用符号化结构重写文本，让小模型更高效答题。
Context Compression Is Not One Thing: Readable Symbolic Re-expression vs. Coherent Summary at Matched Budget
- 将检索文本转为结构化实体关系句，降低令牌开销。
- 在多个数据集上提升13至20个F1百分点，优于三种压缩方法。
- 适合资源受限场景下提升多跳问答性能的实践者。
我们研究小语言模型在多跳问答任务中的上下文压缩问题。提出Telegraph English，一种可读的符号化格式，将检索到的段落重写为结构化的实体-关系陈述，在更低的令牌成本下保留推理证据。在MuSiQue、TwoWiki和HotpotQA的受控实验中，Telegraph English在每个数据集上均优于三种同预算压缩基线（字符级删除、截断、随机子采样），F1提升达13至20个百分点。它还优于同一编码器生成的连贯段落摘要，尤其在最难的数据集上表现更优。预注册的深度-交互假设未成立：优势不随推理深度增加而增长。结果表明，在相同令牌预算下，可读的符号化重写比自然语言或连贯摘要更密集地保留实体内容。
原文摘要 · Abstract (English)
We study context compression for multi-hop question answering with small language models. We propose Telegraph English, a readable symbolic format that rewrites retrieved passages into structured entity-relation statements, preserving reasoning evidence at lower token cost. In controlled experiments on MuSiQue, TwoWiki, and HotpotQA, Telegraph English outperforms three matched-budget compression baselines (character-level deletion, truncation, and random sub-sampling) on every dataset, with gains of 13 to 20 F1 percentage point. It also outperforms a coherent prose summary produced by the same encoder on the hardest dataset. A pre-registered depth-interaction hypothesis is null: the advantage does not grow with reasoning depth within datasets. We interpret these results as evidence that readable symbolic re-expression preserves entity content more densely than either natural language or coherent summarization at matched token budget.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。