Search papers, labs, and topics across Lattice.
Affiliation:
11
0
19
TrajMind achieves a remarkable 41.1% reduction in latency while delivering over 15 percentage points improvement in anomaly diagnosis accuracy.
A learnable moderator can transform multi-agent debates into more efficient reasoning processes, outperforming traditional methods by reducing redundancy and enhancing evidence aggregation.
Persona conditioning in LLMs can be exploited to amplify inference costs by over 200 times, revealing a critical vulnerability in their deployment.
Trip delays in autonomous vehicle fleets are not just averages; they reveal deep inequities that can be mitigated with smarter routing strategies.
PolyUQuest achieves superior answer correctness and verifiability by integrating structural web data into its retrieval process, outperforming traditional RAG systems.
SrDetection uncovers hidden data leakage patterns in Code LLMs, boosting detection accuracy by over 21 points without relying on fragile heuristics.
User feedback prediction can be learned, revealing critical performance gaps in AI assistants that traditional evaluation methods overlook.
Dynamic allocation of compute resources based on attention entropy can yield significant speedups in long-context LLM inference without sacrificing quality.
Reinforcement learning can efficiently merge heterogeneous language models in federated ASR, outperforming genetic algorithms and improving character error rate.
CiteLLM offers a LaTeX-integrated agent that grounds claims in trusted academic sources, bypassing LLM hallucinations by using them only for search and ranking, not content generation.
Automate customer service without complex orchestration or sacrificing privacy by using task-oriented flowcharts to guide small, locally deployed language models.
