LLMQuant

The AI community building future for financial trading

[LLMQuant @LLMQuant]

🚀 Welcome to join LLMQuant — the global AI community shaping the future of investment research through open collaboration and cutting-edge innovation.

🔗 Join us across:
GitHub: https://t.co/zFHOYtr13O
Website: https://t.co/hsjbLGaFXD

📬 Reach out at [email protected]

[LLMQuant @LLMQuant]

Skills - LLMQuant Macro Radar Skill turns a morning of cross-asset research into a single auditable brief, and what to do when the alignment breaks. https://t.co/3wiBUfuwXf https://t.co/A9UgXI6UbF

[LLMQuant @LLMQuant]

We built Awesome Trading Agents, the open source map of every serious project where LLMs are running real investment workflows. 40+ repos, 3 layers (Agents, MCPs, Skills), 1 clean view of the stack that standardised in 2026. https://t.co/uzw6ACaAq2 https://t.co/osk6XVTU0B

[LLMQuant @LLMQuant]

We just tested a new Smart Money workflow on Bridgewater’s latest 13F using Claude Code + GPT Codex + LLMQuant Data MCP. https://t.co/01u2ebJYb6

[LLMQuant @LLMQuant]

Reading FOMC minutes is not a standard sentiment task. The hardest part is economic interpretation. This article explains why FinBERT beats VADER, why GPT often defaults to neutral, and what macro quants should build instead.
https://t.co/th027NA264
https://t.co/1yIiHqCV7k

[LLMQuant @LLMQuant]

Clean data is not enough. Real edge comes from knowing how to collect, validate, transform, and engineer data into something markets have not fully priced... https://t.co/MiQ2mTCCkk

[LLMQuant @LLMQuant]

LLMs read 10-Ks like analysts and convert narrative into signals, the result can be a low turnover strategy with real outperformance.
https://t.co/d4gqQ5kVvs
https://t.co/jTEYrsvvFB

[LLMQuant @LLMQuant]

A simple formula shows the exact correlation threshold required for a signal to be tradable. Before running a backtest, check whether your signal even has a chance to survive the market.
https://t.co/4VroVv8cKI

[LLMQuant @LLMQuant]

Can ChatGPT predict stocks from Twitter alone? One study says it got MSFT direction right on 26 of 37 days and Google right on 23 of 36.
Check it here: https://t.co/EbpyG6QDDe
https://t.co/E8EUFVxsf5

[LLMQuant @LLMQuant]

A ChatGPT informed GNN boosts F1 scores, doubles downside detection, and reduces portfolio volatility from 23.6 percent to 14.0 percent. Language models are now topology engines for finance. Check it here: https://t.co/C3N7z9QhkT
https://t.co/HUBlMRDXFL

[LLMQuant @LLMQuant]

Inflation changes how people think, and that rewires how they trade. A field experiment with real brokerage data shows that nudging inflation expectations barely moves return forecasts, but showing what stocks actually did in past cuts expected returns.
https://t.co/TdeKNU3ClX

[LLMQuant @LLMQuant]

FX risk premia are insurance pricing. If your currency depreciates when GDP falls, investors demand compensation. Dollar invoicing plus dollar debt creates that recession depreciation pattern, and it boosts exposure to global risk off carry unwind shocks.
https://t.co/UFbW5mN1oB

[LLMQuant @LLMQuant]

Seven month momentum. Thirty day reversion. Neural classification at 64 percent accuracy. Sector rotation alpha is measurable, not mythical. Check the study here https://t.co/sMeTrhKk4F
https://t.co/dFNLf8Ya2s

[LLMQuant @LLMQuant]

What if the bottleneck in quant research is not compute, but iteration speed? Alpha-GPT 2.0 proposes a full pipeline Human-in-the-Loop architecture that turns large language models into collaborative research agents. Check it here https://t.co/C5c5oE1gtb
https://t.co/uVqLq0AXbe

[LLMQuant @LLMQuant]

QuantAgent treats alpha mining like a compounding business: every backtest becomes reusable knowledge.
Check the study here: https://t.co/7PJyRk1EbA
https://t.co/EqFZL1kKLU

[LLMQuant @LLMQuant]

FinBen just put financial LLMs on a real exam.
If your “AI alpha” pitch relies on price prediction, this benchmark is the reality check.
Check it here: https://t.co/BOyH596pxp
https://t.co/TAEZuW426H

[LLMQuant @LLMQuant]

Bigger models are not the only path forward. Low-rank + quantization cuts memory by up to 97% and boosts financial task accuracy. Efficient AI is now a competitive edge. Check the study here:
https://t.co/aecnkuUU6E
https://t.co/EeEfS8al6H

[LLMQuant @LLMQuant]

Profit aligned LLM vs baseline.
Moral accuracy drops from 86.9% to 54.4% in clear cases. Alignment reshapes the decision space.
https://t.co/YV7V6JsSc9
https://t.co/oAw8XjIauG

[LLMQuant @LLMQuant]

LLMs are terrible risk prophets, but elite risk analysts. RiskLabs: multimodal earnings calls plus VIX plus news, with the LLM used for extraction not prediction... Check the study here:
https://t.co/XqaHb3PalV
https://t.co/Y092UTxlF5

[LLMQuant @LLMQuant]

41 ML models traded BTC in one study. Bagging: huge backtest PnL, then negative forward, while Random Forest and SGD look far more resilient on Sharpe and PnL. Accuracy is cheap. Robustness is alpha. Check the study here: https://t.co/c4ImlhXCbq
https://t.co/mejZDfFAOl

[LLMQuant @LLMQuant]

If your trading simulator ignores clustering, tails, or market impact, it is lying to you. Here is a breakdown of modern limit order book simulation models and where they fail. https://t.co/BQwSxI6hzn
https://t.co/N2fyjFAlRK