Organization card: benchmarks, models, leaderboards, naming and access notes
Browse files
README.md
CHANGED
|
@@ -6,6 +6,50 @@ colorTo: blue
|
|
| 6 |
sdk: static
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
-
The Fin AI is dedicated to advancing the open-source development of financial AI technologies. Our focus is on building financial large language models (LLMs) to tackle real-world challenges such as financial analysis, risk assessment, fraud detection, and regulatory compliance. We collaborate with academic and industry partners to curate datasets, benchmark models, and create secure AI applications.
|
| 10 |
|
| 11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
sdk: static
|
| 7 |
pinned: false
|
| 8 |
---
|
|
|
|
| 9 |
|
| 10 |
+
# The Fin AI π€
|
| 11 |
+
|
| 12 |
+
**The Fin AI** is an open research community building financial large language models, benchmarks and agents β for financial analysis, risk assessment, fraud detection, regulatory compliance and trading. We work with academic and industry partners to curate datasets, evaluate models and release open tools. π [thefin.ai](https://thefin.ai)
|
| 13 |
+
|
| 14 |
+
## π Leaderboards
|
| 15 |
+
|
| 16 |
+
[Open FinLLM Leaderboard](https://huggingface.co/spaces/TheFinAI/Open-FinLLM-Leaderboard) Β· [Reasoning Leaderboard](https://huggingface.co/spaces/TheFinAI/open-finllm-reasoning-leaderboard) Β· [Greek Leaderboard](https://huggingface.co/spaces/TheFinAI/Open-Greek-Financial-LLM-Leaderboard) Β· [Agent Market Arena](https://huggingface.co/spaces/TheFinAI/Agent-Market-Arena) Β· [all leaderboards](https://huggingface.co/collections/TheFinAI/the-fin-ai-leaderboards-6ac7154319e5aa2b52037515)
|
| 17 |
+
|
| 18 |
+
## π Benchmarks
|
| 19 |
+
|
| 20 |
+
| | Focus | Collection | Paper |
|
| 21 |
+
|---|---|---|---|
|
| 22 |
+
| **FinBen & PIXIU** | Holistic English financial NLP, reasoning, forecasting and trading | [β](https://huggingface.co/collections/TheFinAI/finben-and-pixiu-english-financial-evaluation-658f515911f68f12ea193194) | [arXiv:2402.12659](https://arxiv.org/abs/2402.12659), [arXiv:2306.05443](https://arxiv.org/abs/2306.05443) |
|
| 23 |
+
| **MultiFinBen** | Multilingual & multimodal (OCR, audio, 5 languages) | [β](https://huggingface.co/collections/TheFinAI/multifinben-multilingual-and-multimodal-financial-evaluation-6826f6fc4bc13d8af4fab223) | [arXiv:2506.14028](https://arxiv.org/abs/2506.14028) |
|
| 24 |
+
| **ICE-PIXIU** | Chinese financial tasks | [β](https://huggingface.co/collections/TheFinAI/ice-pixiu-chinese-financial-evaluation-65e5851af7daaa71c1c59902) | [arXiv:2403.06249](https://arxiv.org/abs/2403.06249) |
|
| 25 |
+
| **FLARE-ES** | Spanish financial tasks | [β](https://huggingface.co/collections/TheFinAI/flare-es-spanish-financial-evaluation-65e5855900680b19bc83e03d) | [arXiv:2402.07405](https://arxiv.org/abs/2402.07405) |
|
| 26 |
+
| **Plutus** | Greek financial tasks | [β](https://huggingface.co/collections/TheFinAI/plutus-greek-financial-evaluation-67bc718fb8d897c65f1e87db) | [arXiv:2502.18772](https://arxiv.org/abs/2502.18772) |
|
| 27 |
+
| **Ebisu** | Japanese financial language understanding | [β](https://huggingface.co/collections/TheFinAI/ebisu-japanese-financial-evaluation-6ac7153a5be985dedba08c78) | [arXiv:2602.01479](https://arxiv.org/abs/2602.01479) |
|
| 28 |
+
| **FinTagging** | XBRL tagging (FinNI, FinCL) | [β](https://huggingface.co/collections/TheFinAI/fintagging-xbrl-tagging-68270132372c6608ac069bef) | [arXiv:2505.20650](https://arxiv.org/abs/2505.20650) |
|
| 29 |
+
| **FinAuditing** | Taxonomy-structured multi-document auditing | [β](https://huggingface.co/collections/TheFinAI/finauditing-taxonomy-structured-auditing-68e5f80606e22454027075e7) | [arXiv:2510.08886](https://arxiv.org/abs/2510.08886) |
|
| 30 |
+
| **FinCriticalED** | Fact-level OCR of financial documents | [β](https://huggingface.co/collections/TheFinAI/fincriticaled-fact-level-financial-ocr-6ac7153c3d632745cc3f7f01) | [arXiv:2511.14998](https://arxiv.org/abs/2511.14998) |
|
| 31 |
+
| **CALM** | Credit scoring, fraud, distress and claims | [β](https://huggingface.co/collections/TheFinAI/calm-credit-scoring-and-risk-assessment-6ac7153781499dcd5001572e) | [arXiv:2310.00566](https://arxiv.org/abs/2310.00566) |
|
| 32 |
+
| **Conv-FinRe** | Conversational financial recommendation | [β](https://huggingface.co/collections/TheFinAI/conv-finre-conversational-financial-recommendation-69880c8328d2264727b74fb5) | [arXiv:2602.16990](https://arxiv.org/abs/2602.16990) |
|
| 33 |
+
| **FMD** | Financial misinformation detection | [β](https://huggingface.co/collections/TheFinAI/fmd-financial-misinformation-detection-6ac715408313fc0cb7529013) | [arXiv:2409.16452](https://arxiv.org/abs/2409.16452) |
|
| 34 |
+
| **EnterpriseArena** | Long-horizon CFO-style resource allocation (agents) | [β](https://huggingface.co/collections/TheFinAI/enterprisearena-can-llm-agents-be-cfos-6ac7153d9ffaefeeb3a950d4) | [arXiv:2603.23638](https://arxiv.org/abs/2603.23638) |
|
| 35 |
+
| **Herculean** | Agentic financial analysis (trading, hedging, reporting, auditing) | [β](https://huggingface.co/collections/TheFinAI/herculean-agentic-financial-benchmark-6ac71541a7f8dd780d7ab3dd) | [arXiv:2605.14355](https://arxiv.org/abs/2605.14355) |
|
| 36 |
+
|
| 37 |
+
## π§ Models
|
| 38 |
+
|
| 39 |
+
| | Focus | Collection | Paper |
|
| 40 |
+
|---|---|---|---|
|
| 41 |
+
| **Fin-o1 / Fino1** | Reasoning-enhanced financial LLMs (8B, 14B) | [β](https://huggingface.co/collections/TheFinAI/fino1-fin-o1-financial-reasoning-682702bbd25a7ef4512df24d) | [arXiv:2502.08127](https://arxiv.org/abs/2502.08127) |
|
| 42 |
+
| **Open-FinLLMs** | FinLLaMA, FinLLaMA-instruct, FinLLaVA (multimodal) | [β](https://huggingface.co/collections/TheFinAI/open-finllms-finllama-finllama-instruct-finllava-66b671f2b4958a65e20decbe) | [arXiv:2408.11878](https://arxiv.org/abs/2408.11878) |
|
| 43 |
+
| **Plutus-8B** | Greek financial LLM | [β](https://huggingface.co/collections/TheFinAI/plutus-greek-financial-evaluation-67bc718fb8d897c65f1e87db) | [arXiv:2502.18772](https://arxiv.org/abs/2502.18772) |
|
| 44 |
+
| **FinSeer / StockLLM** | Retrieval-augmented time-series forecasting | [β](https://huggingface.co/collections/TheFinAI/finseer-retrieval-augmented-time-series-forecasting-6ac7153f1e8bbfa51e7bd031) | [arXiv:2502.05878](https://arxiv.org/abs/2502.05878) |
|
| 45 |
+
| **FinMA** | PIXIU instruction-tuned models | [β](https://huggingface.co/collections/TheFinAI/finben-and-pixiu-english-financial-evaluation-658f515911f68f12ea193194) | [arXiv:2306.05443](https://arxiv.org/abs/2306.05443) |
|
| 46 |
+
|
| 47 |
+
## Using our datasets
|
| 48 |
+
|
| 49 |
+
- **Names** follow `{language}-{task}`, e.g. `en-fpb`, `zh-stocka`, `es-tsa`, `gr-qa`, `jp-icr` (`multi-` = multilingual). Paper-named benchmarks keep their names (e.g. `FinCriticalED`, `Herculean`). Old names such as `flare-fpb` still redirect.
|
| 50 |
+
- **Access:** most repositories ask for a short form (name, affiliation, intended use) and are approved **automatically**; third-party content is reviewed manually. Log in and accept once, then `load_dataset("TheFinAI/<name>")` works with your token.
|
| 51 |
+
- **Licenses** follow the original data sources; each dataset card states its source and license.
|
| 52 |
+
|
| 53 |
+
## Contact
|
| 54 |
+
|
| 55 |
+
Open a discussion on any repository, or reach us via [thefin.ai](https://thefin.ai).
|