2/25/2026
Regret-Guided Search Control for Efficient Learning in AlphaZero
Search control method for AlphaZero using regret-guided learning, improving sample efficiency by restarting from valuable game states rather than initial positions.
2/25/2026
SoK: Agentic Skills -- Beyond Tool Use in LLM Agents
Systematization of knowledge on agentic skills—reusable procedural capabilities with explicit conditions, policies, and interfaces for reliable long-horizon workflows in LLM agents.
2/25/2026
Airavat: An Agentic Framework for Internet Measurement
Airavat: agentic framework automating internet measurement workflows and verification against methodological standards using AI agents for tool orchestration.
2/25/2026
The Art of Efficient Reasoning: Data, Reward, and Optimization
Research on efficient reasoning in LLMs using Chain-of-Thought, addressing computational overhead through reward shaping and reinforcement learning optimization.
2/25/2026
See and Fix the Flaws: Enabling VLMs and Diffusion Models to Comprehend Visual Artifacts via Agentic Data Synthesis
Agentic data synthesis approach to help vision-language models and diffusion models detect and mitigate visual artifacts in AI-generated images.
2/25/2026
Does Order Matter : Connecting The Law of Robustness to Robust Generalization
Theoretical analysis connecting law of robustness to robust generalization, resolving open problem on overparameterization requirements for robust interpolation.
2/25/2026
Toward an Agentic Infused Software Ecosystem
Vision paper proposing Agentic Infused Software Ecosystem (AISE) framework rethinking software development tools and practices for autonomous AI agents.
2/25/2026
CrystaL: Spontaneous Emergence of Visual Latents in MLLMs
CrystaL method enables spontaneous emergence of visual latent representations in MLLMs through improved supervision for latent chain-of-thought reasoning.
2/25/2026
Multimodal MRI Report Findings Supervised Brain Lesion Segmentation with Substructures
Report-supervised learning for brain lesion segmentation using multimodal MRI and radiology report constraints on substructures.
2/25/2026
MIP Candy: A Modular PyTorch Framework for Medical Image Processing
MIP Candy is a modular PyTorch framework for medical image processing with support for volumetric data, multiple formats, and domain-specific training.
2/25/2026
VAUQ: Vision-Aware Uncertainty Quantification for LVLM Self-Evaluation
VAUQ provides vision-aware uncertainty quantification for evaluating Large Vision-Language Model outputs and reducing hallucinations in vision-conditioned tasks.
2/25/2026
Localized Dynamics-Aware Domain Adaption for Off-Dynamics Offline Reinforcement Learning
Localized domain adaptation method for offline reinforcement learning addressing dynamics mismatch between source and target domains.
2/25/2026
Probing Graph Neural Network Activation Patterns Through Graph Topology
Analyzes how graph topology interacts with GNN learned preferences through massive activations, investigating oversmoothing and oversquashing artifacts.
2/25/2026
Cooperative-Competitive Team Play of Real-World Craft Robots
Multi-agent deep RL system for training cooperative-competitive robot teams in simulation with transfer to real-world robotic applications.
2/25/2026
"Are You Sure?": An Empirical Study of Human Perception Vulnerability in LLM-Driven Agentic Systems
Study examines security vulnerabilities in LLM-driven agents through Agent-Mediated Deception attacks, revealing human perception susceptibility to compromised copilots.
2/25/2026
SparkMe: Adaptive Semi-Structured Interviewing for Qualitative Insight Discovery
SparkMe uses LLMs to automate semi-structured qualitative interviews with adaptive topic exploration balanced against systematic coverage.
2/25/2026
PVminer: A Domain-Specific Tool to Detect the Patient Voice in Patient Generated Data
PVminer uses ML/NLP to detect patient voice in healthcare text data, scaling qualitative analysis of patient-generated messages across health systems.
2/25/2026
XMorph: Explainable Brain Tumor Analysis Via LLM-Assisted Hybrid Deep Intelligence
XMorph framework combines deep learning with LLM assistance for interpretable brain tumor classification and segmentation on medical images with computational efficiency.
2/25/2026
Why Pass@k Optimization Can Degrade Pass@1: Prompt Interference in LLM Post-training
Study reveals that optimizing Pass@k metric for LLM code generation can degrade Pass@1 performance due to prompt interference, analyzing trade-offs in inference-aware fine-tuning methods.
2/25/2026
Learning from Trials and Errors: Reflective Test-Time Planning for Embodied LLMs
Reflective Test-Time Planning enables embodied LLMs to learn from failures through in-action and post-action reflection modes, improving robot task reasoning and reducing repeated mistakes.
2/25/2026
Test-Time Training with KV Binding Is Secretly Linear Attention
Analysis revealing test-time training with KV binding functions as linear attention rather than memorization-based meta-learning.
2/25/2026
A Survey on the Optimization of Large Language Model-based Agents
Comprehensive survey on optimization techniques for LLM-based agents covering prompt design, fine-tuning, and performance improvements.
2/25/2026
"Don't Do That!": Guiding Embodied Systems through Large Language Model-based Constraint Generation
STPR framework uses LLMs to generate formal constraints for robot navigation from complex natural language spatial and conditional specifications.
2/25/2026
Sensory-Motor Control with Large Language Models via Iterative Policy Refinement
Method enabling LLMs to control embodied agents by generating and iteratively refining control policies from natural language descriptions.
2/25/2026
Programming by Backprop: An Instruction is Worth 100 Examples When Finetuning LLMs
Programming by Backprop enables LLMs to learn procedural knowledge from declarative instructions during training, improving few-shot capabilities.
2/25/2026
AutoEDA: Enabling EDA Flow Automation through Microservice-Based LLM Agents
AutoEDA applies microservice-based LLM agents to automate Electronic Design Automation workflows, replacing Tcl scripting with natural language.
2/25/2026
Hidden Dynamics of Massive Activations in Transformer Training
Comprehensive analysis of massive activation patterns in transformer training using Pythia model family with public dataset release.
2/25/2026
TASER: Table Agents for Schema-guided Extraction and Recommendation
TASER uses AI agents to extract structured data from unformatted financial tables for QA and data normalization tasks.
2/25/2026
Hybrid Deep Searcher: Scalable Parallel and Sequential Search Reasoning
Hybrid Deep Searcher combining parallel and sequential search reasoning for large reasoning models with RAG enabling test-time scaling.
2/25/2026
DS-STAR: Data Science Agent for Solving Diverse Tasks across Heterogeneous Formats and Open-Ended Queries
DS-STAR: Data science agent for multi-source data integration and exploration synthesizing open-ended insights across heterogeneous formats.
2/25/2026
TimeOmni-1: Incentivizing Complex Reasoning with Time Series in Large Language Models
TimeOmni-1: Multimodal time series dataset and methods for advancing complex reasoning with temporal data in large language models.
2/25/2026
A Framework for Studying AI Agent Behavior: Evidence from Consumer Choice Experiments
ABxLab framework evaluating LLM-powered agent decision-making in realistic consumer choice scenarios beyond task competence metrics.
2/25/2026
NewtonBench: Benchmarking Generalizable Scientific Law Discovery in LLM Agents
NewtonBench: Benchmark for evaluating LLM agents in scientific law discovery addressing memorization, scalability, and authentic scientific process.
2/25/2026
Evaluating and Mitigating LLM-as-a-judge Bias in Communication Systems
Systematic investigation of biases in LLM-as-a-judge systems for autonomous content evaluation with mitigation strategies.
2/25/2026
LLMs Process Lists With General Filter Heads
Mechanistic analysis revealing that LLMs use compact filter head representations to encode general filtering operations in list-processing tasks.
2/25/2026
MindPower: Enabling Theory-of-Mind Reasoning in VLM-based Embodied Agents
MindPower framework enabling theory-of-mind reasoning in VLM-based embodied agents for robot-centric decision-making and action generation.
2/25/2026
From Moderation to Mediation: Can LLMs Serve as Mediators in Online Flame Wars?
Study on using LLMs as mediators in online conflicts to foster empathy and de-escalate flame wars instead of just moderating.
2/25/2026
STAR: Similarity-guided Teacher-Assisted Refinement for Super-Tiny Function Calling Models
STAR: Method for distilling function calling capabilities from large LLMs into smaller models using similarity-guided teacher assistance.
2/25/2026
BrowseComp-$V^3$: A Visual, Vertical, and Verifiable Benchmark for Multimodal Browsing Agents
BrowseComp benchmark for evaluating multimodal browsing agents with visual verification and deep search capabilities in open-world environments.
2/25/2026
Automated Generation of Microfluidic Netlists using Large Language Models
First practical application of LLMs for automated microfluidic netlist generation, connecting practitioners with design automation techniques.
2/25/2026
Towards Attributions of Input Variables in a Coalition
Theoretical framework for variable partition attribution in Shapley value-based explainable AI methods addressing attribution conflicts.
2/25/2026
Augmenting Lateral Thinking in Language Models with Humor and Riddle Data for the BRAINTEASER Task
Fine-tuned DeBERTa system augmented with humor and riddle data to improve lateral thinking and creative reasoning in language models.
2/25/2026
A Problem-Oriented Perspective and Anchor Verification for Code Optimization
Investigation of LLM capabilities for code optimization and performance enhancement using problem-oriented perspective and anchor verification.
2/25/2026
Oracular Programming: A Modular Foundation for Building LLM-Enabled Software
Framework for modular composition and reliable execution of LLM-enabled software with enforced contracts.
2/25/2026
Using the Path of Least Resistance to Explain Deep Networks
Gradient-based attribution method for explaining deep neural network decisions.
2/25/2026
Distributional Vision-Language Alignment by Cauchy-Schwarz Divergence
Method for improving vision-language alignment using Cauchy-Schwarz divergence.
2/25/2026
Bridging Gaps in Natural Language Processing for Yor\`ub\'a: A Systematic Review of a Decade of Progress and Prospects
Systematic review of NLP progress and challenges for Yoruba language.
2/25/2026
Semantic Parallelism: Redefining Efficient MoE Inference via Model-Data Co-Scheduling
Semantic parallelism optimization for efficient MoE model inference across multiple devices.
2/25/2026
Diffusion Generative Recommendation with Continuous Tokens
Diffusion-based recommendation system using continuous tokens with LLM integration.
2/25/2026
How much does context affect the accuracy of AI health advice?
Evaluation of LLM accuracy for health advice across languages and contextual factors.