4/14/2026
CricBench: A Multilingual Benchmark for Evaluating LLMs in Cricket Analytics
CricBench: benchmark for evaluating LLMs on multilingual cricket analytics and domain-specific Text-to-SQL tasks.
4/14/2026
Can Small Training Runs Reliably Guide Data Curation? Rethinking Proxy-Model Practice
Research questions whether small proxy model training reliably guides data curation decisions for full-scale frontier AI model pretraining.
4/14/2026
Disco-RAG: Discourse-Aware Retrieval-Augmented Generation
Disco-RAG improves retrieval-augmented generation by capturing discourse structure and synthesizing knowledge from dispersed evidence.
4/14/2026
Enhanced-FQL($\lambda$), an Efficient and Interpretable RL with novel Fuzzy Eligibility Traces and Segmented Experience Replay
Enhanced-FQL(λ) reinforcement learning framework with fuzzy eligibility traces and interpretable fuzzy rules for continuous control.
4/14/2026
Merging Triggers, Breaking Backdoors: Defensive Poisoning for Instruction-Tuned Language Models
Defensive poisoning technique merges triggers to remove backdoors in instruction-tuned LLMs vulnerable to data poisoning attacks.
4/14/2026
What Users Leave Unsaid: Under-Specified Queries Limit Vision-Language Models
HAERAE-Vision benchmark with 653 real-world underspecified visual questions reveals vision-language model limitations with informal queries.
4/14/2026
GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO
GanitLLM: Bengali mathematical reasoning model with difficulty-aware curriculum-based GRPO training pipeline.
4/14/2026
Controlling Multimodal Conversational Agents with Coverage-Enhanced Latent Actions
Coverage-enhanced latent actions framework for controlling multimodal conversational agents with reinforcement learning.
4/14/2026
The Poisoned Apple Effect: Strategic Manipulation of Mediated Markets via Technology Expansion of AI Agents
Game-theoretic analysis of how expanding AI agent capabilities affects strategic interaction in bargaining, negotiation, and persuasion.
4/14/2026
Powerful Training-Free Membership Inference Against Autoregressive Language Models
EZ-MIA: Training-free membership inference attack against fine-tuned language models to audit privacy risks from data memorization.
4/14/2026
XD-MAP: Cross-Modal Domain Adaptation via Semantic Parametric Maps for Scalable Training Data Generation
Cross-modal domain adaptation approach transferring image dataset knowledge to LiDAR for synthetic training data generation.
4/14/2026
Parallelism and Generation Order in Masked Diffusion Language Models: Limits Today, Potential Tomorrow
Analyzes parallelism and generation order in Masked Diffusion Language Models across 8 models and 58 benchmarks.
4/14/2026
StreetDesignAI: A Multi-Persona Evaluation System for Inclusive Infrastructure Design
Uses persona-based evaluation with LLMs to support inclusive cycling infrastructure design by simulating diverse user experiences.
4/14/2026
Who Gets Which Message? Auditing Demographic Bias in LLM-Generated Targeted Text
Systematic analysis of demographic bias in LLM-generated targeted messaging across GPT-4o, Llama-3.3, and Mistral-Large models.
4/14/2026
MERMAID: Memory-Enhanced Retrieval and Reasoning with Multi-Agent Iterative Knowledge Grounding for Veracity Assessment
MERMAID: Multi-agent system for fact-checking using LLMs with memory-enhanced retrieval and iterative reasoning to assess veracity of claims.
4/14/2026
Beyond RAG for Agent Memory: Retrieval by Decoupling and Aggregation
Agent memory system beyond RAG addressing agent-specific needs: bounded coherent dialogue retrieval with decoupling and aggregation.
4/14/2026
Why Steering Works: Toward a Unified View of Language Model Parameter Dynamics
Unified framework explaining LLM steering methods (fine-tuning, LoRA, activation interventions) as dynamic weight updates from control signals.
4/14/2026
El Agente Estructural: An Artificially Intelligent Molecular Editor
El Agente Estructural multimodal agent for autonomous molecular geometry generation and manipulation using natural language and vision.
4/14/2026
Fake-HR1: Rethinking Reasoning of Vision Language Model for Synthetic Image Detection
Fake-HR1 hybrid-reasoning model for synthetic image detection balancing chain-of-thought reasoning with computational efficiency.
4/14/2026
Pyramid MoA: A Probabilistic Framework for Cost-Optimized Anytime Inference
Pyramid MoA hierarchical mixture-of-agents architecture with decision-theoretic router for cost-optimized anytime LLM inference.
4/14/2026
Reasoning as Gradient: Scaling MLE Agents Beyond Tree Search
Gome agent for machine learning engineering using gradient-based optimization instead of tree search, scaling LLM-based reasoning.
4/14/2026
How Controllable Are Large Language Models? A Unified Evaluation across Behavioral Granularities
SteerEval benchmark for evaluating LLM controllability across language features, sentiment, and personality at multiple specification levels.
4/14/2026
Prompt Injection as Role Confusion
Analysis of prompt injection attacks as role confusion where models infer text source by content style rather than origin.
4/14/2026
Resource Consumption Threats in Large Language Models
Survey of resource consumption threats in LLMs including excessive generation attacks, resource efficiency requirements, and mitigation strategies.
4/14/2026
Detection Is Cheap, Routing Is Learned: Why Refusal-Based Alignment Evaluation Fails
Study of LLM alignment evaluation focusing on routing from concept detection to behavioral policy, using Chinese language models as case study.
4/14/2026
Cognitive Training for Language Models: Towards General Capabilities via Cross-Entropy Games
Curriculum learning framework using cross-entropy games to automatically build general capabilities and discover skills in language models.
4/14/2026
Measuring and curing reasoning rigidity: from decorative chain-of-thought to genuine faithfulness
Introduces Step-Level Reasoning Capacity metric and LC-CoSR training method to measure and reduce reasoning rigidity in chain-of-thought reasoning.
4/14/2026
ForestPrune: High-ratio Visual Token Compression for Video Multimodal Large Language Models via Spatial-Temporal Forest Modeling
Training-free spatial-temporal token compression method for video MLLMs achieving high-ratio visual token reduction via forest modeling.
4/14/2026
MSA: Memory Sparse Attention for Efficient End-to-End Memory Model Scaling to 100M Tokens
Memory-sparse attention mechanism enabling LLMs to scale context to 100M tokens through efficient memory modeling instead of full attention.
4/14/2026
DecepGPT: Schema-Driven Deception Detection with Multicultural Datasets and Robust Multimodal Learning
Multimodal deception detection using schema-driven approach with audiovisual analysis across multicultural datasets for forensics applications.
4/14/2026
Policy-Guided Threat Hunting: An LLM enabled Framework with Splunk SOC Triage
LLM-enabled threat hunting framework for SOC analysts integrating Splunk with policy-guided decision making for APT detection.
4/14/2026
GameplayQA: A Benchmarking Framework for Decision-Dense POV-Synced Multi-Video Understanding of 3D Virtual Agents
Benchmark framework for evaluating multimodal LLMs as perceptual backbones for autonomous agents in 3D environments with decision-dense scenarios.
4/14/2026
Learning to Focus and Precise Cropping: A Reinforcement Learning Framework with Information Gaps and Grounding Loss for MLLMs
Reinforcement learning framework enabling multimodal LLMs to autonomously crop and focus on regions of interest for improved perception in complex visual scenes.
4/14/2026
C2F-Thinker: Coarse-to-Fine Reasoning with Hint-Guided Reinforcement Learning for Multimodal Sentiment Analysis
Coarse-to-fine reasoning framework using reinforcement learning for interpretable multimodal sentiment analysis with MLLMs and hint-guided training.
4/14/2026
DarwinNet: An Evolutionary Network Architecture for Agent-Driven Protocol Synthesis
Bio-inspired self-evolving network architecture for autonomous agents using evolutionary approaches instead of static human-defined protocols.
4/14/2026
Beyond Message Passing: A Semantic View of Agent Communication Protocols
Framework analyzing agent communication protocols for LLM systems across three layers: communication, syntactic, and semantic. Systematically organizes 18 representative protocols.
4/14/2026
When simulations look right but causal effects go wrong: Large language models as behavioral simulators
Evaluates whether LLMs can infer causal intervention effects from natural language descriptions using behavioral simulation on climate-psychology interventions.
4/14/2026
LitPivot: Developing Well-Situated Research Ideas Through Dynamic Contextualization and Critique within the Literature Landscape
LitPivot: tool supporting iterative research idea development through dynamic literature contextualization and AI-driven critique.
4/14/2026
Valence-Arousal Subspace in LLMs: Circular Emotion Geometry and Multi-Behavioral Control
Method to identify valence-arousal emotional subspace in LLM representations for emotion steering and behavioral control.
4/14/2026
Can LLMs Reason About Attention? Towards Zero-Shot Analysis of Multimodal Classroom Behavior
Privacy-preserving system using LLMs to analyze student attention in classroom videos without storing identifiable footage.
4/14/2026
Zero-Shot Quantization via Weight-Space Arithmetic
Zero-shot quantization method using weight-space arithmetic to improve post-training quantization robustness across models.
4/14/2026
The Augmentation Trap: AI Productivity and the Cost of Cognitive Offloading
Economic analysis of AI productivity gains showing sustained tool use erodes worker expertise over time.
4/14/2026
Inside the Scaffold: A Source-Code Taxonomy of Coding Agent Architectures
Taxonomy of LLM-based coding agent architectures via source-code analysis, categorizing control loops, tool definitions, and context strategies.
4/14/2026
How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models
Analysis of policy routing circuits in alignment-trained LLMs, localizing attention gates and amplifier heads controlling refusal behavior.
4/14/2026
StableTTA: Training-Free Test-Time Adaptation that Improves Model Accuracy on ImageNet1K to 96%
StableTTA: training-free test-time adaptation improving image classification accuracy via novel ensemble strategies.
4/14/2026
EduIllustrate: Towards Scalable Automated Generation Of Multimodal Educational Content
EduIllustrate: benchmark evaluating LLMs on automated generation of diagram-rich educational content combining visuals and reasoning.
4/14/2026
VideoStir: Understanding Long Videos via Spatio-Temporally Structured and Intent-Aware RAG
VideoStir: retrieval-augmented generation system for understanding long videos with multimodal LLMs using spatio-temporal structure.
4/14/2026
The Defense Trilemma: Why Prompt Injection Defense Wrappers Fail?
Theoretical analysis proving limitations of continuous wrapper defenses against prompt injection attacks in LLMs.
4/14/2026
MoBiE: Efficient Inference of Mixture of Binary Experts under Post-Training Quantization
MoBiE: binarization framework for efficient inference in Mixture-of-Experts LLMs via post-training quantization.
4/14/2026
Latent Structure of Affective Representations in Large Language Models
Analysis of emotional representation geometry in LLM latent spaces for transparency and safety.