Qwen3.5-0.8B × TinyCeNN Memory Fusion

Experimental accepted-only Memory Fusion adapter for Qwen/Qwen3.5-0.8B. Native Qwen3.5 Gated DeltaNet layers remain unchanged.

Target anchors: [3, 7, 11, 15, 19, 23]
Accepted anchors: [3, 7]
Base revision: 2fc06364715b967f1860aea9cf38778875588b17
Source commit: 3ac52ab6495f6872fdf5f50af4cb433e0542a2bc

Acceptance gates: NMSE ≤ 0.20, cosine ≥ 0.90, incremental ΔNLL ≤ 0.015, cumulative ΔNLL ≤ 0.05. chat_examples.json contains deterministic original-vs-Memory-Fusion chat tests. The unaccepted current layer, when present, is stored only under training/.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for vtava/Qwen3.5-0.8B-MemoryFusion

Base model

Qwen/Qwen3.5-0.8B-Base

Finetuned

Qwen/Qwen3.5-0.8B

Finetuned

(383)

this model