- Score 83
- Official
SPLICE: Structured Prompt Local Iterative Combinatorial Evolution
arXiv:2611.18883v1 Announce Type: new Abstract: Natural language optimization (NLO) improves an agent by having a model critique and rewrite its natural language policy parameters: system prompts, tool instructions, reasoning directives, few-shot demonstrations. These artifacts are multi-section documents in which some sections encode contracts (safety clauses, tool schemas, output formats) that m
So what
What this event means by reading role—not a longer recap.
- BuilderIf you auto-tune system prompts, SPLICE's per-block mutability could let you optimize freely while pinning tool schemas and output formats verbatim.
- ResearcherIt maps three known NLO failure modes to formal guarantees and reports wins on four of seven benchmarks, a concrete baseline for prompt-search work.
Score dimensions
Higher total means read first. Each bar is one factor we use to rank the system pool. How we score