Papers/2609.13154
🧪 Test?View on arXiv

Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task Categories

Not provided

prompt engineeringcompressionNLPdeterministic methods
2609.13154
Builder Relevance
80%
2h ago

Abstract

This paper explores a training-free, deterministic pipeline for compressing prompts for large language models, achieving significant token reduction while maintaining output quality.

Reality Card

Core Claim

The most aggressive configuration of the proposed pipeline achieves a mean token reduction of 40.3% with a BERTScore-F1 of 0.876.

Method / Result

40.3% mean token reduction at BERTScore-F1 of 0.876.

Limitations

The study indicates commonsense reasoning as a systematic failure mode under aggressive compression, which may affect generalizability.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.
← Back to all papers