Search papers, labs, and topics across Lattice.
Peking University, Affiliation:
3
0
7
3
Adapters forget as much as they learn: a simple post-training normalization unlocks significant gains in both specialization and retained capabilities for LoRA.
Enhancing on-policy reinforcement learning can paradoxically reduce the diversity of successful behaviors, leading to a trade-off that challenges future trainability.
AudioLLMs don't have to choose between reasoning and perception: a unified schema can boost both.
