Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
4
0
5
19
Rollback-Induced Reflection (RIR) is proposed, a unified recovery framework that restores execution to a selected prior state while carrying forward reusable knowledge distilled from the abandoned trajectory to guide subsequent decisions.
Task learnability can significantly enhance RL efficiency in LLMs, leading to better performance with less data.
BOUND achieves up to 5.6 EM points improvement over traditional methods by correcting local search-control errors in real-time.
Self-evolving LLM agents can drastically reduce reasoning overhead by transforming atomic actions into reusable Standard Operating Procedures, leading to higher success rates and fewer interaction rounds.