大模型的读写能力本质纠缠，无法分离。
Reading Without a Reader: Large Language Models Collapse Reading and Writing into a Single Entangled Code
- 用统一自回归路径实现读写，共享一个编码机制。
- 读写编码相似度达0.23~0.35，远高于随机水平。
- 读写不可分离，适合研究认知架构与语言模型关系者阅读。
在人类大脑中，阅读与书写具有双重分离性：腹侧解码通路（纯失读症）和前额顶叶编码通路（纯失写症）共享部分字形核心。而仅含解码器的大语言模型（LLM）通过单一自回归路径同时完成读写任务，该路径基于文本训练（一种文化发明，非演化本能）。我们通过指数 𝒞∈[0,1]（使用 CKA、Procrustes 残差、互k-NN）对比输入侧“阅读编码”𝑊𝐸 与输出侧“写作编码”𝑊𝑈，校准独立初始化基线与固定上限。在 GPT-2、OPT 与 Pythia（14M–1.4B 参数）上，未绑定模型表现出一个耦合但未达上限的代码（𝒞=0.23–0.35，显著高于基线），且𝑊𝑈 在每十分位上漂移约3.2倍于𝑊𝐸。更关键的是：理解与生成行为无法分离，无法进行类似失读/失写症的对照测试。对最小词对而言，失读症类比在理论上为空——贪婪生成意味着全局argmax，从而胜出配对排名。差异损伤指标不具备尺度可识别性：重尾损伤导致线性标准化坍缩至较大项，秩变换虽能修正却存在饱和，其零假设自动饱和。两项指标均包含目标词的对数概率，后者即可解释其方差绝大部分并制造虚假耦合。因此，我们撤回耦合统计量、跨层级桥梁与分离度量。在一个读写共用下一个词分布的模型中，不存在输出侧成对隔离任一能力的可能：纠缠无需指数即可显现。这提示我们，大模型应以类比而非同源性，置于可能心智空间之中。
原文摘要 · Abstract (English)
In the literate human brain, reading and writing doubly dissociate: a ventral decoding route (pure alexia) and a fronto-parietal encoding route (pure agraphia), sharing a partial orthographic core. A decoder-only large language model (LLM) drives both from one autoregressive path optimized on text (a \emph{cultural} invention, not an evolved instinct). We ask how entangled it is, comparing an input-side ``reading code'' $\mathbf{W}_{E}$ with an output-side ``writing code'' $\mathbf{W}_{U}$ via an index $\mathcal{E}\in[0,1]$ (CKA, Procrustes residual, mutual $k$-NN) calibrated against an independent-init floor and tied ceiling. On GPT-2, OPT and Pythia (14M--1.4B), untied models hold one \emph{coupled but sub-ceiling} code ($\mathcal{E}=0.23$--$0.35$, far above floor) on a non-monotonic couple-then-differentiate trajectory, $\mathbf{W}_{U}$ drifting $\sim$3.2$\times$ farther than $\mathbf{W}_{E}$ in every decile. Equally informative is a negative: the matching behavioural test, that comprehension and production fail together rather than dissociate, cannot be run. For minimal pairs the alexia analogue is empty by theorem: greedy production implies a vocabulary-wide argmax, so it wins the pairwise ranking. Differential-damage indices are not scale-identified: heavy-tailed damage makes linear standardizations collapse onto their larger term, and the rank transform fixing this is bounded, so its null saturates. Both scores also contain the target's log-probability, which alone explains most of their variance and manufactures the apparent coupling. We withdraw a coupling statistic, a cross-level bridge and a separation measure. In a model reading and writing off one next-token distribution, no output-side pair isolates either ability: entanglement needing no index to see. By analogy, not homology, this situates LLMs in the space of possible minds.
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。