Reading Without a Reader: Large Language Models Collapse Reading and Writing into a Single Entangled Code

2026/07/02 by Diego Saldaña Ulloa
Biochemistry, Genetics and Molecular Biology · Computer Science · #cs.AI #cs.CL #cs.LG #q-bio.NC

paper · pdf

arxiv created 2026/07/30 · arxiv updated 2026/07/31

Abstract

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 cultural invention, not an evolved instinct). We ask how entangled it is, comparing an input-side ``reading code'' W<sub>E</sub> with an output-side ``writing code'' W_U via an index E∈[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 coupled but sub-ceiling code (E=0.23--0.35, far above floor) on a non-monotonic couple-then-differentiate trajectory, W_U drifting ∼3.2× farther than 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.

Citations

• Leveraging Human Production-Interpretation Asymmetries to Test LLM Cognitive Plausibility
• Palatable Conceptions of Disembodied Being: Terra Incognita in the Space of Possible Minds
• Understanding Token Probability Encoding in Output Embeddings
• The Platonic Representation Hypothesis
• Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokens
• Similarity of Neural Network Models: A Survey of Functional and Representational Measures
• Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling
• Eliciting Latent Predictions from Transformers with the Tuned Lens
• Talking About Large Language Models
• Talking about Large Language Models
• OPT: Open Pre-trained Transformer Language Models
• Training Compute-Optimal Large Language Models
• Quantifying Memorization Across Neural Language Models
• On the Dangers of Stochastic Parrots
• Extracting Training Data from Large Language Models
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• Experience Grounds Language
• Scaling Laws for Neural Language Models
• Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
• RoBERTa: A Robustly Optimized BERT Pretraining Approach
• What Does BERT Look At? An Analysis of BERT's Attention
• Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned
• Similarity of Neural Network Representations Revisited
• Are Sixteen Heads Really Better than One?
• Graph Attention Networks
• Attention Is All You Need
• Offline bilingual word vectors, orthogonal transformations and the inverted softmax
• Tying Word Vectors and Word Classifiers: A Loss Framework for Language Modeling
• Using the Output Embedding to Improve Language Models
• Toward a mechanistic psychology of dialogue
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Related

• Lost in Translation: The Algorithmic Gap Between LMs and the Brain
• Large Language Models Are Human-Like Internally
• Neuroscience and Literacy: An Integrative View
• Brain-Like Language Processing via a Shallow Untrained Multihead Attention Network
• Mind Scramble: Unveiling Large Language Model Psychology Via Typoglycemia