H2O
Heavy-hitter oracle for efficient generative inference. NeurIPS 2023.
At a glance
- Type
- Heavy-hitter oracle KV eviction
- Tier
- T3
- Created
- 2023-06-24 (arxiv 2306.14048 submitted June 24 2023; NeurIPS 2023)
- 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
- evict-oldest
- unit
- kv-token
- governance
- deterministic
- conflict
- n/a
When to use
Optimised for: not applicable - research paper
Anti-fit: not applicable - research paper
Pros & cons
Pros
Heavy-hitter oracle KV eviction — keeps tokens with high attention scores, evicts the rest; NeurIPS 2023.
Cons
Heavy-hitter heuristic can miss tokens that become important later; eviction is irreversible.
Claims & capabilities
Heavy-Hitter Oracle for KV-cache eviction balancing recent and high-impact tokens. Headline: up to 29x throughput on OPT-6.7B/OPT-30B with 20% heavy hitters retained, 1.9x latency reduction; baselines: DeepSpeed Zero-Inference, Hugging Face Accelerate, FlexGen; primary datasets: OPT/LLaMA/GPT-NeoX evaluated on HELM and lm-eval-harness (COPA, MathQA, OpenBookQA, PiQA, RTE, Winogrande, XSUM, CNN/Daily Mail), AlpacaEval, MT-Bench.
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.
Related systems
Referenced by (1)
- NAMMs cites — S2 isInfluential citation