LLM optimization
How Quantization and Batching Cut LLM Energy Costs
New research explores how quantization, batching strategies, and serving optimizations dramatically reduce LLM energy consumption while maintaining performance.
Research
LLM Reliability
Six Sigma Agents: Consensus Architecture Hits 99.99% LLM Reliabil
New research achieves enterprise-grade 99.99966% reliability in LLM systems through consensus-driven decomposed execution, bringing Six Sigma quality standards to AI agents.
Conformal Prediction
Conformal Prediction Meets Generative AI: New Uncertainty Method
New research introduces adaptive cluster-based density estimation for conformal prediction in generative models, enabling statistical guarantees on AI-generated content quality and reliability.
AI agents
Training Small AI Agents with Synthetic Worlds and Rubric Rewards
New research presents a framework for building capable small language model agents using synthetic tasks, simulated environments, and structured rubric-based rewards—democratizing agentic AI development.
LLM Security
Drunk Prompts: Novel Jailbreak Method Exposes LLM Safety Gaps
Researchers discover that simulating intoxicated speech patterns can bypass AI safety guardrails. The 'In Vino Veritas' attack reveals fundamental weaknesses in how LLMs handle linguistic degradation.
AI Detection
Training Humans to Detect AI Text: Rubric-Based Calibration
New research introduces cognitive calibration methods to improve human detection of LLM-generated Korean text, shifting from intuition to expertise-based assessment.
LLM Agents
New Framework for Automated Testing of LLM Agent Reliability
Researchers introduce methods and a framework for automated structural testing of LLM-based agents, addressing critical reliability challenges in agentic AI systems through systematic evaluation approaches.
AI Systems
SETA: A New Framework for Debugging Multi-Component AI Systems
Researchers introduce SETA, a statistical method for identifying which component in complex AI pipelines causes failures—critical for debugging multi-stage systems like video generation workflows.
neuro-symbolic AI
Tensor Logic: Bridging Symbolic AI and Neural Networks
New research unifies Datalog symbolic reasoning with neural computation via tensor contractions, enabling differentiable logic programming with potential implications for AI reasoning systems.
neural networks
Differential Geometry Unlocks Neural Network Information Flow
New research applies differential geometry to analyze how information propagates through neural networks, offering mathematical tools to understand deep learning architectures at a fundamental level.
LLM Agents
Diagnosing Tool Failures in Multi-Agent LLM Systems
New research introduces a systematic framework for identifying why LLM agents fail to invoke tools correctly, addressing a critical reliability gap in multi-agent AI systems.
AI Video Generation
Why AI Video Models Need Physics: From Generation to Simulation
New research argues current AI video generators like Sora lack true physical understanding. The paper proposes a shift from pattern-matching to physics-grounded world models for reliable simulation.