MoT (Memory-of-Thought)
https://arxiv.org/abs/2305.05181
Before test time, LLM reasons over unlabeled data and stores high-confidence intermediate thoughts as external memory. At test time, retrieves relevant stored thoughts to augment reasoning. No annotated data, no parameter updates.
Research code · Verified 2026-10-28 · fresh as of 2026-10-28 · 26 cited facts, 0 with a third-party source · 1 recorded change · Suggest a correction
At a glance
- Type
- Pre-test self-reflection cache vendor fix
- Tier
- T3
- Latest release
- no-release no-release vendor fix
- License
- no-license repo fix
- GitHub
- 24★ Python repo fix
- Pricing
- not applicable — research paper fix
- Funding
- not applicable — not commercial fix
- Freshness
- 2026-10-28 fix
- Layer
- agent-infra repo fix
Memory taxonomy
- storage
- vector
- retrieval
- similarity
- persistence
- session
- update
- append-only
- unit
- chunk
- governance
- n/a
- conflict
- llm-arbitrate
When to use
Optimised for: not applicable - research paper
Anti-fit: not applicable - research paper
Pros & cons
Pros
Chain-of-thought combined with memory recall for deliberative reasoning.
Cons
Research-only; recall+reasoning loop adds inference cost.
Claims & capabilities
Significant gains on arithmetic, commonsense, factual, NLI reasoning with ChatGPT. EMNLP 2023.
Technical surface
- API surface
- not applicable — research paper
- Backend storage
- not applicable — research paper
- Deployment
- not applicable — research paper
- 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 (25 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 Pre-test self-reflection cache vendor
- What it does & what's distinct Before test time, LLM reasons over unlabeled data and stores high-confidence intermediat… vendor
- Claims / benefits Significant gains on arithmetic, commonsense, factual, NLI reasoning with ChatGPT. EMNLP… vendor
- Modalities text (reasoning tasks — arithmetic commonsense factual NLI) vendor
- Created 2023-06 vendor
- Latest release no-release no-release vendor
- License no-license repo
- GitHub 24★ Python repo
- Citations 55 cites 18/yr paper
- Founders / pedigree Peking University; Xiaonan Li and Xipeng Qiu; EMNLP 2023 vendor
- Memory volume / scale no scale claim — research vendor
- Performance Significant gains on arithmetic, commonsense, factual, and NLI reasoning with ChatGPT. E… vendor
- Reproducibility No reproductions found paper
- Code/weights release Code public + weights closed repo
- Versioning immutable append-log vendor
- Tombstoning append-only no delete vendor
- Schema evolution schema with migration vendor
- Contradiction handling always append (no resolution) vendor
- Forgetting policy none (permanent) vendor
- Pros Chain-of-thought combined with memory recall for deliberative reasoning. vendor
- Cons Research-only; recall+reasoning loop adds inference cost. vendor
- Project & sources arxiv 2305.05181 vendor
- commit-trajectory [{"ym":"2023-06","cum":3},{"ym":"2023-11","cum":12}] repo
- citation-trajectory [] paper
- taxonomy-three-layer-rollup agent-infra repo
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)
- 2026-10-22
- Lifecycle first recorded as “research-complete”