# Orthographic Constraint Satisfaction and Human Difficulty Alignment in Large Language Models

arXiv: https://arxiv.org/abs/2511.21086 · HTML: https://mlsanity.com/paper/2511.21086/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 30 references that are in the map.
- First posted on arXiv 26 Nov 2025.

## Builds on (most cited first, 14 of 14)

- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Faith and Fate: Limits of Transformers on Compositionality](https://mlsanity.com/paper/2305.18654/) (2023)
- [Lexically Constrained Decoding for Sequence Generation Using Grid Beam Search](https://mlsanity.com/paper/1704.07138/) (2017)
- [NeuroLogic A*esque Decoding: Constrained Text Generation with Lookahead Heuristics](https://mlsanity.com/paper/2112.08726/) (2021)
- [Interpreting Language Models with Contrastive Explanations](https://mlsanity.com/paper/2202.10419/) (2022)
- [Puzzle Solving using Reasoning of Large Language Models: A Survey](https://mlsanity.com/paper/2402.11291/) (2024)
- [Flexible and Efficient Grammar-Constrained Decoding](https://mlsanity.com/paper/2502.05111/) (2025)
- [CUTE: Measuring LLMs' Understanding of Their Tokens](https://mlsanity.com/paper/2409.15452/) (2024)
- [Why is constrained neural language generation particularly challenging?](https://mlsanity.com/paper/2206.05395/) (2022)
- [Tokenization and Morphology in Multilingual Language Models: A Comparative Analysis of mT5 and ByT5](https://mlsanity.com/paper/2410.11627/) (2024)
- [Language Models are Crossword Solvers](https://mlsanity.com/paper/2406.09043/) (2024)
- [Beyond Film Subtitles: Is YouTube the Best Approximation of Spoken Vocabulary?](https://mlsanity.com/paper/2410.03240/) (2024)
- [Multilingual LLMs Struggle to Link Orthography and Semantics in Bilingual Word Processing](https://mlsanity.com/paper/2501.09127/) (2025)
- [Unmasking the Imposters: How Censorship and Domain Adaptation Affect the Detection of Machine-Generated Tweets](https://mlsanity.com/paper/2406.17967/) (2024)

## Abstract (arXiv)

Large language models must satisfy hard orthographic constraints during controlled text generation, yet systematic cross-family evaluation remains limited. We evaluate 39 configurations spanning three model families (Qwen3, Claude Haiku 4.5, GPT-5-mini) on 58 word puzzles requiring character-level constraint satisfaction. Cross-family differences produce substantially larger performance gaps (2.0-2.2x, F1 = 0.761 vs. 0.343) than parameter scaling within families (83% gain from 4B to 32B scaling), and a partial-correlation analysis rules out tokenizer design as a confound for within-family scaling. Thinking budget sensitivity proves heterogeneous: high-capacity models show strong returns (+0.102 to +0.136 F1), while mid-sized variants saturate or degrade, showing inconsistent compute benefits. Using difficulty ratings from 10,000 human solvers per puzzle, we establish modest but consistent calibration (rho = 0.28-0.42) across all families, yet identify systematic failures on common words with unusual orthography ("data", "loll", "acai": 83-91% human success, 94-98% model miss rate). These failures point to over-reliance on distributional plausibility that penalizes orthographically atypical but constraint-valid patterns.