YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
🧠 Hybrid Episodic Memory & Parametric Routing with LFM-2.5-350M
Official implementation of a resource-constrained, sample-efficient continual learning framework utilizing hybrid memory routing and LFM-2.5-350M embeddings.
🚀 Key Features
- O(1) Streaming Adaptability: Dynamic routing between the parametric prediction head and episodic memory lookup: $$p_t(y) = (1-\alpha) p_{\text{head}}(y) + \alpha p_{\text{mem}}(y)$$
- LFM-2.5-350M Embedded Features: Ultra-efficient representation encoding for downstream low-latency streaming routing.
- Zero-Shot Fictional Knowledge Ingestion: Proven 100.0% immediate recall on completely unseen, dynamically streamed factual injections.
- Optimized Storage Policies: High performance stability down to a capacity of $M=10$ with Balanced Eviction memory schemas.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support