TriAgent: Divergence-Aware Multi-Agent Committees for Cost-Efficient Financial Sentiment Analysis
Quick Answer
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5.
Quick Take
It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.
Key Points
- TriAgent uses a three-tiered approach: word-level, sentence-level, and cross-sentence reasoning.
- Achieves F1 score of ~0.87, outperforming a 3-persona voting system at F1=0.66.
- Shared Consensus Dictionary answers 95% of Chinese queries with F1=0.99 from English cache.
- SDI serves as a hallucination detector with AUC=0.90.
- Saves $9.3M/year at a 10M-user scale compared to GPT-4o-mini.
DeepSignal Analysis
What happened
TriAgent is a multi-agent system designed for financial sentiment analysis, integrating VADER, FinBERT, and Qwen2.5. It reportedly achieves an F1 score of approximately 0.87 while saving $9.3 million annually at a scale of 10 million users compared to GPT-4o-mini. Additionally, it detects LLM hallucinations with an AUC of 0.90.
Key evidence
- TriAgent combines three models: VADER for word-level analysis, FinBERT for sentence-level, and Qwen2.5 for cross-sentence reasoning.
- The system achieves an F1 score of ~0.87 and saves $9.3 million per year at a 10 million user scale compared to GPT-4o-mini.
- TriAgent's Semantic Divergence Index (SDI) serves as a post-hoc LLM hallucination detector, achieving an AUC of 0.90.
Why it matters
The introduction of TriAgent addresses the cost inefficiencies in processing financial sentiment analysis queries, where many are easily classifiable. By utilizing a multi-agent approach, it optimizes resource allocation and reduces operational costs significantly. The ability to detect hallucinations in LLM outputs is also crucial for maintaining the reliability of financial analyses, which can have substantial implications for decision-making in finance.
Paper Resources
📖 Reader Mode
~2 min readAbstract: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.
| Subjects: | Computation and Language (cs.CL); Computational Engineering, Finance, and Science (cs.CE); Databases (cs.DB); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.19794 [cs.CL] |
| (or arXiv:2607.19794v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.19794 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jiacheng Ding [view email]
[v1]
Wed, 22 Jul 2026 06:20:32 UTC (343 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →RF-Agent: A Practical Framework for Building Language Agents for RFIC Design
RF-Agent introduces a novel framework for RF circuit design using , creating a unique RF-domain reasoning dataset with over 11,000 samples. The study reveals that domain-specific supervised fine-tuning and semantic retrieval strategies significantly enhance RF reasoning performance, particularly for smaller models.