---
format: "aidr-story-markdown/v1"
id: "9965cd0c4f04d783a5506169afefb3faea9c3d7ed1e83b8eb0a5fda9cb375659"
canonical_url: "https://aidr.today/9965cd0c?lang=en"
title: "An Input-Dependent Readout Phase for Resonant Sequence Models"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-12T04:00:00.000Z"
category: "Research"
topics: ["llm"]
source_urls: ["https://arxiv.org/abs/2610.12496"]
summary: "ResonatorLM uses a fixed learned phase per head to read a bank of damped complex oscillators, and making that readout phase a function of the current token adds input dependence to the read while keeping the convolution exact and costing 0.2 percent more parameters. In 6M-parameter character-level models on WikiText-2, the input-dependent phase matches the original at the training length and reduces perplexity from 12.81 to 3.76 when trained on 256 tokens and tested on 1024."
---

# An Input\-Dependent Readout Phase for Resonant Sequence Models

> [Open the canonical story](<https://aidr.today/9965cd0c?lang=en>)

**Published:** 2026-10-12T04:00:00.000Z
**Category:** Research
**Topics:** llm

## Summary

ResonatorLM uses a fixed learned phase per head to read a bank of damped complex oscillators, and making that readout phase a function of the current token adds input dependence to the read while keeping the convolution exact and costing 0\.2 percent more parameters\. In 6M\-parameter character\-level models on WikiText\-2, the input\-dependent phase matches the original at the training length and reduces perplexity from 12\.81 to 3\.76 when trained on 256 tokens and tested on 1024\.

## Sources

- [Story source](<https://arxiv.org/abs/2610.12496>)