Research Preview

WaveTransformer
O(n) Attention
via Wave Propagation

Replace quadratic self-attention with wave propagation on a learned lattice. Linear scaling in sequence length. No KV cache needed. 5.8x better perplexity at matched compute.

O(n) scalingNo KV cachePyTorch nativeValidated on language modeling
5.8x
Better Perplexity
vs standard attention at iso-compute
O(n)
Time Complexity
Linear in sequence length
0
KV Cache
No memory bottleneck
4x
Fewer Parameters
At matched performance

How It Compares

Standard attention scales quadratically. WaveTransformer replaces the attention matrix with wave propagation, achieving linear scaling.

PropertyStandard TransformerWaveTransformer
Time complexityO(n²)O(n)
Memory (inference)O(n²) KV cacheO(n) lattice state
Long sequencesQuadratic wallLinear scaling
Parameter efficiencyBaseline4x fewer at iso-quality
Perplexity (iso-compute)Baseline5.8x better

Target Applications

Long-Document Processing

Process 100K+ token documents without quadratic memory blowup. Legal, medical, scientific texts.

Edge / Mobile Inference

No KV cache means dramatically lower memory footprint. Run on devices that can't hold standard transformer state.

Real-Time Streaming

O(n) per-token cost enables true streaming without growing memory. Chat, transcription, live analysis.

Efficient Pre-Training

4x fewer parameters at matched quality. Train larger effective models on the same hardware budget.

Retrieval-Augmented Gen

Process retrieved passages linearly. No quadratic penalty for adding more context documents.

Scientific Simulation

Sequence-to-sequence modeling of physical systems with inherently long temporal correlations.

Current Status

WaveTransformer has been validated on small-scale language modeling tasks (E15a/b experiments). Large-scale validation and public release forthcoming.

Core architecture validated (5.8x PPL improvement)
O(n) scaling confirmed experimentally
PyTorch implementation complete
Large-scale benchmark suite
Public model weights release
HuggingFace integration

Want early access?

We're offering early benchmark access and consulting for organizations that want to evaluate WaveTransformer on their workloads.