# Beyond Sequence Order: Syntax-Informed Positional Embeddings for Transformers

arXiv: https://arxiv.org/abs/2608.06111 · HTML: https://mlsanity.com/paper/2608.06111/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 35 references that are in the map.
- First posted on arXiv 6 Aug 2026.

## Builds on (most cited first, 23 of 23)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://mlsanity.com/paper/2006.03654/) (2020)
- [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://mlsanity.com/paper/1901.02860/) (2019)
- [DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing](https://mlsanity.com/paper/2111.09543/) (2021)
- [Self-Attention with Relative Position Representations](https://mlsanity.com/paper/1803.02155/) (2018)
- [BERT Rediscovers the Classical NLP Pipeline](https://mlsanity.com/paper/1905.05950/) (2019)
- [BLiMP: The Benchmark of Linguistic Minimal Pairs for English](https://mlsanity.com/paper/1912.00582/) (2019)
- [Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference](https://mlsanity.com/paper/2412.13663/) (2024)
- [A Systematic Assessment of Syntactic Generalization in Neural Language Models](https://mlsanity.com/paper/2005.03692/) (2020)
- [Linguistically-Informed Self-Attention for Semantic Role Labeling](https://mlsanity.com/paper/1804.08199/) (2018)
- [Transformer Grammars: Augmenting Transformer Language Models with Syntactic Inductive Biases at Scale](https://mlsanity.com/paper/2203.00633/) (2022)
- [Hierarchical Poset Decoding for Compositional Generalization in Language](https://mlsanity.com/paper/2010.07792/) (2020)
- [Structural Guidance for Transformer Language Models](https://mlsanity.com/paper/2108.00104/) (2021)
- [Pushdown Layers: Encoding Recursive Structure in Transformer Language Models](https://mlsanity.com/paper/2310.19089/) (2023)
- [Dependency Transformer Grammars: Integrating Dependency Structures into Transformer Language Models](https://mlsanity.com/paper/2407.17406/) (2024)
- [Tree-Planted Transformers: Unidirectional Transformer Language Models with Implicit Syntactic Supervision](https://mlsanity.com/paper/2402.12691/) (2024)
- [Sneaking Syntax into Transformer Language Models with Tree Regularization](https://mlsanity.com/paper/2411.18885/) (2024)
- [Composition, Attention, or Both?](https://mlsanity.com/paper/2210.12958/) (2022)

## Abstract (arXiv)

Positional embeddings (PE) in Transformers encode token distance and order but are largely agnostic to textitsyntactic structure. We introduce Syntax-informed Positional Embeddings (SiPE), which learns a lightweight syntactic prior from dependency parses during pretraining and injects it across all three dominant PE families (absolute, relative, rotary), for both encoders and decoders, leaving self-attention and the rest of the architecture untouched. We isolate where and how the prior should enter the model, and find it depends on the architecture: for autoregressive decoders that use relative PE, the prior is strongest when coupled multiplicatively with the relative-position term of the attention score, outperforming injection into the input embeddings, into self-attention, or into the positional and attention terms jointly---while for encoders it is best added directly to the input embeddings, composing with each encoder's native positional mechanism. We find that models pre-trained with SiPE improve on the SyntaxGym benchmark by up to 10.3% while simultaneously reducing perplexity by 9.0% over a base model with no syntactic supervision---a metric nearly every existing syntax-injection method instead degrades. Crucially, these gains extend beyond syntactic generalization: SiPE also improves real-world language understanding, raising scores on the GLUE benchmark by up to 8.2% over a model trained without it. Unlike existing syntactic language models that marginalize over many parses at inference or discard syntax at runtime, SiPE conditions on a single parse, establishing a new Pareto frontier between syntactic supervision and inference cost.