Reproduce “Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Dohoon Kim, Junho Park, Hyunwoong Bae, Seyoon Kim, Taesup Moon

Axis-LoRA adapts frozen embedding models for personal-memory retrieval by adding four low-rank branches for temporal, spatial, entity, and event cues while retaining the base similarity score. It trains on counterfactual memories. Across four benchmark splits, Qwen3-Embedding-0.6B raises mean R@10 from 0.5892 to 0.6020, and all three encoders improve their means. This shows improved ranking, not query-specific routing or inference savings.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.