Search papers, labs, and topics across Lattice.
Institute of Software, Chinese Academy of Sciences
3
0
7
3
Bash-Commenter achieves a remarkable 33.40% BLEU-4 score, setting a new benchmark for automated comment generation in Bash scripting.
Value function approximation in continual RL gets a boost: dynamically rehearsing Q-values and regularizing from the start significantly reduces catastrophic forgetting in multi-cyclic environments.
Forget struggling with cryptic SQL: a new LLM fine-tuned with human preferences generates comments so good, they beat Qwen3-14B by up to 13% on standard metrics.
