Formalizes four concurrency anomalies in multi-agent LLM systems and mechanically verifies a hierarchy of sound detectors and preventions realized in Rust runtimes using TLA+ and Verus.
Elle: Inferring isolation anomalies from experimental observations.ArXiv abs/2003.10554(2020)
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 3roles
background 1polarities
background 1representative citing papers
Boomslang introduces a front-end/back-end pipeline with superpositions in its IR to enable general-purpose checking of arbitrary transaction isolation levels via SMT solving.
Lemonshark enables early finality for transactions in asynchronous DAG-BFT by identifying conditions where commitment is sufficient but not necessary for safe results, cutting latency by up to 65%.
citing papers explorer
-
Verified Detection and Prevention of Concurrency Anomalies in Multi-Agent Large Language Model Systems
Formalizes four concurrency anomalies in multi-agent LLM systems and mechanically verifies a hierarchy of sound detectors and preventions realized in Rust runtimes using TLA+ and Verus.
-
Making TransactionIsolation Checking Practical
Boomslang introduces a front-end/back-end pipeline with superpositions in its IR to enable general-purpose checking of arbitrary transaction isolation levels via SMT solving.
-
Lemonshark: Asynchronous DAG-BFT With Early Finality
Lemonshark enables early finality for transactions in asynchronous DAG-BFT by identifying conditions where commitment is sufficient but not necessary for safe results, cutting latency by up to 65%.