Search papers, labs, and topics across Lattice.
7
0
12
Achieve state-of-the-art robustness in conversational emotion recognition by distilling knowledge from a complete-view teacher model, even when modalities are missing or conflicting.
Deterministic LLM inference gets a 2x speedup by verifying only the 1% of tokens with shaky confidence.
LLMs that can generate HTML are finally useful: HTMLCure's closed-loop repair engine turns superficially correct but broken pages into high-quality training data, rivaling the performance of much larger models.
MLLMs can achieve 10% gains on multimodal reasoning benchmarks by using ground-truth anchored data curation and scaffold-stripping to avoid cognitive drift during self-evolution.
Jointly ranking answers to complex knowledge graph queries with multiple free variables is now tractable, thanks to a neural-symbolic search method that avoids full enumeration.
Forget agonizing over checkpointing and restarts: LiveR slashes LLM training downtime from minutes to seconds by hot-swapping model state between parallel training worlds.
Explicitly modeling the dependency between dialogue context and current utterance as an "interpretation cue" significantly boosts conversational multimodal understanding.
