TriAgent: Divergence-Aware Multi-Agent Committees for Cost-Efficient Financial Sentiment Analysis

A new multi-agent committee called TriAgent, described in arXiv:2607.19794v1, achieves cost-efficient financial sentiment analysis by routing queries through a word-level lexicon (VADER), a sentence-level domain transformer (FinBERT), and a cross-sentence reasoner (Qwen2.5, 0.5B-14B-4bit, with Mistral-7B and Phi-3.5-mini cross-family checks) using a three-way Semantic Divergence Index (SDI). The system saves $9.3M/year at 10M-user scale versus a GPT-4o-mini baseline, achieves F1=0.87 with the LLM as a critic, and SDI doubles as a hallucination detector (AUC=0.90) and achieves a Sharpe ratio of 3.50 in back-tests.

arXiv:2607.19794v1 Announce Type: new Abstract: Production LLM-based financial sentiment analysis faces a structural cost trap: most queries are trivially classifiable, yet expensive cloud reasoners process them all, and the bill scales linearly with user count. We present TriAgent, a multi-agent committee stratified by contextual granularity -- a word-level lexicon VADER , a sentence-level domain transformer FinBERT , and a cross-sentence reasoner Qwen2.5, 0.5B-14B-4bit, with Mistral-7B and Phi-3.5-mini cross-family checks . A three-way Semantic Divergence Index SDI measures pairwise disagreement across granularities and routes each query accordingly. Our central finding is the critic plateau: when the LLM is re-tasked as a critic over the smaller agents' outputs, F1 plateaus at ~0.87 across 1.5B-7B Qwen bootstrap 95% CIs overlap , while a same-size 3-persona vote drops to F1=0.66, which is driven by granularity-stratified diversity. Three corollaries follow from the same SDI signal: i a Shared Consensus Dictionary on multilingual sentence-BERT answers 95% of Chinese queries from an English cache at F1=0.99 -- cross-border canonicalization at zero marginal cost; ii SDI doubles as a post-hoc LLM-hallucination detector at AUC=0.90; iii the SDI single-stage strategy attains the best risk-adjusted return Sharpe=3.50 on a 20-ticker back-test, dominating both always-FinBERT 1.36 and always-LLM 0.11 . At 10M-user scale, TriAgent saves $9.3M/year vs. a GPT-4o-mini baseline. Code, lexicons, and the SCD are released.