Search papers, labs, and topics across Lattice.
13
0
13
83
Reasoning supervision can be compressed over 450x with uncertainty-guided contraction, boosting LLM efficiency without sacrificing capability.
Frontier multimodal agents struggle to autonomously acquire and utilize task-relevant information in unfamiliar physical environments, failing 77% of the time on a new exploration benchmark.
Multi-agent collaboration often performs worse than solo reasoning when peers hallucinate, but Bayesian tracking of advisor reliability rooted in internal belief states keeps collective performance strictly above autonomous baselines even under high misinformation.
Robotic world-action models do not need pretrained internet video backbones to scale: training native planning and dynamics modules from scratch across 30,000 hours unlocks strong zero-shot manipulation and drives a 17.7-point gain on an embodiment comprising under 2% of the data.
Agents falter in partially observable settings, revealing that merely gathering visual evidence isn't enough without robust perceptual state management.
Agents can discover tool behaviors but fail to adapt effectively, often resorting to inefficient exhaustive searches in dynamic environments.
$\Sigma$-Mem not only tracks agent reliability but also adapts dynamically to feedback, outperforming traditional methods in multi-agent coordination.
IDEAgent achieves a staggering 3.89x improvement in generating diverse and high-quality research ideas compared to existing methods.
LLMs may create a "novelty mirage," consistently rating generated research questions as innovative, while human experts disagree.
GRAIL reweights token advantages in reinforcement learning, leading to significant accuracy gains without the need for expensive process-level supervision.
By explicitly grounding reasoning steps to visual objects, Chain-of-Glimpse enables more accurate and interpretable video understanding, outperforming object-agnostic methods on multiple benchmarks.
Current VLMs can ace image quizzes, but completely fumble when asked to stack blocks in a physically plausible way, revealing a critical gap in understanding real-world physics.
