4
Step-Level Preference Learning for Generative Agents in Social Simulations
This research introduces a step-level preference learning method to better align the intermediate decision-making processes of generative agents with human preferences.
4
Energy-Efficient Federated Learning via Adaptive Encoder Freezing for MRI-to-CT Conversion: A Green AI-Guided Research
Researchers have developed an adaptive layer-freezing technique for federated learning to reduce the computational burden of training models in resource-constrained healthcare environments.
4
SAGA: Schema-Aware Grounding for Agentic Text-to-SPARQL Generation
Researchers introduced a schema-aware grounding method to improve the accuracy of agentic text-to-SPARQL query generation for knowledge base question answering.
4
Integration Matters: Rollout-Based Training for Constrained Diffusion Models
The authors introduce a rollout-based training method to improve the performance of constrained diffusion models in satisfying complex feasibility requirements.
4
SportD: Can VLMs Physically Strategize?
This research evaluates the strategic decision-making capabilities of vision-language models by testing their performance in simulated soccer scenarios.
4
RetroAgent: Harnessing LLMs to Search Over Structured Memory for Agentic Retrosynthesis Planning
RetroAgent utilizes large language models to navigate structured memory for more effective multi-step retrosynthesis planning in chemical research.
4
Can We Trust Item Response Theory for AI Evaluation?
The study examines the limitations and potential misapplications of using item response theory to evaluate artificial intelligence benchmarks.
4
VAN-AD: Visual Masked Autoencoder with Normalizing Flow For Time Series Anomaly Detection
The authors introduce a visual masked autoencoder combined with normalizing flows to enhance the generalization capabilities of time series anomaly detection models.
4
Alipay-PIBench: A Realistic Payment Integration Benchmark for Coding Agents
A new benchmark called Alipay-PIBench has been introduced to evaluate the performance of AI coding agents in handling complex payment integration tasks.
4
From Stateless to Situated: Building a Psychological World for LLM-Based Agents
This research proposes a framework to transition LLM-based agents from stateless interactions to situated models capable of maintaining temporal continuity for psychological support.
4
When a Verified World Model Still Loses: Play-Adequacy vs Prediction-Accuracy in LLM-Synthesized Code World Models
The study demonstrates that high transition accuracy in LLM-synthesized world models does not necessarily correlate with effective planning performance.
4
Fully Offline Reinforcement Learning
Researchers have developed SOReL, a Bayesian model-based reinforcement learning method that enables hyperparameter selection without requiring online environment interactions.
4
Stop Thinking, Start Looking: Efficient Post-Training for Multimodal Document Question Answering via Reasoning-Free Alignment
A new post-training method for multimodal document question answering focuses on reasoning-free alignment to reduce computational costs and improve visual grounding.
4
MCPEvol-Bench: Benchmarking LLM Agent Performance Across Dynamic Evolutions of MCP Servers
Researchers introduced a new benchmark designed to evaluate how effectively AI agents adapt to the continuous updates and changes within Model Context Protocol servers.
4
ANet Patu-1: The Value of Connection in the Agent Network
This paper models the value of connectivity within networks of AI agents to determine optimal collaboration protocols.
4
Beyond Generalist LLMs: Specialist Agentic Systems for Structured Code Workflow Execution
Researchers investigate the efficacy of specialized agentic systems compared to general-purpose LLMs for automating business process workflows.
4
SciDiagramEdit: Learning to Edit Scientific Diagrams from Paper Revisions
A new method called SciDiagramEdit uses natural language instructions to automate the complex process of editing scientific diagrams and figures.
4
Similarity as Reward Alignment: Robust and Versatile Preference-based Reinforcement Learning
A new contrastive framework called SARA aims to improve the robustness of preference-based reinforcement learning against noisy or inaccurate human labeling.
4
Just Keep Prompting: Evaluating Repetitive Socratic Prompting in VLMs
Researchers have developed a new evaluation framework called JKP to test the epistemic stability of vision-language models when subjected to repetitive adversarial questioning.
4
Flow Matching in Feature Space for Stochastic World Modeling
A new flow-matching technique for feature-space world modeling addresses the trade-off between reconstruction quality and predictive accuracy in stochastic environments.
4
Beyond Entropy: Correctness-Aware Advantage Shaping via Contrastive Policy Optimization
Researchers introduced Contrastive Policy Optimization to improve reinforcement learning by using token-level disagreement to better distinguish between useful uncertainty and harmful confusion.
4
Assessing AI in Introductory Physics Problem Solving
A study evaluated the performance of OpenAI's o4-mini model on undergraduate physics problems to assess the reasoning capabilities of current inference-scaling LLMs.
4
LIGO-PINN: Learned Initialization via Gated Optimization to Alleviate Convergence Failures in Physics Informed Neural Networks
LIGO-PINN introduces a gated optimization technique to improve the training stability and convergence of physics-informed neural networks in complex domains.
4
Demographically-Conditioned Synthetic Medical Images for Bias Mitigation and Bias Detection in Disease Classifiers
The study demonstrates that using demographically-conditioned synthetic medical images can help detect and mitigate bias in diagnostic classifiers.