2026-10-13 04:35 UTC
A zero-training memory harness for open chat models (Qwen and friends). The same model that reads and writes text also classifies its own context as long-term / short-term memory while decoding; a small deterministic eviction engine then enforces the token budget.
Design inspired by Context Language Models (Shao et al., arXiv:2609.37725): context is treated as an editable artifact mirrored to disk, and context management is delegated to the model via prompting with no fine-tuning, works off the shelf with any OpenAI compatible server (vLLM, Ollama, llama.cpp, LM Studio).
https://github.com/glassresistor/clm-mem
#llm #ai #jev
Replies (0)
No replies.