Search papers, labs, and topics across Lattice.
This paper investigates the efficacy of linguistic rules as standalone prompt compressors for large language models (LLMs), eliminating the need for costly LM forward passes during compression. By employing offline evolutionary search to identify optimal combinations of lexical, syntactic, semantic, and discourse cues, the authors demonstrate that these linguistic rules can achieve competitive performance compared to advanced prompt-compression strategies. The findings reveal that while performance is robust under moderate compression, it declines with more aggressive compression, highlighting the nuanced interplay between linguistic signals at varying compression levels.
Linguistic rules can replace expensive LM scoring in prompt compression, achieving comparable performance with significantly lower computational costs.
Prompt compression shortens LLM input to reduce inference cost, yet existing methods score token importance through LM forward passes. It remains questionable whether such nuanced, costly token selection is necessary. Compression requires identifying informative content, a problem that linguistic research has long addressed through cues that can be operationalized as deterministic rules. We therefore ask: can \textbf{linguistic rules alone} serve as effective prompt compressors, without LM-based scoring at compression time? To address this, we conduct offline evolutionary search over lexical, syntactic, semantic, and discourse seeds to find competitive rule combinations. The resulting linguistic compressor requires no LM forward pass at deployment and uses only CPU-side processing for compression. We evaluate it with a dual-path protocol to balance compression quality and reconstruction fidelity. Across short passages, multi-document reasoning, and dialogue-memory QA datasets, evolved compressors achieve performance similar to that of recent advanced prompt-compression strategies. Performance is strongest under light-to-moderate compression and degrades as compression becomes more aggressive, while the Direct and Reconstruction paths exhibit distinct patterns. Evolutionary analysis reveals that effective compression fuses signals across linguistic levels and, as the compression ratio increases, rules shift from token pruning to sentence extraction.
