AWhat changed
Co-designed a structured generation language and execution runtime with radix-tree LRU caching of attention KV states, enabling prefix reuse across dependent generation calls.
Scientific scope
Co-designed a structured generation language and execution runtime with radix-tree LRU caching of attention KV states, enabling prefix reuse across dependent generation calls. SGLang's original programming/runtime contribution is separate from later serving features and FlashInfer kernels. Reported speedups are benchmark-specific; no claim that shared prefixes permit reuse across incompatible model state or attention semantics.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. SGLang's original programming/runtime contribution is separate from later serving features and FlashInfer kernels. Reported speedups are benchmark-specific; no claim that shared prefixes permit reuse across incompatible model state or attention semantics.
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.
- Efficiently Programming Large Language Models using SGLang
- Zheng et al.: structured LM execution with radix-tree KV reuse — original project
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