苹果研究：语言模型树结构表达式序列化的通信瓶颈与往返研究
The Communication Bottleneck: A Round-Trip Study of Tree-Structured Expression Serialization in Language Models
苹果研究者提出往返协议，用生成器把算术表达式转成文字题、再由独立提取器还原，并以符号等价作为精确判定，测试十六个模型的所有两两组合。结果显示该通道有损且不对称：交换生成与提取模型可使准确率相差最多 60.4 个百分点，最佳组合达 92.9%；至少 73.6% 的失败源于生成端，难度由树结构而非模型家族决定。
AuthorsXavier Suau, Alex Ferrando de las Morenas, Luca Zappella, Samy Bengio
When language models reason in chain-of-thought or exchange free-text intermediates, they serialize structured information into natural language. How much tree-structured compositional content survives this bottleneck? We propose a round-trip protocol that answers this question empirically for tree-structured expressions. A generator converts a procedurally generated arithmetic expression into a word problem, a separate extractor recovers the expression from the word problem alone, and symbolic equivalence provides an exact oracle. Evaluating all pairwise combinations of sixteen models yields a communication matrix whose marginals separate generation quality from extraction quality. Three main findings emerge. First, the channel is lossy and asymmetric: swapping which model generates and which extracts shifts accuracy by up to 60.4 points, and the best pair reaches 92.9% by combining different models on each end rather than the same model on both. Second, at least 73.6% of round-trip failures originate at generation, and difficulty is driven by tree structure (operator count, depth, right-branching) rather than model family. Third, the channel is trainable: ∼ 3600 fine-tuning examples that share the evaluation’s operators and tree shapes lift every open-weight model above untrained Gemini-3.1-Pro, an upper bound under matched semantics. A disjoint-domain regime with new operators and vocabulary also raises every open-weight model, confirming the gain is not an artifact of matched semantics, though a gap to the frontier remains. Together these results identify tree-structured expression serialization as a primary limiting factor when models communicate hierarchical structure through natural language.
Related readings and updates.
With Apple Intelligence, we’re integrating powerful generative AI right into the apps and experiences people use every day, all while protecting their privacy. At the 2025 Worldwide Developers Conference we introduced a new generation of language foundation models specifically developed to enhance the Apple Intelligence features in our latest software releases. We also introduced the new Foundation Models framework, which gives app developers…
Lightweight syntactic analysis tools like Semgrep and Comby leverage the tree structure of code, making them more expressive than string and regex search. Unlike traditional language frameworks (e.g., ESLint) that analyze codebases via explicit syntax tree manipulations, these tools use query languages that closely resemble the source language. However, state-of-the-art matching techniques for these tools require queries to be complete and…
来源：Apple Machine Learning Research（RSS） · machinelearning.apple.com