Search papers, labs, and topics across Lattice.
3
0
6
0
Current research agents still struggle with retrieval robustness and hallucination control, even when evaluated in a static, verifiable research environment.
LLMs still fail to follow complex instructions that entangle content, formatting, control flow, and real-world constraints, despite progress on simpler benchmarks.
LLMs can reason better on graphs if you teach them to selectively extract and denoise subgraphs, outperforming one-size-fits-all approaches in zero-shot settings.
