M+ (MemoryLLM)
https://arxiv.org/abs/2502.00592
Extends MemoryLLM with scalable long-term memory.
Active · Verified 2026-10-28 · fresh as of 2026-10-28 · 20 cited facts, 0 with a third-party source · Suggest a correction
At a glance
- Type
- Scalable long-term latent memory vendor fix
- Tier
- T4
- Latest release
- not applicable — not OSS fix
- License
- not applicable — not OSS fix
- GitHub
- not applicable — no GitHub repo fix
- Pricing
- not applicable — not commercial fix
- Funding
- not applicable — not commercial fix
- Freshness
- 2026-10-28 fix
- Layer
- agent-infra repo fix
Memory taxonomy
- storage
- parametric
- retrieval
- parametric-recall
- persistence
- parametric-permanent
- update
- parametric-edit
- unit
- weight
- governance
- opaque
- conflict
- n/a
When to use
Optimised for: not applicable - research paper
Anti-fit: not applicable - research paper
Pros & cons
Pros
Extends MemoryLLM with scalable long-term storage; UCSD + Amazon.
Cons
Preprint; production deployment unclear; model is open weights but limited tooling.
Claims & capabilities
Long-term memory mechanism with co-trained retriever extending MemoryLLM via 1B-param latent-space memory pool compressing past info into hidden states. Headline: extends knowledge retention from <20K to >160K tokens at similar GPU memory; baselines: MemoryLLM-7B and Llama-3.1-8B with 16K/128K fixed contexts; primary datasets: LongBook-QA (351 book-question pairs averaging 192K tokens) and authors' LongBook Event QA.
Technical surface
- API surface
- not applicable — research paper
- Backend storage
- not applicable — research paper
- Deployment
- not applicable — not a deployable product
- Embedding model
- not applicable — research paper
- Multi-tenancy
- not applicable — research paper
- MCP
- not applicable — research paper, no deployed product
- A2A
- not applicable — research paper, no deployed product
- OpenTelemetry
- not applicable — research paper, no deployed product
Similar systems
Other recent method papers — theorized, no distinct product in the catalog, ranked by inbound references.
- Compressive Transformer T3
Maintains recent states in full resolution while compressing older memories with learned compression functions. DeepMind.
- MemGPT v2 / agent-tools T3
Already in catalog as the foundational MemGPT paper. Note: Letta is the productionised successor (cross-listed).
- Transformer-XL T3
Extends context through segment-level recurrence + caching of hidden states from prior segments. Foundational long-context architecture.
- Generative Agents T3
Park et al. — landmark agent-simulation paper. Reflection + memory stream + retrieval enable believable agent behavior.
- MemoryBank T3
Enhances LLMs with long-term memory. Early influential paper.
- Reflexion T3
Language agents with verbal reinforcement learning.
Sources (19 cited facts, 0 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Scalable long-term latent memory vendor
- What it does & what's distinct Extends MemoryLLM with scalable long-term memory. vendor
- Claims / benefits Long-term memory mechanism with co-trained retriever extending MemoryLLM via 1B-param la… vendor
- Modalities text vendor
- Created 2025-02 vendor
- Mindshare 26 cites · 1.7/mo · 313* / 32 forks paper
- Citations 26 cites 26/yr paper
- Founders / pedigree no data — searched arxiv.org/abs/2502.00592; author affiliations not found in preview vendor
- Memory volume / scale no scale claim — research vendor
- Performance >160k tokens retention vs <20k MemoryLLM extends MemoryLLM-7B retention 8x at similar GP… vendor
- Reproducibility Original team + downstream attention: 32 forks of https://github.com/wangyu-ustc/MemoryL… repo
- Code/weights release Code public + weights closed — https://github.com/wangyu-ustc/MemoryLLM · 313* / 32 fork… repo
- Versioning snapshots vendor
- Schema evolution schema locked vendor
- Pros Extends MemoryLLM with scalable long-term storage; UCSD + Amazon. vendor
- Cons Preprint; production deployment unclear; model is open weights but limited tooling. vendor
- Project & sources arxiv 2502.00592 vendor
- citation-trajectory [{"year":2025,"cum":9,"infl":1}] paper
- taxonomy-three-layer-rollup agent-infra repo
Related systems
References (2)
- MemGPT v2 / agent-tools cites — S2 isInfluential citation
- MemoryLLM (original) cites · extends — Extends MemoryLLM with scalable long-term memory.
Referenced by (1)
- MIRIX cites — S2 isInfluential citation