Search papers, labs, and topics across Lattice.
7
1
12
LLM agents can enhance training-data strategies in over half of their attempts, but their inconsistency reveals critical limitations in recursive self-improvement.
Jointly optimizing action predictions and state transitions leads to significant performance gains in GUI agent training, outperforming conventional methods.
Vortex achieves up to 4.7 times higher throughput for large language models, revolutionizing how researchers can prototype and evaluate sparse attention algorithms.
Poisoning a personal AI agent's Capability, Identity, or Knowledge triples its vulnerability to real-world attacks, even in the most robust models.
Skip the expensive supervised fine-tuning: this RL-only method teaches LLMs to use tools by showing them how in-context, then gradually removing the crutches until they're tool-using pros in zero-shot.
Forget prompting monolithic models – ImageEdit-R1 uses reinforcement learning to orchestrate a team of specialized agents, outperforming even closed-source diffusion models on complex image editing tasks.
RLHF struggles with long contexts because the reward signal for *finding* the right information vanishes, but can be revived by directly rewarding the model for selecting relevant context.
