governance settings. The study contributes to the design of explainable and economically rigorous AI agents for decentralized financial systems.

DAO-AI: Evaluating Collective Decision-Making through Agentic AI in Decentralized Governance

Multi-agent systems have demonstrated the ability to improve performance on a variety of predictive tasks by leveraging collaborative decision making. However, the lack of effective evaluation methodologies has made it difficult to estimate the risk of bias, making deployment of such systems unsafe in high stakes domains such as consumer finance, where biased decisions can translate directly into regulatory breaches and financial loss. To address this challenge, we need to develop fairness evaluation methodologies for multi-agent predictive systems and measure the fairness characteristics of these systems in the financial tabular domain. Examining fairness metrics using large-scale simulations across diverse multi-agent configurations, with varying communication and collaboration mechanisms, we reveal patterns of emergent bias in financial decision-making that cannot be traced to individual agent components, indicating that multi-agent systems may exhibit genuinely collective behaviors. Our findings highlight that fairness risks in financial multi-agent systems represent a significant component of model risk, with tangible impacts on tasks such as credit scoring and income estimation. We advocate that multi-agent decision systems must be evaluated as holistic entities rather than through reductionist analyses of their constituent components.

Emergent Bias and Fairness in Multi-Agent Decision Systems

Retrieval-Augmented Generation (RAG) is becoming increasingly essential for Question Answering (QA) in the financial sector, where accurate and contextually grounded insights from complex public disclosures are crucial. However, existing financial RAG systems face two significant challenges: (1) they struggle to process heterogeneous data formats, such as text, tables, and figures; and (2) they encounter difficulties in balancing general-domain applicability with company-specific adaptation. To overcome these challenges, we present VeritasFi, an innovative hybrid RAG framework that incorporates a multi-modal preprocessing pipeline alongside a cutting-edge two-stage training strategy for its re-ranking component. VeritasFi enhances financial QA through three key innovations: (1) A multi-modal preprocessing pipeline that seamlessly transforms heterogeneous data into a coherent, machine-readable format. (2) A tripartite hybrid retrieval engine that operates in parallel, combining deep multi-path retrieval over a semantically indexed document corpus, real-time data acquisition through tool utilization, and an expert-curated memory bank for high-frequency questions, ensuring comprehensive scope, accuracy, and efficiency. (3) A two-stage training strategy for the document re-ranker, which initially constructs a general, domain-specific model using anonymized data, followed by rapid fine-tuning on company-specific data for targeted applications. By integrating our proposed designs, VeritasFi presents a groundbreaking framework that greatly enhances the adaptability and robustness of financial RAG systems, providing a scalable solution for both general-domain and company-specific QA tasks. In additional, VeritasFi has been deployed as an online meeting agent, for which it has served more than 2000 people.

VeritasFi: An Adaptable, Multi-tiered RAG Framework for Multi-modal Financial Question Answering

Automated Question Generation (AQG) plays a crucial role in enabling scalable assessment and personalized learning, serving as a cornerstone of modern Intelligent Tutoring Systems (ITS). However, existing AQG systems, particularly those built upon Large Language Models (LLMs), face a significant cognitive ceiling in generating high-stakes, exam-ready content for specialized domains like finance (e.g., CFA Exams). They often generate linguistically fluent but cognitively shallow items, failing to ensure difficulty calibration, distractor plausibility, computational correctness, and multi-concept integration. To address these challenges, we propose Fin-Agent-QG, a cognitive-collaborative multi-agent framework that emulates the workflow of professional financial exam committees. Fin-Agent-QG coordinates specialized agents for different cognitive phases of question construction: a Teacher–Critic loop for conceptual and difficulty alignment, a Quantitative Analyst Agent utilizing Code-as-Chain-of-Thought for computational reasoning, and a Knowledge Graph Construction Agent with a Case Strategist for multi-concept integration via a Financial Knowledge Graph. Empirical evaluations by both LLM metrics and human experts demonstrate that Fin-Agent-QG produces exam-ready financial questions with greater cognitive depth, accuracy, and conceptual coherence than traditional AQG systems. This work bridges the gap between automated generation and expert-level financial assessment design.

Fin-Agent-QG: A Collaborative Multi-Agent Framework for Higher-Order Cognitive and Difficulty-Aware Financial Exam Question Generation

We present ArgBoost, a transparent multi-agent forecasting framework for directional cryptocurrency price prediction that integrates structured LLM-based reasoning with interpretable machine learning. ArgBoost classifies next-day VWAP (Volume Weighted Average Price) movements for Ethereum by aggregating outputs from specialized LLM agents operating across distinct data domains. The agent-generated claims are organized into an argumentation graph and analyzed using an Explainable Boosting Machine (EBM) trained on rolling windows. Using daily Ethereum data from January 2024 to April 2025, ArgBoost outperforms Monte Carlo baselines and traditional technical strategies in both predictive and trading performance. The model demonstrates particular strength in tail market conditions, accurately detecting sharp upswings and downturns. Feature importance analysis shows that interaction terms involving coordinated relationships among agents and contextual market conditions jointly enhance predictability, offering a transparent and modular approach to evidence-driven decision support in volatile financial markets.

ArgBoost: Interpretable Multi-Agent Reasoning for Directional Cryptocurrency Price Forecasting

Rice is one of the most widely cultivated crops globally and has been developed into numerous varieties. The quality of rice during cultivation is primarily determined by its cultivar and characteristics. Traditionally, rice classification and quality assessment rely on manual visual inspection, a process that is both time-consuming and prone to errors. However, with advancements in machine vision technology, automating rice classification and quality evaluation based on its cultivar and characteristics has become increasingly feasible, enhancing both accuracy and efficiency. This study proposes a real-time evaluation mechanism for comprehensive rice grain assessment, integrating a one-stage object detection approach, a deep convolutional neural network, and traditional machine learning techniques. The proposed framework enables rice variety identification, grain completeness grading, and grain chalkiness evaluation. The rice grain dataset used in this study comprises approximately 20,000 images from six widely cultivated rice varieties in China. Experimental results demonstrate that the proposed mechanism achieves a mean Average Precision (mAP) of 99.14\% in the object detection task and an accuracy of 97.89\% in the classification task. Furthermore, the framework attains an average accuracy of 97.56\% in grain completeness grading within the same rice variety, contributing to an effective quality evaluation system.

An Overall Real-Time Mechanism for Classification and Quality Evaluation of Rice

Pig counting is an important task in pig marketing and farming. Most of the existing methods use supervised object detection for counting. However, supervised pig counting relies heavily on expensive data labeling, especially in the dense counting scenario. To address this issue, we propose BU-CLIP, which adopts the Contrastive Language Image Pre-training model (CLIP) for unsupervised pig counting. We tailor the image encoder of CLIP and the loss function to make it more suitable for pig counting. Our method replaces CLIP's image encoder with a bilinear model combining ConvNeXt and ResNet50 backbones, while the pooling operation is performed via a multi-head attention module. We reconstruct the loss function by using a multi-modal full-ranking loss, which captures the intrinsic correspondence between text and image. The proposed model is tested on a dense pig counting dataset, and extensive experiments demonstrate that our method outperforms unsupervised state-of-the-art counting methods and achieves almost the same results as supervised state-of-the-art counting methods.

Bilinear Magic Meets CLIP: Unsupervised Learning in Pig Counting Scenario

High-throughput plant phenotyping is an emerging interdisciplinary field of research that lies at the intersection of computer vision, plant science, artificial intelligence, data analytics and visualization, and genomics. To develop novel algorithms that advance image-based high-throughput plant phenotyping research with an aim of precision farming and global food security, the publication of benchmark datasets is indispensable. Thus, this paper introduces ReproPheno, a large-scale, open-source dataset comprising multi-view image sequences of plants spanning their entire life cycle captured using three modalities of cameras, i.e., visible light, fluorescent, and hyperspectral, within the LemnaTec Scanalyzer 3D high-throughput plant phenotyping facility. By enabling comprehensive analysis across multiple sensing modalities and temporal scales, this dataset opens the door to several novel research directions in computer vision, including multimodal co-segmentation, hyperspectral dataset distillation, and 3D model reconstruction of living organisms with growing architectural complexity. Furthermore, the paper presents ReproPhenoNet, a novel algorithm that uses You Only Look Once (YOLO) deep neural-network-based object detector to detect flowers and fruits from visible light and hyperspectral image sequences, respectively. These detections form the foundation for the quantitative computation of reproductive-stage plant phenotypes.

ReproPheno and ReproPhenoNet: A Large-Scale Multimodal Benchmark Dataset and Deep Learning Framework for Reproductive-Stage Plant Phenotyping

Timely detection of plant diseases in greenhouse environments is essential for minimizing crop loss, reducing chemical use, and supporting sustainable production. This paper presents an AI-enabled, reconfigurable edge device designed for leaf-level disease assessment using multimodal sensing and embedded intelligence. A 3D-printed handheld prototype was developed with a custom PCB integrating VIS/NIR spectral sensors, a thermal module, and an RGB camera for noninvasive measurement of physiological stress. A lightweight deep learning model combined with handcrafted spectral indices performs fully on-device analysis to detect early indicators such as abnormal thermal patterns, pigment loss, and reflectance changes linked to infections. The system is built on a low-power Raspberry Pi platform and supports model reconfiguration for rapid adaptation to different crops and disease types. Real-time inference is achieved through a segmentation-based pipeline deployed on both an Android app and an Ubuntu graphical interface. Validation trials demonstrated early disease detection with previously collected images and an average latency of 2.52 seconds on Android and 2.19 seconds on Ubuntu. We are currently calibrating our framework in the greenhouse environment for real-time assessments. A web dashboard provides live visualization of thermal overlays, spectral response, and health scores. By combining custom hardware, multimodal sensing, and on-device AI, the system offers a portable and scalable solution for early disease detection in greenhouse crops.

AI-Enabled Reconfigurable Edge Device for Plant Health Assessment in Greenhouse Environment

Weed detection is a critical task in precision agriculture, as weeds significantly reduce crop yields and increase production costs. This work presents an efficient, low-cost, and low-power UAV-deployable weed detection pipeline that relies solely on consumer grade RGB cameras. We propose novel lightweight U-Net and Attention U-Net architectures optimized for real-time semantic segmentation on an edge device. To enhance segmentation accuracy, we integrate additional RGB derived features. Experiments on the CoFly-WeedDB detection dataset demonstrate that both architectures perform effectively on RGB imagery, with further improvements when incorporating hue and edge detection features. The proposed lightweight U-Net architectures, made more efficient through quantization, achieves IoU scores above 50% and Dice scores exceeding 60% on the CoFly-WeedDB dataset across multiple augmentation levels. These findings highlight the practicality of deploying lightweight deep learning models for precision weed detection in resource-constrained agricultural environments.

Next from AAAI 2026

DAO-AI: Evaluating Collective Decision-Making through Agentic AI in Decentralized Governance

Similar lecture

FinEval: A Chinese Financial Domain Knowledge Evaluation Benchmark for Large Language Models