Sailor2: Sailing in South-East Asia with Inclusive Multilingual LLMs
Paper
•
2502.12982
•
Published
•
9
ONNXConfig for all
non-profit
AI & ML interests
Make all hub models available for conversion to ONNX format.
Recent Activity
View all activity
OWG's activity
wannaphong
authored
a
paper
3 days ago
louisbrulenaudet
posted
an
update
6 days ago
Post
2969
I am pleased to introduce my first project built upon Hugging Face’s smolagents framework, integrated with Alpaca for financial market analysis automation 🦙🤗
The project implements technical indicators such as the Relative Strength Index (RSI) and Bollinger Bands to provide momentum and volatility analysis. Market data is retrieved through the Alpaca API, enabling access to historical price information across various timeframes.
AI-powered insights are generated using Hugging Face’s inference API, facilitating the analysis of market trends through natural language processing with DuckDuckGo search integration for real-time sentiment analysis based on financial news 🦆
Link to the GitHub project: https://github.com/louisbrulenaudet/agentic-market-tool
The project implements technical indicators such as the Relative Strength Index (RSI) and Bollinger Bands to provide momentum and volatility analysis. Market data is retrieved through the Alpaca API, enabling access to historical price information across various timeframes.
AI-powered insights are generated using Hugging Face’s inference API, facilitating the analysis of market trends through natural language processing with DuckDuckGo search integration for real-time sentiment analysis based on financial news 🦆
Link to the GitHub project: https://github.com/louisbrulenaudet/agentic-market-tool
vumichien
authored
a
paper
10 days ago
Post
4451
Introducing OpenR1-Math-220k!
open-r1/OpenR1-Math-220k
The community has been busy distilling DeepSeek-R1 from inference providers, but we decided to have a go at doing it ourselves from scratch 💪
What’s new compared to existing reasoning datasets?
♾ Based on AI-MO/NuminaMath-1.5: we focus on math reasoning traces and generate answers for problems in NuminaMath 1.5, an improved version of the popular NuminaMath-CoT dataset.
🐳 800k R1 reasoning traces: We generate two answers for 400k problems using DeepSeek R1. The filtered dataset contains 220k problems with correct reasoning traces.
📀 512 H100s running locally: Instead of relying on an API, we leverage vLLM and SGLang to run generations locally on our science cluster, generating 180k reasoning traces per day.
⏳ Automated filtering: We apply Math Verify to only retain problems with at least one correct answer. We also leverage Llama3.3-70B-Instruct as a judge to retrieve more correct examples (e.g for cases with malformed answers that can’t be verified with a rules-based parser)
📊 We match the performance of DeepSeek-Distill-Qwen-7B by finetuning Qwen-7B-Math-Instruct on our dataset.
🔎 Read our blog post for all the nitty gritty details: https://huggingface.co/blog/open-r1/update-2
open-r1/OpenR1-Math-220k
The community has been busy distilling DeepSeek-R1 from inference providers, but we decided to have a go at doing it ourselves from scratch 💪
What’s new compared to existing reasoning datasets?
♾ Based on AI-MO/NuminaMath-1.5: we focus on math reasoning traces and generate answers for problems in NuminaMath 1.5, an improved version of the popular NuminaMath-CoT dataset.
🐳 800k R1 reasoning traces: We generate two answers for 400k problems using DeepSeek R1. The filtered dataset contains 220k problems with correct reasoning traces.
📀 512 H100s running locally: Instead of relying on an API, we leverage vLLM and SGLang to run generations locally on our science cluster, generating 180k reasoning traces per day.
⏳ Automated filtering: We apply Math Verify to only retain problems with at least one correct answer. We also leverage Llama3.3-70B-Instruct as a judge to retrieve more correct examples (e.g for cases with malformed answers that can’t be verified with a rules-based parser)
📊 We match the performance of DeepSeek-Distill-Qwen-7B by finetuning Qwen-7B-Math-Instruct on our dataset.
🔎 Read our blog post for all the nitty gritty details: https://huggingface.co/blog/open-r1/update-2
lewtun
authored
a
paper
16 days ago
Post
10142
We are reproducing the full DeepSeek R1 data and training pipeline so everybody can use their recipe. Instead of doing it in secret we can do it together in the open!
🧪 Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
🧠 Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
🔥 Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
🧪 Step 1: replicate the R1-Distill models by distilling a high-quality reasoning corpus from DeepSeek-R1.
🧠 Step 2: replicate the pure RL pipeline that DeepSeek used to create R1-Zero. This will involve curating new, large-scale datasets for math, reasoning, and code.
🔥 Step 3: show we can go from base model -> SFT -> RL via multi-stage training.
Follow along: https://github.com/huggingface/open-r1
- 5 replies
Post
1608
R1 is out! And with a lot of other R1 releated models...
Post
3859
I was initially pretty sceptical about Meta's Coconut paper [1] because the largest perf gains were reported on toy linguistic problems. However, these results on machine translation are pretty impressive!
https://x.com/casper_hansen_/status/1875872309996855343
Together with the recent PRIME method [2] for scaling RL, reasoning for open models is looking pretty exciting for 2025!
[1] Training Large Language Models to Reason in a Continuous Latent Space (2412.06769)
[2] https://huggingface.co/blog/ganqu/prime
https://x.com/casper_hansen_/status/1875872309996855343
Together with the recent PRIME method [2] for scaling RL, reasoning for open models is looking pretty exciting for 2025!
[1] Training Large Language Models to Reason in a Continuous Latent Space (2412.06769)
[2] https://huggingface.co/blog/ganqu/prime
Post
2293
This paper (
HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs (2412.18925)) has a really interesting recipe for inducing o1-like behaviour in Llama models:
* Iteratively sample CoTs from the model, using a mix of different search strategies. This gives you something like Stream of Search via prompting.
* Verify correctness of each CoT using GPT-4o (needed because exact match doesn't work well in medicine where there are lots of aliases)
* Use GPT-4o to reformat the concatenated CoTs into a single stream that includes smooth transitions like "hmm, wait" etc that one sees in o1
* Use the resulting data for SFT & RL
* Use sparse rewards from GPT-4o to guide RL training. They find RL gives an average ~3 point boost across medical benchmarks and SFT on this data already gives a strong improvement.
Applying this strategy to other domains could be quite promising, provided the training data can be formulated with verifiable problems!
* Iteratively sample CoTs from the model, using a mix of different search strategies. This gives you something like Stream of Search via prompting.
* Verify correctness of each CoT using GPT-4o (needed because exact match doesn't work well in medicine where there are lots of aliases)
* Use GPT-4o to reformat the concatenated CoTs into a single stream that includes smooth transitions like "hmm, wait" etc that one sees in o1
* Use the resulting data for SFT & RL
* Use sparse rewards from GPT-4o to guide RL training. They find RL gives an average ~3 point boost across medical benchmarks and SFT on this data already gives a strong improvement.
Applying this strategy to other domains could be quite promising, provided the training data can be formulated with verifiable problems!
- 1 reply
Post
6898
We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute 🔥
How? By combining step-wise reward models with tree search algorithms :)
We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think"
We're open sourcing the full recipe and sharing a detailed blog post.
In our blog post we cover:
📈 Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
🎄 Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets.
🧭 Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM
Here's the links:
- Blog post: HuggingFaceH4/blogpost-scaling-test-time-compute
- Code: https://github.com/huggingface/search-and-learn
Enjoy!
How? By combining step-wise reward models with tree search algorithms :)
We show that smol models can match or exceed the performance of their much larger siblings when given enough "time to think"
We're open sourcing the full recipe and sharing a detailed blog post.
In our blog post we cover:
📈 Compute-optimal scaling: How we implemented DeepMind's recipe to boost the mathematical capabilities of open models at test-time.
🎄 Diverse Verifier Tree Search (DVTS): An unpublished extension we developed to the verifier-guided tree search technique. This simple yet effective method improves diversity and delivers better performance, particularly at large test-time compute budgets.
🧭 Search and Learn: A lightweight toolkit for implementing search strategies with LLMs and built for speed with vLLM
Here's the links:
- Blog post: HuggingFaceH4/blogpost-scaling-test-time-compute
- Code: https://github.com/huggingface/search-and-learn
Enjoy!
- 2 replies
shivi
authored
a
paper
2 months ago
shivi
authored
2
papers
3 months ago
hazemessam
authored
a
paper
3 months ago
louisbrulenaudet
posted
an
update
3 months ago
Post
1991
I’ve published a new dataset to simplify model merging 🤗
This dataset facilitates the search for compatible architectures for model merging with @arcee_ai’s mergekit, streamlining the automation of high-performance merge searches 📖
Dataset : louisbrulenaudet/mergekit-configs
This dataset facilitates the search for compatible architectures for model merging with @arcee_ai’s mergekit, streamlining the automation of high-performance merge searches 📖
Dataset : louisbrulenaudet/mergekit-configs
- 1 reply
louisbrulenaudet
posted
an
update
4 months ago
Post
1342
Introducing Lemone-router, a series of classification models designed to produce an optimal multi-agent system for different branches of tax law.
Trained on a base of 49k lines comprising a set of synthetic questions generated by GPT-4 Turbo and Llama 3.1 70B, which have been further refined through evol-instruction tuning and manual curation and authority documents, these models are based on an 8-category decomposition of the classification scheme derived from the Bulletin officiel des finances publiques - impôts :
It achieves the following results on the evaluation set:
- Loss: 0.4734
- Accuracy: 0.9191
Link to the collection: louisbrulenaudet/lemone-router-671cce21d6410f3570514762
Trained on a base of 49k lines comprising a set of synthetic questions generated by GPT-4 Turbo and Llama 3.1 70B, which have been further refined through evol-instruction tuning and manual curation and authority documents, these models are based on an 8-category decomposition of the classification scheme derived from the Bulletin officiel des finances publiques - impôts :
label2id = {
"Bénéfices professionnels": 0,
"Contrôle et contentieux": 1,
"Dispositifs transversaux": 2,
"Fiscalité des entreprises": 3,
"Patrimoine et enregistrement": 4,
"Revenus particuliers": 5,
"Revenus patrimoniaux": 6,
"Taxes sur la consommation": 7
}
id2label = {
0: "Bénéfices professionnels",
1: "Contrôle et contentieux",
2: "Dispositifs transversaux",
3: "Fiscalité des entreprises",
4: "Patrimoine et enregistrement",
5: "Revenus particuliers",
6: "Revenus patrimoniaux",
7: "Taxes sur la consommation"
}It achieves the following results on the evaluation set:
- Loss: 0.4734
- Accuracy: 0.9191
Link to the collection: louisbrulenaudet/lemone-router-671cce21d6410f3570514762