2
Transfer Learning Across Policy Regimes in Adaptive Multi-Agent Systems
This paper frames institutional and regulatory changes in adaptive socio-technical systems as a transfer-learning problem within multi-agent environments.
2
Efficiently Adapting Spoken Language Models for the Singaporean Context
Researchers adapted an open-source spoken language model to the multilingual Singaporean context using parameter-efficient fine-tuning and synthetic datasets.
2
Large Multimodal Model-Based Environment-Aware Mobility Management
This paper explores the integration of large multimodal models to improve mobility management in wireless networks by predicting user trajectories and making real-time decisions.
2
Multimodal Reward Hacking in Reinforcement Learning
This paper investigates the phenomenon of reward hacking in multimodal large language models aligned via reinforcement learning across various tasks and model scales.
2
What You Train Is What You Get: Gender Bias, Training Composition, and Post-Hoc Mitigation in Audio Deepfake Detection
A study on the ASVspoof5 dataset reveals how gender composition in training data affects the performance and bias of audio deepfake detection models.
2
Workload-Driven Optimization for On-Device Real-Time Subtitle Translation
The study explores optimization techniques for low-latency, on-device English-to-Traditional-Chinese subtitle translation under strict resource constraints.
2
Exploring Agentic Workflows for Generating High Quality Math Visual Aids
The paper examines how agentic workflows can be used to generate accurate and pedagogically effective mathematical diagrams for middle school education.
2
The Patchwork Problem in LLM-Generated Code
This paper identifies the patchwork problem in LLM-generated code, where locally correct code snippets fail to integrate coherently into larger software projects.
2
From Checker to Forecaster: Code-Owned Evaluation of Model-Generated Strategic Routes Under Delayed Ground Truth
This study introduces RouteCast, an evaluation framework designed to assess model-generated strategic routes when ground truth feedback is delayed or private.
2
中信建投：关注能源结构的绿色低碳转型、新能源及氢基燃料的发展
New government guidelines in China emphasize the transition toward green energy and low-carbon industrial development for the upcoming five-year plan.
2
Listen to the Features: Voice Anonymization Driven by Content Embedding Matching over Signal Reconstruction
Researchers developed a voice anonymization model that prioritizes content preservation over realistic speech generation by decoding content embeddings without waveform reconstruction loss.
2
STAMP: Provenance-Guided Credit Assignment for Deep Search Agents
The proposed STAMP framework addresses the reward-credit mismatch in deep-search agents by using a reference-based verifier to evaluate whether cited documents support specific claims.
2
CRiT-QA: Evaluating Multi-hop Reasoning with Counterfactual Chains and Distractor Traps
The CRiT-QA benchmark evaluates the multi-hop reasoning capabilities of large language models by testing their reliance on context versus internal knowledge using counterfactual chains and distractor traps.
2
Measure the Sim-to-Real Gap: Designing an Affordable Real-World Benchmark Platform for Reinforcement Learning in AIoT Systems
This paper proposes an affordable physical benchmark platform to evaluate the transferability of reinforcement learning models from simulation to real-world AIoT systems.
2
AdvNav: Behavior-Guided Black-Box Adversarial Attacks on Vision-Language Navigation
The paper proposes AdvNav, a behavior-guided black-box adversarial attack method targeting the vulnerabilities of Vision-and-Language Navigation systems.
2
Self-Regulated Reading with AI Support: An Eight-Week Study with Students
A longitudinal study of college students reveals how AI chatbot interactions influence cognitive engagement and reading habits during academic tasks.
2
AgentLens: Production-Assessed Trajectory Reviews for Coding Agent Evaluation
AgentLens is a new benchmark designed to evaluate coding agents by assessing the quality of their entire interaction trajectory rather than just final task outcomes.
2
Efficiently Learning Branching Networks for Multitask Algorithmic Reasoning
This study presents a method for training branching neural networks to perform multiple algorithmic reasoning tasks simultaneously.
2
Agentic Service-Oriented Computing: A Manifesto for the Next Frontier of Service-Oriented Computing
This manifesto outlines how LLM-powered autonomous agents are transforming software systems and proposes integrating them with service-oriented computing principles.
2
Keep Policy Gradient in Charge: Sibling-Guided Credit Distillation for Long-Horizon Tool-Use Agents
A novel reinforcement learning technique called Sibling-Guided Credit Distillation improves how agents learn to use tools by better attributing rewards to specific actions.
2
When Does Personality Composition Matter for Multi-Agent LLM Teams?
Researchers investigated how personality-based prompting affects the task performance of multi-agent LLM teams, finding that communication styles significantly influence collaborative outcomes.
2
Learning When to Trust in Contextual Social Bandits
A new study identifies a failure mode called contextual sycophancy, where reinforcement learning agents receive biased feedback from evaluators in specific critical scenarios.
2
TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale
TerraZero provides a procedural simulation environment and training stack to support the development of autonomous driving agents through large-scale self-play.
2
Understanding Sources of Demographic Predictability in Brain MRI via Disentangling Anatomy and Contrast
A study investigates how anatomical and contrast factors in brain MRI scans contribute to demographic predictability, highlighting potential biases in medical AI.