POLAR:A Per-User Association Test in Embedding Space
2026-03-18T08:52:19Z•022949b92fb95a6242eda8a24a2e76a56661f475bc44bd5c7a00c67eebdf2fe2
AI reproducibilityBenjamini-HochbergPOLARSUnSETagent variabilityautomated analysis agentsbot detectioncorrelated graphs hypothesis testingembeddingsextremism detectiongraph algorithmslexical associationmarket data (NYSE TAQ)masked language modelsmotif countingnatural language processingnonstandard errorsper-user analysispermutation testprivacy/deanonymization risksocial media (Reddit/Twitter)spiral of silencestatistical testingsurvey methodstimeline summarization
What happened
Collection of recent CS/SI arXiv papers spanning methods for per-user lexical association in embedding space (POLAR) to detect bot/ideological alignment; empirical study of agent-to-agent variability ("nonstandard errors") among autonomous coding agents on market-data analyses; survey-based measurement of the "spiral of silence" on Reddit; stakeholder-aware timeline summarization (SUnSET); polynomial-time motif-counting tests for correlated Erdős–Rényi graphs; bounds on degree vs. leading-eigenvector misalignment in assortative/structured networks; FS_GPlib, a unified, highly accelerated graph
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_si
- Record identifier
- 022949b92fb95a6242eda8a24a2e76a56661f475bc44bd5c7a00c67eebdf2fe2
- Enrichment time
- 2026-03-18T08:52:19Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.