Search papers, labs, and topics across Lattice.
2
0
4
3
Standard LLM agents lose critical plan information when evicted from context, leading to a staggering 34.7% drop in task success rates.
Current multimodal agents navigating document collections achieve human-level accuracy through brute-force search, highlighting a critical gap in strategic reasoning and planning.
