SoftCoT
https://aclanthology.org/2025.acl-long.1137/
Soft chain-of-thought for efficient reasoning with LLMs. ACL 2025.
At a glance
- Type
- Soft chain of thought
- Tier
- T3
- Created
- 2025-02-17 (arxiv 2502.12134 submitted February 17 2025; ACL 2025)
- Latest release
- not applicable — not OSS
- License
- not applicable — not OSS
- GitHub
- not applicable — no GitHub repo
- Pricing
- not applicable — not commercial
- Funding
- not applicable — not commercial
Taxonomy
- storage
- kv-cache
- retrieval
- attention
- persistence
- session
- update
- read-only
- unit
- kv-token
- governance
- inspectable
- conflict
- n/a
When to use
Optimised for: not applicable - research paper
Anti-fit: not applicable - research paper
Pros & cons
Pros
Soft chain-of-thought as continuous memory rather than discrete tokens.
Cons
Research-only; integration into production inference stacks unclear.
Claims & capabilities
Continuous-space chain-of-thought reasoning using lightweight fixed assistant model to generate instance-specific soft thought tokens, mapped via trainable projection to LLM representation space; parameter-efficient fine-tuning without catastrophic forgetting; does not modify underlying LLM; ACL 2025; experimental results on five reasoning benchmarks demonstrate enhanced LLM reasoning
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.