AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Distribution-Aware Multi-Granularity Phase Coding: Towards Lower Conversion Error for Spike-Driven Large Language Models

Hanyuan Zheng, Haozhen Zhang, Tianshuo Chen, Zhaogeng Liu, Yi Chang, Bin Gu

ICLR 2026 · Architectures and efficiency › Spiking networks · approach: Spiking transformers

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add neuron training with multi-granularity phase coding over LAS (untrained special case)LAS · LLaMA-2-7B T=8avg PPL: 31.73 -> 7.16Improved
add neuron training over LASLAS · LLaMA-3-8B T=8avg PPL: 8.16 -> 7.22Improved
use training-free conversion instead of SpikeLLM's trainingOurs vs SpikeLLM · LLaMA-2-7Btime cost: 5h 54m -> 2m 01sImproved
increase timestep T from 8 to 10Ours (Grain=2) · LLaMA-2-7Bavg PPL: 7.16 -> 5.73Improved
use multiple granularities (3) instead of one (ablation)Ours · LLaMA-2-7B T=8avg ACC: 67.03 -> 67.31Improved
use multiple granularities (2) instead of one (ablation)Ours · LLaMA-2-7B T=8avg PPL: 7.01 -> 7.16Hurt
set number of granularities equal to T (ablation)Ours · LLaMA-3-8B T=8avg ACC: 71.40 vs 71.48No clear effect
add alternating optimization instead of joint (ablation)Ours · LLaMA-3-8B 6/3ArcC: 47.18 -> 49.23Improved
add alternating optimization instead of joint (ablation)Ours · LLaMA-2-7B 8/2Wikitext2 PPL: 6.58 -> 6.71Hurt
decouple h and theta (ablation)Ours · LLaMA-2-7B 8/2Wikitext2 PPL: 7.62 -> 6.71Improved
decouple h and theta (ablation)Ours · LLaMA-3-8B 6/3Wikitext2 PPL: 7.63 -> 8.53Hurt
quantize weights to 4 bitOurs (Grain=2) · LLaMA-2-7B T=8Wikitext2 PPL: 7.56 -> 8.91Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy