LLM Agents
New research introduces Tool-R0, a framework enabling LLM agents to autonomously learn tool usage through self-evolution, eliminating the need for curated training datasets while achieving state-of-the-art performance.
LLM Agents
Researchers introduce a new benchmark for evaluating how general LLM agents perform when given additional compute resources at inference time, addressing a critical gap in agent evaluation.
LLM Interpretability
New research introduces ADAPT, a hybrid optimization technique that combines discrete and continuous methods to visualize and understand internal features of large language models.
LLM fine-tuning
New research introduces proxy methods that preserve gradient influence signals while dramatically reducing computational costs for selecting optimal training data in large language model fine-tuning.
LLM
New research applies software product line variability modeling to systematically optimize LLM inference hyperparameters like temperature and sampling strategies.
LLM Safety
New research explores whether constraining specific parameter regions in large language models can ensure safety, examining the theoretical foundations of alignment through architectural constraints.
LLM
New research reveals that LLMs reason better using their own examples rather than human-provided ones, suggesting the process of generation matters more than example quality.
Agentic AI
New research proposes proxy state-based evaluation for multi-turn tool-calling LLM agents, addressing the challenge of scalable reward verification in complex agentic workflows.
AI Research
Researchers establish mathematical framework for understanding how generative AI models can survive training on contaminated data, offering crucial insights for maintaining synthetic media quality.
world models
World models enable AI to simulate reality by learning internal representations of environments. This foundational architecture powers next-gen video generation, robotics, and autonomous systems.
LLM fine-tuning
New research introduces learnable direction sampling for zero-order optimization, dramatically reducing memory requirements for fine-tuning large language models without sacrificing performance.
LLM Agents
New research applies martingale theory to analyze how information degrades in tool-using LLM agents operating under the Model Context Protocol, establishing mathematical bounds on agent reliability.