1
ARMOR: Stabilizing On-Policy LLM RL with Off-Policy Anchor Samples
The ARMOR framework introduces off-policy anchor samples to stabilize reinforcement learning in large language models and prevent over-optimization.
1
Pipette: An Embodied Simulation Platform, Benchmark, and Data-Efficient Augmentation Framework for Wet-Lab Robotics
Pipette is a new simulation platform and benchmark created to improve the training and data efficiency of robotics systems used in biomedical laboratory environments.
1
When Does Restricting a Coding Agent to execute_code Help? A Regime $\times$ Agent-Design Ablation
This study performs a comparative analysis of different code-execution environments to determine their impact on the performance of AI coding agents.
1
Flout at Your Own Risk: LLMs Struggle with Pragmatic Cooperativity Under Epistemic Asymmetry
The study investigates the limitations of LLMs in collaborative settings, specifically their difficulty in managing pragmatic communication when information is asymmetrically distributed.
1
Explaining Process Control Optimisation Recommendations via GradientSHAP and Implicit Differentiation
Researchers propose a method to improve the interpretability of automated industrial process control recommendations using gradient-based explanation techniques.
1
Transformer-Guided Swarm Intelligence for Frugal Neural Architecture Search
A new neural architecture search framework utilizes swarm intelligence and transformer controllers to enable efficient model design on consumer-grade hardware.
1
Mini Amusement Parks (MAPs): A Testbed for Modelling Business Decisions
Researchers have developed a new simulation testbed designed to evaluate how AI models handle complex, long-horizon business decision-making.
1
Agentic generation of verifiable rules for deterministic, self-expanding reaction classification
A new multi-agent framework automates the generation of verifiable rules for classifying chemical reactions to improve computer-assisted synthesis planning.
1
MG$^2$-RAG: Multi-Granularity Graph for Multimodal Retrieval-Augmented Generation
MG2-RAG is a proposed multimodal retrieval-augmented generation framework that uses multi-granularity graphs to improve reasoning and preserve fine-grained visual data.
1
Towards a Bridge Layer Between Bibliographic and Formalized Mathematical Knowledge
Researchers have developed a relational database bridge to connect bibliographic metadata with formalized mathematical proof libraries.
1
Multi-Scale ViT Inference with Habitat-Fit Priors and kNN Retrieval for Multi-Species Plant Identification
A multi-scale vision transformer approach is detailed for identifying multiple plant species in high-resolution photographs using limited training data.
1
Automated identification of Ichneumonoidea wasps via YOLO-based deep learning: Integrating HiresCam for Explainable AI
Researchers have developed a deep learning framework using YOLO and explainable AI techniques to automate the taxonomic identification of parasitoid wasps.
1
Gaussian Process Aggregation for Root-Parallel Monte Carlo Tree Search with Continuous Actions
A new method using Gaussian Process Regression has been proposed to optimize value estimation in Monte Carlo Tree Search for continuous action spaces.
1
A short review on the maximum clique problem algorithms with classical, AI, and quantum methods
This review paper examines the evolution of algorithms for the maximum clique problem, comparing classical, AI-based, and quantum computing approaches.
1
WrAFT: a Modularized Automated Writing Evaluation System for Argumentative Essays
WrAFT is a modular automated writing evaluation system designed to provide scoring and feedback for argumentative essays using various large language models.
1
CrimeNER Demo: Named-Entity Recognition in the Crime Domain
A new platform called CrimeNER Demo has been introduced to facilitate named-entity recognition and classification of crime-related information in documents.
1
When AI Blurs the Boundaries of Contribution: An Empirical Study of Authorship Calibration
This study examines how users perceive their own authorship when collaborating with generative AI tools.
1
Man, Machine, and Masterpiece: Artistic Ownership in the AI Era
Researchers developed a tool to quantify human versus AI contributions in creative works to address ongoing debates regarding artistic ownership.
1
Concept-Guided Spatial Regularization for World Models in Atari Pong
Researchers evaluated the performance of five prominent world-model agents in Atari Pong to better understand their isolated capabilities.
1
Fast-Fading Channel and Power Optimization of the Magnetic Inductive Cellular Network
This research proposes a model for optimizing power and channel performance in magnetic inductive cellular networks used in underground environments.
1
Heterogeneous Element-Aware Cross-Version Differencing of Scientific Documents via Layout-Aware Alignment and Structure-Aware Reasoning
A new approach for comparing scientific document versions integrates layout-aware alignment and structural reasoning to handle complex elements like tables and formulas.
1
Team RAS in 11th ABAW Competition: Multimodal Ambivalence Recognition Approach
Researchers developed a text-centered multimodal system designed to recognize human ambivalence and hesitancy in video data.
1
Knowing You at First Glance: Inferring Apparent Personality from Faces
A new study explores methods for AI agents to infer personality traits from facial images to improve human-robot interaction.
1
HABIB_TAZ at SemEval-2026 Task 11: Disentangling Formal Logic from Content via Synthetic Training and Multi-Objective Optimization
The HABIB_TAZ system uses synthetic training and multi-objective optimization to help language models separate formal logic from real-world content biases.