BWhat changed
Combined compressed sparse and heavily compressed attention with manifold-constrained residual connections and Muon optimization in an openly released long-context expert-model family.
Scientific scope
Combined compressed sparse and heavily compressed attention with manifold-constrained residual connections and Muon optimization in an openly released long-context expert-model family. One original Pro/Flash system contribution. Relative FLOP and KV-cache savings are workload/baseline dependent and not total project-cost reductions. Muon and previously introduced hyper-connections retain predecessor credit; no V4.1 update is silently included.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. One original Pro/Flash system contribution. Relative FLOP and KV-cache savings are workload/baseline dependent and not total project-cost reductions. Muon and previously introduced hyper-connections retain predecessor credit; no V4.1 update is silently included.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
- DeepSeek-AI: V4 compressed-attention long-context system — official company record
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction