# Out-of-Length Scene Text Recognition: A Two-Axis Diagnosis and a Training-Free Fix

arXiv: https://arxiv.org/abs/2607.23194 · HTML: https://mlsanity.com/paper/2607.23194/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 31 of its 46 references that are in the map.
- First posted on arXiv 25 Jul 2026.

## Builds on (most cited first, 30 of 31)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation](https://mlsanity.com/paper/2108.12409/) (2021)
- [Extending Context Window of Large Language Models via Positional Interpolation](https://mlsanity.com/paper/2306.15595/) (2023)
- [An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition](https://mlsanity.com/paper/1507.05717/) (2015)
- [Synthetic Data for Text Localisation in Natural Images](https://mlsanity.com/paper/1604.06646/) (2016)
- [Rotary Position Embedding for Vision Transformer](https://mlsanity.com/paper/2403.13298/) (2024)
- [Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition](https://mlsanity.com/paper/2103.06495/) (2021)
- [What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis](https://mlsanity.com/paper/1904.01906/) (2019)
- [Scene Text Recognition with Permuted Autoregressive Sequence Models](https://mlsanity.com/paper/2207.06966/) (2022)
- [Reading Text in the Wild with Convolutional Neural Networks](https://mlsanity.com/paper/1412.1842/) (2014)
- [Round and Round We Go! What makes Rotary Positional Encodings useful?](https://mlsanity.com/paper/2410.06205/) (2024)
- [SVTR: Scene Text Recognition with a Single Visual Model](https://mlsanity.com/paper/2205.00159/) (2022)
- [Vision Transformer for Fast and Efficient Scene Text Recognition](https://mlsanity.com/paper/2105.08582/) (2021)
- [ESIR: End-to-end Scene Text Recognition via Iterative Image Rectification](https://mlsanity.com/paper/1812.05824/) (2018)
- [Rethinking Text Line Recognition Models](https://mlsanity.com/paper/2104.07787/) (2021)
- [Reading and Writing: Discriminative and Generative Modeling for Self-Supervised Text Recognition](https://mlsanity.com/paper/2207.00193/) (2022)
- [Revisiting Scene Text Recognition: A Data Perspective](https://mlsanity.com/paper/2307.08723/) (2023)
- [Length Extrapolation of Transformers: A Survey from the Perspective of Positional Encoding](https://mlsanity.com/paper/2312.17044/) (2023)
- [SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition](https://mlsanity.com/paper/2411.15858/) (2024)
- [Context Perception Parallel Decoder for Scene Text Recognition](https://mlsanity.com/paper/2307.12270/) (2023)
- [LISTER: Neighbor Decoding for Length-Insensitive Scene Text Recognition](https://mlsanity.com/paper/2308.12774/) (2023)
- [Out of Length Text Recognition with Sub-String Matching](https://mlsanity.com/paper/2407.12317/) (2024)
- [On the Generalization of Handwritten Text Recognition Models](https://mlsanity.com/paper/2411.17332/) (2024)
- [Spiral RoPE: Rotate Your Rotary Positional Embeddings in the 2D Plane](https://mlsanity.com/paper/2602.03227/) (2026)
- [Context-aware Rotary Position Embedding](https://mlsanity.com/paper/2507.23083/) (2025)
- [Improving Long Handwritten Text Line Recognition with Convolutional Multi-way Associative Memory](https://mlsanity.com/paper/1911.01577/) (2019)
- [A 2D Semantic-Aware Position Encoding for Vision Transformers](https://mlsanity.com/paper/2505.09466/) (2025)

## Abstract (arXiv)

Scene Text Recognition (STR) models are trained almost exclusively on word crops of at most 25 characters, yet real deployments (signage, product labels, dense captions) require reading much longer text. This paper diagnoses that failure and then closes it. The diagnosis separates out-of-length failure into two simultaneously extrapolating axes (the encoder's width axis and the decoder's time axis) and shows that encoder width, not decoder length, is the dominant failure mode. Representation-side fixes bring only partial relief: training-free rotary rescalings recover at most 2-4 points of character error rate (CER), and a weighted fine-tuning recipe recovers 6-8 points while improving standard-benchmark accuracy, yet word accuracy on the Long Text Benchmark (LTB) stays near zero, because the residual gap lies in the decoding mechanism rather than the representation. We then close that gap at inference time, on an unmodified word-level checkpoint: the long image is sliced into overlapping crops at the model's training width, each decoded independently and in-distribution, and the reads stitched by geometry-anchored edit-distance alignment. This procedure reaches 42.79-43.05% bucket-average word accuracy on LTB across two base checkpoints, matching the published state of the art (41.57%) and beating it by 11-12 points on the hardest bucket, at wall-clock parity with plain decoding; applied unchanged to the public PARSeq checkpoint it reaches 47.11%. Once chunking is applied fine-tuning no longer helps: the decoding-side fix alone matches purpose-built architectures. We release the diagnosis harness and implementation.