flamiinngo/market-analysis-mixtral-8x7b

Market-Analysis Mixtral-8x7B — Financial News Q&A

A LoRA fine-tune of Mixtral-8x7B-Instruct-v0.1 adapted for market analysis and financial news interpretation: classifying news by market area, reading sentiment, identifying who a story matters to, and answering forward-looking questions from earnings calls.

Built with Adaption Labs' AutoScientist for the AutoScientist Challenge (Market-Analysis & News category).

Results

Head-to-head win rate against the base model:

Evaluation	Base	Adapted
Training distribution	28	72
Market-Analysis category (all tasks)	40	60

Both numbers are wins in a paired comparison, not accuracy percentages. The two figures differ because the category evaluation spans a wider range of tasks than the training data covers — the model is strongest on the task types it was trained on and holds a smaller edge across the category as a whole.

Worth stating plainly: an earlier checkpoint trained on shorter, more templated answers scored 64–36 on the category evaluation while scoring only 45–55 on its own training distribution. Concise answers generalised better; richer answers won more decisively in-distribution. This release is the latter.

Usage

Extract the adapter — the archive unpacks flat, so give it its own directory:

mkdir -p market-adapter
tar --zstd -xf adaption_mixtral_8x7b_instruc_financial_news_classifie_45d894da.tgz -C market-adapter

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

BASE = "mistralai/Mixtral-8x7B-Instruct-v0.1"
ADAPTER = "./market-adapter"

tokenizer = AutoTokenizer.from_pretrained(ADAPTER)
model = AutoModelForCausalLM.from_pretrained(
    BASE, torch_dtype=torch.bfloat16, device_map="auto"
)
model = PeftModel.from_pretrained(model, ADAPTER)
model.eval()

messages = [{
    "role": "user",
    "content": "What area of the market does this headline concern, and why does "
               "it matter? Citi cuts Microsoft price target, citing foreign "
               "exchange headwinds",
}]
inputs = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True, return_tensors="pt"
).to(model.device)

out = model.generate(inputs, max_new_tokens=200, do_sample=False)
print(tokenizer.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))

Hardware: Mixtral-8x7B is a 46.7B-parameter mixture-of-experts model — roughly 94 GB in bf16, or about 25 GB with 4-bit quantisation. The adapter itself is 54 MB.

Task types it was trained on: topic classification, sentiment assessment, market implication, subject identification, who-it-matters-to, and earnings-call analysis.

Training

• Base: mistralai/Mixtral-8x7B-Instruct-v0.1
• Method: LoRA supervised fine-tuning via AutoScientist (recipe auto-selected)
• Data: 5,929 adapted rows from flamiinngo/market-news-qa

Parameter	Value
Rank (r)	16
lora_alpha	32
lora_dropout	0.0
Target modules	q_proj, k_proj, v_proj, o_proj
Peak learning rate	1e-4
Schedule	cosine, warmup 0.1
Epochs	3 (249 steps)

Training curve

Epoch	Step	Eval loss
0.64	53	1.0593
1.23	102	1.0188
1.82	151	1.0000
2.41	200	0.9922
3.00	249	0.9912

Training loss fell from 1.414 to 0.876. Evaluation loss was still declining at the end of epoch 3, though the last epoch gained only 0.001 — close to converged, with no sign of overfitting.

Dataset

flamiinngo/market-news-qa — 10,011 market-analysis Q&A pairs across five task types, median answer 16 words, 85% of answers distinct. Derived from zeroshot/twitter-financial-news-topic (MIT), zeroshot/twitter-financial-news-sentiment (MIT), and lamini/earnings-calls-qa (CC-BY-4.0).

Also on Kaggle: dataset · adapted training data

Limitations

• Not investment advice. This model classifies and comments on financial news. It does not predict prices and must not be used to make investment decisions.
• It can state figures confidently and be wrong. The training data includes specific financial numbers. For any company or period outside that data, treat numeric claims as unverified.
• Training headlines skew US markets, 2020–2022. Coverage of other regions and later periods is thinner.
• Category labels carry upstream annotation noise, so some classifications reflect debatable source labels.
• Reasoning clauses in the training data are category-level, not per-headline human judgment. The model learned correct categories with generally apt explanations, not independent analysis of each story.
• Evaluated by automated head-to-head comparison against one base model. Win rate is not a measure of factual accuracy in absolute terms.
• English only.

License

The adapter is released under Apache 2.0, matching the base model's licence. Note that the training data is CC-BY-4.0 and requires attribution — see the dataset card.

Acknowledgements

• Adaption Labs — AutoScientist platform and the AutoScientist Challenge
• zeroshot and Lamini — upstream open datasets
• Mistral AI — Mixtral-8x7B base model