Research

LLM

LLMs as Neural Architects: Auto-Generating Image Captioning Model

New research demonstrates LLMs can design complete neural network architectures for image captioning under strict API constraints, opening new possibilities for automated AI system design.

LLM Training

Entropy-Driven Gradient Compression Advances LLM Training

New EDGC method uses entropy to dynamically compress gradients during LLM training, reducing communication overhead while preserving model accuracy across distributed systems.

LLM Infrastructure

FPGA-Based CXL Memory Architecture Tackles LLM KV-Cache Bottlenec

New research proposes CXL-SpecKV, a disaggregated FPGA architecture using CXL memory pooling and speculative prefetching to overcome memory bottlenecks in large language model inference at datacenter scale.

AI Safety

AI Models Can Learn to Hide Thoughts From Safety Monitors

New research reveals language models can learn to conceal internal states from activation-based monitoring systems, raising critical questions for AI safety and detection systems.

diffusion models

GenSOT: Harmonic Path-Integral Diffusion for Optimal Transport

New research introduces Generative Stochastic Optimal Transport (GenSOT), combining harmonic path-integral methods with optimal transport theory to improve guided diffusion model generation.

AI Safety

New Research Proposes Behavioral Guidance for Trustworthy AI Agen

Researchers present a framework for making multi-turn LLM agents more trustworthy through behavioral guidance, addressing critical safety concerns as AI systems become more autonomous.

LLM

New KV Cache Method Enables Sublinear Memory Growth for LLMs

Researchers introduce Adaptive Soft Rolling KV Freeze with entropy-guided recovery, achieving sublinear memory scaling for long-context LLM inference without significant quality loss.

AI security

AI Security Battles: Is Robust Alignment an Endless Struggle?

New research examines whether AI security and alignment efforts face fundamental limitations, analyzing the cycle of safety measures and adversarial bypasses in modern AI systems.

agentic AI

Deep Dive: Architectural Patterns for Building Agentic AI Systems

New research surveys the core architectural patterns enabling autonomous AI agents, from single-agent designs to multi-agent orchestration frameworks that power complex AI workflows.

LLM Agents

Dual-Strategy Framework Boosts LLM Agent Decision Making

New research introduces a co-adaptive dual-strategy framework combining fast intuitive reasoning with slow deliberative thinking to improve LLM-based agent performance.

LLM Agents

SABER Framework Tackles Error Cascades in LLM Agents

New research introduces SABER, a safeguarding framework that identifies how small errors in LLM agent actions can cascade into significant failures, proposing intervention mechanisms.

AI agents

Research Asks: Can AI Agents Build and Run Data Systems?

New arXiv research explores whether AI agents can autonomously build, operate, and utilize complete data infrastructure, examining the boundaries of agentic AI capabilities.