Model

Qwen3.8 Flash-Next

A large-scale pretrained language model in Alibaba's Qwen series, incorporating a PLE memory module with approximately 51 billion parameters, designed for encoding and transferring n-gram statistical information.

Timeline (last 90 days)

Sep 25

Transferring the PLE (n-gram memory) pre-trained by Qwen3.8 Flash-Next to the Qwen3.5-0.8B model achieved a 5.05% perplexity reduction on the frozen validation set.

Unverified50%

All Facts (1)

Unverified

Transferring the PLE (n-gram memory) pre-trained by Qwen3.8 Flash-Next to the Qwen3.5-0.8B model achieved a 5.05% perplexity reduction on the frozen validation set.

50%

Source Articles

Qwengram-0.8B Experiment: Cross-Model N-gram Memory Transfer Reduces Perplexity by 5.05%Sep 25