01Entry
PARSER: Read in Parallel, Reason in Depth for Long-Context LLM Agents
PARSER splits long-document QA into parallel chunk reads plus iterative scatter-gather reasoning — targets the accuracy cliff agents hit on multi-hop questions past ~100k tokens. Directly relevant if your agents do retrieval over long docs.