LLM-based Medical Assistant Personalization with Short- and Long-Term Memory Coordination (NAACL 2024)
Overview
This entry indexes the NAACL 2024 paper "LLM-based Medical Assistant Personalization with Short- and Long-Term Memory Coordination."
| Field | Content | |-------|---------| | Title | LLM-based Medical Assistant Personalization with Short- and Long-Term Memory Coordination | | Venue | NAACL 2024 (long paper) | | Link | https://aclanthology.org/2024.naacl-long.132/ | | Category | Personalization |
The paper studies how to personalize LLM-based medical assistants by coordinating short-term memory (within a conversation/session) and long-term memory (persistent user context across sessions), so that responses can adapt to each user's medical history and interaction patterns.
Context in the Field
In large-scale search, recommendation, and personalization systems, information retrieval has long faced challenges around efficiency, scalability, and understanding user intent. Traditional pipelines separate retrieval, ranking, and generation, which struggles to meet LLM-era demands for natural language interaction, multi-hop reasoning, and up-to-date knowledge. This work sits at the intersection of personalization and LLM-based assistants.
Typical components in such systems include:
1. Input and representation — encoding queries, documents, and user context as dense/sparse representations or structured prompts; 2. Core modules — retrievers, rerankers, planners, memory modules, and tool interfaces; 3. Learning strategies — supervised fine-tuning, contrastive learning, distillation, reinforcement learning, and synthetic data; 4. Inference strategies — single-pass retrieval, iterative retrieval, parallel sub-queries, early stopping, and budget control.
Key Takeaways for Search / Rec / Personalization
1. Architecture: cascaded retrieval + rerank + generation remains mainstream, but agentic paradigms treat "whether and how often to retrieve" as a learnable decision; 2. Data: high-quality instruction data and click/session logs are critical; synthetic data must guard against knowledge leakage and distribution shift; 3. Evaluation: the gap between offline metrics and online satisfaction is widening; LLM-as-judge should be cross-validated with human evaluation; 4. Product: latency, cost, explainability, and safety are hard constraints for real deployment — do not optimize only for academic benchmarks.
Engineering Checklist
| Item | Question | Recommendation | |------|----------|----------------| | Data | Does training/indexing contain PII? How are versions managed? | Partitioned indexes, anonymization, rollback-safe embedding versions | | Latency | What is the p99 budget? How many retrieval steps? | Cascades + early stopping, hot-query caching, async reranking | | Quality | Do offline gains translate to online CTR/satisfaction? | Interleaving experiments, human audits, citation verification | | Safety | Does open retrieval introduce poisoning/bias? | Source whitelists, adversarial detection, output filtering | | Cost | Token and GPU usage per query? | Small-model routing, distillation, hybrid sparse+dense retrieval |
Glossary
| Term | Meaning | |------|---------| | IR | Information Retrieval | | RAG | Retrieval-Augmented Generation | | LTR | Learning to Rank | | nDCG | Normalized Discounted Cumulative Gain, a ranking quality metric | | Agentic Search | Search modeled as sequential decision-making with tool use | | Gen-IR | Generative Information Retrieval |
Related Entries
- A Survey of Personalization: From RAG to Agent (arXiv 2504.10147)
- Bridging Personalization and Control in Scientific Personalized Search (arXiv 2411.02790)
- Can Large Language Models Understand Preferences in Personalized Recommendation? (arXiv 2501.13391)
- IntentRec: Predicting User Session Intent with Hierarchical Multi-Task Learning (arXiv 2408.05353)
- Unified Embedding-Based Personalized Retrieval in Etsy Search (arXiv 2306.04833)
- User-LLM: Efficient LLM Contextualization with User Embeddings (WWW, DOI 10.1145/3701716.3715463)
Notes
This post is based on the paper's abstract and public metadata. Specific numerical results should be verified against the original PDF at the ACL Anthology link above. Readers are encouraged to combine the paper with related surveys, open-source implementations, and industry write-ups to form a triangle of "paper contributions — open-source reproduction — production constraints."