Paper: arXiv 2512.00630
Authors: Zhiming Lian
Abstract
Financial text classification has increasingly become an important aspect in quantitative trading systems and related tasks, such as financial sentiment analysis and the classification of financial news. In this paper, we assess the performance of the large language model Qwen3-8B on both tasks. Qwen3-8B is a state-of-the-art model that exhibits strong instruction-following and multilingual capabilities, and is distinct from standard models, primarily because it is specifically optimized for efficient fine tuning and high performance on reasoning-based benchmarks, making it suitable for financial applications. To adapt this model, we apply Noisy Embedding Instruction Finetuning and based on our previous work, this method increases robustness by injecting controlled noise into the embedding layers during supervised adaptation. We improve efficiency further with Rank-stabilized Low-Rank Adaptation low-rank optimization approach, and FlashAttention, which allow for faster training with lower GPU memory. For both tasks, we benchmark Qwen3-8B against standard classical transformer models, such as T5, BERT, and RoBERTa, and large models at scale, such as LLaMA1-7B, LLaMA2-7B, and Baichuan2-7B. The findings reveal that Qwen3-8B consistently surpasses these baselines by obtaining better classification accuracy and needing fewer training epochs. The synergy of instruction-based fine-tuning and memory-efficient optimization methods suggests Qwen3-8B can potentially serve as a scalable, economical option for real-time financial NLP applications. Qwen3-8B provides a very promising base for advancing dynamic quantitative trading systems in the future.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 7.5/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper applies standard deep learning architectures (LoRA, Qwen3) with minimal novel mathematical derivations, but is heavily data-driven with explicit benchmarking on financial datasets, hyperparameter tuning, and empirical comparisons against multiple baselines.
Research Flowchart
flowchart TD Goal["Research Goal: Assess Qwen3-8B for Financial Text Classification"] SubMethod["Methodology<br>Noisy Embedding Instruction Finetuning<br>Rank-stabilized Low-Rank Adaptation<br>FlashAttention"] Inputs["Data/Inputs<br>Financial Sentiment Analysis &<br>Financial News Classification Tasks"] Processes["Computational Processes<br>Qwen3-8B Fine-tuning<br>vs. BERT, RoBERTa, T5, LLaMA, Baichuan2"] Outcomes["Key Findings<br>Qwen3-8B surpasses baselines<br>Fewer epochs needed<br>Scalable & economical solution"] Goal --> SubMethod SubMethod --> Inputs Inputs --> Processes Processes --> Outcomes
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a training approach that uses Noisy Embedding Instruction Finetuning together with rLoRA and FlashAttention for efficient, robust finetuning of Qwen3-8B on financial text classification.
Source sentence
we proposed a training approach that utilizes Noisy Embedding Instruction Finetuning in conjunction with both rLoRA and FlashAttention, which produces an effecient and robust finetuning strategy.
Assumes
No explicit assumptions found in the text.
Data
- financial sentiment classification dataset text-news unknown
Source sentence
One is a financial sentiment classification dataset with 2,879 neutral, 1,362 positive, and 604 negative samples, resulting in a reasonably balanced distribution of sentiment categories.
- Twitter-based financial news dataset text-news public
Source sentence
The second dataset is a Twitter-based financial news dataset publicly available on Kaggle that we use for model fine-tuning to classify financial news articles into 20 categories.
Twitter / X Kaggle
Costs
Not applicable
Finds 2
- Qwen3-8B achieves the highest accuracy among all baselines, reaching 0.8415 on sentiment classification and 0.9315 on topic classification. performance
Source sentence
Qwen3-8B achieves better performance than all baselines, reaching 0.8415 on sentiment classification and 0.9315 on topic classification.
- Qwen3-8B converges rapidly, attaining stable performance within only three epochs, whereas traditional non-LLM methods typically need more than ten epochs. performance
Source sentence
Qwen3-8B quickly converges with stable performance after three epochs of fine-tuning whilst requiring relatively fewer iterations, while prior non-LLM approaches usually take over ten epochs of tuning to achieve a similar accuracy for financial classification.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 40 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.