A hybrid interviewer pipeline with a Gemma 27B open-source model, a hand-built dialogue tree, reliability-weighted aggregation, and cluster imputation scored 3rd on eRisk 2026 depression screening, beating the team's paid GPT-5-nano baseline at 25% cost.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DS@GT ARC at eRisk 2026: Hybrid Multi-Agent LLM System with Structured Algorithmic Guidance for Conversational Depression Screening
A hybrid interviewer pipeline with a Gemma 27B open-source model, a hand-built dialogue tree, reliability-weighted aggregation, and cluster imputation scored 3rd on eRisk 2026 depression screening, beating the team's paid GPT-5-nano baseline at 25% cost.