4/8/2026
Pramana: Fine-Tuning Large Language Models for Epistemic Reasoning through Navya-Nyaya
Research on fine-tuning LLMs for epistemic reasoning using Navya-Nyaya logic. Addresses hallucination and brittleness in LLM reasoning capabilities.
4/8/2026
Operational Noncommutativity in Sequential Metacognitive Judgments
Theoretical framework exploring order effects in sequential cognitive processes and non-commutativity in metacognition using operational methods.
4/8/2026
Proximity Measure of Information Object Features for Solving the Problem of Their Identification in Information Systems
Proximity measure quantifies similarity of multi-source information object features for entity identification and matching across heterogeneous data sources.
4/8/2026
ReVEL: Multi-Turn Reflective LLM-Guided Heuristic Evolution via Structured Performance Feedback
ReVEL hybrid framework uses LLM-guided iterative evolution with structured performance feedback to design effective heuristics for NP-hard problems.
4/8/2026
Algebraic Structure Discovery for Real World Combinatorial Optimisation Problems: A General Framework from Abstract Algebra to Quotient Space Learning
Framework identifies algebraic structures in combinatorial optimization problems, constructs quotient spaces to reduce search space and improve solution quality.
4/8/2026
PaperOrchestra: A Multi-Agent Framework for Automated AI Research Paper Writing
PaperOrchestra multi-agent framework automates AI research paper writing by transforming unstructured materials into submission-ready LaTeX manuscripts.
4/8/2026
MMORF: A Multi-agent Framework for Designing Multi-objective Retrosynthesis Planning Systems
MMORF multi-agent framework uses language models with specialized agents for multi-objective retrosynthesis planning balancing quality, safety, and cost.
4/8/2026
MedGemma 1.5 Technical Report
MedGemma 1.5 4B model expands medical capabilities with high-dimensional imaging (CT/MRI/histopathology), anatomical localization, and improved document understanding.
4/8/2026
Uncertainty-Guided Latent Diagnostic Trajectory Learning for Sequential Clinical Diagnosis
LLM-based sequential clinical diagnosis system models uncertainty-guided evidence acquisition over time using diagnostic trajectory learning.
4/8/2026
Non-monotonic causal discovery with Kolmogorov-Arnold Fuzzy Cognitive Maps
Kolmogorov-Arnold Fuzzy Cognitive Maps extend neuro-symbolic modeling to handle non-monotonic causal dependencies in complex dynamic systems.
4/8/2026
IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents
IntentScore is a plan-aware reward model trained on 398K offline GUI interactions to evaluate and score actions for computer-use agents across multiple operating systems.
4/8/2026
Bypassing the CSI Bottleneck: MARL-Driven Spatial Control for Reflector Arrays
Multi-agent reinforcement learning replaces channel modeling with spatial intelligence for autonomous control of reconfigurable intelligent surface arrays.
4/8/2026
Learning to Focus: CSI-Free Hierarchical MARL for Reconfigurable Reflectors
Hierarchical multi-agent reinforcement learning optimizes reconfigurable intelligent surfaces for mmWave networks without channel state information estimation.
4/8/2026
Instruction-Tuned LLMs for Parsing and Mining Unstructured Logs on Leadership HPC Systems
Instruction-tuned LLMs parse and mine unstructured HPC system logs from heterogeneous sources to extract patterns and diagnose operational issues.
4/8/2026
ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspaces
ClawsBench benchmark evaluates LLM agents on realistic productivity tasks (email, scheduling, documents) in simulated multi-service environments with stateful workflows.
4/8/2026
Attribution Bias in Large Language Models
AttriBench: Demographically-balanced benchmark for measuring attribution bias in LLMs when attributing quotes to original authors.
4/8/2026
From Governance Norms to Enforceable Controls: A Layered Translation Method for Runtime Guardrails in Agentic AI
Framework for translating governance norms into enforceable runtime guardrails for agentic AI systems with multi-step execution.
4/8/2026
EAGLE: Edge-Aware Graph Learning for Proactive Delivery Delay Prediction in Smart Logistics Networks
Graph neural network approach for predicting delivery delays in logistics networks using warehouse and transportation data.
4/8/2026
Simulating the Evolution of Alignment and Values in Machine Intelligence
Evolutionary theory simulation of how alignment affects populations of AI models over time and belief propagation dynamics.
4/8/2026
Pressure, What Pressure? Sycophancy Disentanglement in Language Models via Reward Decomposition
Reward decomposition approach to disentangle pressure capitulation from evidence blindness in LLM sycophancy behavior.
4/8/2026
Breakthrough the Suboptimal Stable Point in Value-Factorization-Based Multi-Agent Reinforcement Learning
Theoretical analysis and solutions for value factorization convergence to suboptimal stable points in multi-agent reinforcement learning.
4/8/2026
Graph of Skills: Dependency-Aware Structural Retrieval for Massive Agent Skills
Graph of Skills: Dependency-aware skill retrieval system for managing and scaling thousands of reusable skills in agent systems.
4/8/2026
TRACE: Capability-Targeted Agentic Training
TRACE: Framework for targeted training of LLM agents on capability gaps identified in specific environments and task distributions.
4/8/2026
Dynamic Agentic AI Expert Profiler System Architecture for Multidomain Intelligence Modeling
Agentic AI system that profiles user expertise levels to adapt interaction depth using LLaMA-based modular architecture.
4/8/2026
From Retinal Evidence to Safe Decisions: RETINA-SAFE and ECRT for Hallucination Risk Triage in Medical LLMs
RETINA-SAFE benchmark and ECRT framework for detecting hallucination risks in medical LLMs with insufficient or conflicting evidence.
4/8/2026
ETR: Entropy Trend Reward for Efficient Chain-of-Thought Reasoning
ETR: Training method for efficient chain-of-thought reasoning by optimizing entropy trends rather than global uncertainty reduction.
4/8/2026
LatentAudit: Real-Time White-Box Faithfulness Monitoring for Retrieval-Augmented Generation with Verifiable Deployment
LatentAudit: White-box monitoring system for RAG hallucination detection using Mahalanobis distance on residual stream activations.
4/8/2026
TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems
TFRBench: Benchmark for evaluating reasoning capabilities of time-series forecasting systems beyond numerical accuracy metrics.
4/8/2026
LLM-as-Judge for Semantic Judging of Powerline Segmentation in UAV Inspection
Using LLMs as judges to evaluate lightweight segmentation models for drone-based power line inspection under distribution shift.
4/8/2026
Towards Effective In-context Cross-domain Knowledge Transfer via Domain-invariant-neurons-based Retrieval
Domain-invariant neurons approach for cross-domain knowledge transfer to boost LLM reasoning in expertise-scarce specialized domains.
4/8/2026
Reason Analogically via Cross-domain Prior Knowledge: An Empirical Study of Cross-domain Knowledge Transfer for In-Context Learning
Empirical study on using cross-domain demonstrations to improve in-context learning when expert annotations in target domain are scarce.
4/8/2026
HYVE: Hybrid Views for LLM Context Engineering over Machine Data
HYVE framework for LLMs to better process machine data (logs, metrics, traces) through hybrid structured/unstructured representations.
4/8/2026
CODESTRUCT: Code Agents over Structured Action Spaces
CODESTRUCT: LLM-based code agents using structured AST action spaces instead of text matching for reliable code editing and repository interaction.
4/8/2026
Multi-Agent Pathfinding with Non-Unit Integer Edge Costs via Enhanced Conflict-Based Search and Graph Discretization
Research on multi-agent pathfinding algorithms handling non-unit edge costs and continuous-time actions for real-world robotic/logistics scenarios.
4/8/2026
PRISM-MCTS: Learning from Reasoning Trajectories with Metacognitive Reflection
PRISM-MCTS learning approach using reasoning trajectories with metacognitive reflection, inspired by reasoning models like OpenAI o1, for efficient low-resource NLP methods.
4/8/2026
Automated Auditing of Hospital Discharge Summaries for Care Transitions
Automated framework using locally-deployed LLMs to audit hospital discharge summaries at scale, enforcing transition-of-care documentation requirements for patient safety.
4/8/2026
Adaptive Serverless Resource Management via Slot-Survival Prediction and Event-Driven Lifecycle Control
Adaptive serverless resource management framework using slot-survival prediction and event-driven architecture to optimize cold start latency and utilization.
4/8/2026
OntoTKGE: Ontology-Enhanced Temporal Knowledge Graph Extrapolation
OntoTKGE model for temporal knowledge graph extrapolation leveraging ontological knowledge to handle sparse historical interactions and enable behavioral pattern inheritance.
4/8/2026
Can We Trust a Black-box LLM? LLM Untrustworthy Boundary Detection via Bias-Diffusion and Multi-Agent Reinforcement Learning
GMRL-BD algorithm using bias-diffusion and multi-agent RL to detect untrustworthy topic boundaries of LLMs, identifying domains where model answers cannot be reliably trusted.
4/8/2026
Auditable Agents
Auditable Agents framework establishing accountability, auditability, and auditing definitions for LLM agents with external effects, addressing post-deployment answer-ability.
4/8/2026
SCMAPR: Self-Correcting Multi-Agent Prompt Refinement for Complex-Scenario Text-to-Video Generation
SCMAPR stage-wise multi-agent refinement framework for complex scenario text-to-video generation that refines and self-corrects ambiguous prompts through agent collaboration.
4/8/2026
Thinking Diffusion: Penalize and Guide Visual-Grounded Reasoning in Diffusion Multimodal Language Models
Thinking Diffusion method adding reasoning penalization and guidance to diffusion multimodal LLMs combining Chain-of-Thought reasoning with parallel generation capabilities.
4/8/2026
OmniDiagram: Advancing Unified Diagram Code Generation via Visual Interrogation Reward
OmniDiagram unified framework for code generation across diverse diagram types and languages using visual interrogation reward for alignment with visual specifications.
4/8/2026
UniCreative: Unifying Long-form Logic and Short-form Sparkle via Reference-Free Reinforcement Learning
UniCreative approach using reference-free reinforcement learning to balance long-form coherence and short-form expressiveness in LLM-based creative writing generation.
4/8/2026
Market-Bench: Benchmarking Large Language Models on Economic and Trade Competition
Market-Bench comprehensive benchmark evaluating LLM capabilities in economically-relevant tasks via configurable multi-agent supply chain model with LLM retailer agents.
4/8/2026
ActivityEditor: Learning to Synthesize Physically Valid Human Mobility
ActivityEditor dual-LLM-agent framework for zero-shot cross-regional human trajectory generation, synthesizing physically valid mobility patterns without region-specific historical data.
4/8/2026
Inventory of the 12 007 Low-Dimensional Pseudo-Boolean Landscapes Invariant to Rank, Translation, and Rotation
Analysis of 12,007 rank-invariant pseudo-Boolean landscapes introducing stronger notion of rank landscape equivalence under translation and rotation symmetries.
4/8/2026
Experience Transfer for Multimodal LLM Agents in Minecraft Game
Echo memory framework for multimodal LLM agents enabling transfer of reusable knowledge across Minecraft tasks by decomposing experience into five interpretable dimensions.
4/8/2026
SignalClaw: LLM-Guided Evolutionary Synthesis of Interpretable Traffic Signal Control Skills
SignalClaw framework using LLMs as evolutionary skill generators to synthesize interpretable traffic signal control strategies balancing effectiveness and explainability.
4/8/2026
A canonical generalization of OBDD
Introduces Tree Decision Diagrams generalizing OBDD for Boolean function representation with improved succinctness and tractable operations like model counting and conditioning.