A deployed customer-service pipeline that attributes failures to retrieval, ranking, or decision stages improved session accuracy from 86.5% to 92.5% by fine-tuning the reranker and final decision model.
Title resolution pending
1 Pith paper cite this work, alongside 2 external citations. Polarity classification is still indexing.
1
Pith paper citing it
2
external citations · OpenAlex
fields
cs.IR 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Evidence-in-the-Loop: Trace-Driven Optimization for Customer-Service LLM Agents
A deployed customer-service pipeline that attributes failures to retrieval, ranking, or decision stages improved session accuracy from 86.5% to 92.5% by fine-tuning the reranker and final decision model.