Search papers, labs, and topics across Lattice.
Affiliation:
9
0
15
First-principles simulations can now accurately model biomolecular systems with over 100,000 atoms at a fraction of the cost.
E2E speech models are vulnerable to a stealthy DoS attack that can drastically increase their output length and resource consumption without altering the original input.
EHHN achieves up to 12.4 percentage points improvement in next activity prediction accuracy while slashing peak GPU memory usage by 24 times compared to traditional methods.
Navigating with fewer than 8 VLM calls per episode, Goal2Pixel redefines efficiency in vision-language navigation tasks.
LLMs can reason more efficiently by sharing intermediate thoughts during parallel search, achieving better accuracy with less computation.
Achieve more realistic and accurate trajectory predictions by modeling agents as cognitive entities that infer beliefs and align with social norms, even with limited information.
By treating a cluster's DRAM as a single cache, DPC slashes data redundancy and coherence overhead, achieving up to 12.4x speedups.
Domain-specific LLM applications can consume surprisingly more energy than generic LLMs, especially when designed with complex, agentic RAG pipelines for enhanced accuracy.
Generalizing to unseen compositions? This plug-and-play method leverages structure in the embedding space to adapt prompts, significantly boosting open-vocabulary zero-shot learning.
