Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning
Paper
•
2503.07572
•
Published
•
31
IOPO Experiments
classroom
AI & ML interests
None defined yet.
Recent Activity
View all activity
iopo-exp's activity
lewtun
authored
a
paper
2 days ago
Post
1823
Introducing OlympicCoder: a series of open reasoning models that can solve olympiad-level programming problems 🧑💻
- 7B open-r1/OlympicCoder-7B
- 32B open-r1/OlympicCoder-32B
We find that OlympicCoder models outperform Claude 3.7 Sonnet, as well as others over 100x larger 💪
Together with the models, we are releasing:
📊CodeForces-CoTs: new dataset of code problems from the most popular competitive coding platform, with R1 traces in C++ and Python open-r1/codeforces-cots
🏆 IOI'2024: a new benchmark of VERY hard programming problems where even frontier models struggle to match human performance open-r1/ioi
For links to the models and datasets, check out our latest progress report from Open R1: https://huggingface.co/blog/open-r1/update-3
- 7B open-r1/OlympicCoder-7B
- 32B open-r1/OlympicCoder-32B
We find that OlympicCoder models outperform Claude 3.7 Sonnet, as well as others over 100x larger 💪
Together with the models, we are releasing:
📊CodeForces-CoTs: new dataset of code problems from the most popular competitive coding platform, with R1 traces in C++ and Python open-r1/codeforces-cots
🏆 IOI'2024: a new benchmark of VERY hard programming problems where even frontier models struggle to match human performance open-r1/ioi
For links to the models and datasets, check out our latest progress report from Open R1: https://huggingface.co/blog/open-r1/update-3
- 1 reply
alvarobartt
posted
an
update
17 days ago
Post
2843
🔥 Agents can do anything!
@microsoft
Research just announced the release of Magma 8B!
Magma is a new Visual Language Model (VLM) with 8B parameters for multi-modal agents designed to handle complex interactions across virtual and real environments; and it's MIT licensed!
Magma comes with exciting new features such as:
- Introduces the Set-of-Mark and Trace-of-Mark techniques for fine-tuning
- Leverages a large amount of unlabeled video data to learn the spatial-temporal grounding and planning
- A strong generalization and ability to be fine-tuned for other agentic tasks
- SOTA in different multi-modal benchmarks spanning across UI navigation, robotics manipulation, image / video understanding and spatial understanding and reasoning
- Generates goal-driven visual plans and actions for agentic use cases
Model: microsoft/Magma-8B
Technical Report: Magma: A Foundation Model for Multimodal AI Agents (2502.13130)
Magma is a new Visual Language Model (VLM) with 8B parameters for multi-modal agents designed to handle complex interactions across virtual and real environments; and it's MIT licensed!
Magma comes with exciting new features such as:
- Introduces the Set-of-Mark and Trace-of-Mark techniques for fine-tuning
- Leverages a large amount of unlabeled video data to learn the spatial-temporal grounding and planning
- A strong generalization and ability to be fine-tuned for other agentic tasks
- SOTA in different multi-modal benchmarks spanning across UI navigation, robotics manipulation, image / video understanding and spatial understanding and reasoning
- Generates goal-driven visual plans and actions for agentic use cases
Model: microsoft/Magma-8B
Technical Report: Magma: A Foundation Model for Multimodal AI Agents (2502.13130)
Post
4878
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
about 1 month ago
Post
10242
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
3891
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
2310
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
6933
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
JW17
authored
2
papers
4 months ago
nlee-208
authored
a
paper
4 months ago
nlee-208
updated
a
model
6 months ago
nlee-208
authored
2
papers
6 months ago
nlee-208
updated
a
model
7 months ago
alvarobartt
posted
an
update
7 months ago
Post
3013
🤗 Serving Meta Llama 3.1 405B on Google Cloud is now possible via the HF中国镜像站 Deep Learning Containers (DLCs) for Text Generation Inference (TGI)
In this post, we showcase how to deploy https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct-FP8 on an A3 instance with 8 x H100 GPUs on Vertex AI
Thanks to the HF中国镜像站 DLCs for TGI and Google Cloud Vertex AI, deploying a high-performance text generation container for serving Large Language Models (LLMs) has never been easier. And we’re not going to stop here – stay tuned as we enable more experiences to build AI with open models on Google Cloud!
Read the full post at https://huggingface.co/blog/llama31-on-vertex-ai
In this post, we showcase how to deploy https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct-FP8 on an A3 instance with 8 x H100 GPUs on Vertex AI
Thanks to the HF中国镜像站 DLCs for TGI and Google Cloud Vertex AI, deploying a high-performance text generation container for serving Large Language Models (LLMs) has never been easier. And we’re not going to stop here – stay tuned as we enable more experiences to build AI with open models on Google Cloud!
Read the full post at https://huggingface.co/blog/llama31-on-vertex-ai
kashif
authored
a
paper
9 months ago
JW17
authored
a
paper
9 months ago
alvarobartt
posted
an
update
10 months ago
Post
3230
🔥 Prometheus 2 was recently released by Kaist AI as an alternative and closely mirroring both human and GPT-4 evaluation, and surpassing the former Prometheus!
prometheus-eval/prometheus-7b-v2.0
prometheus-eval/prometheus-8x7b-v2.0
🌬️Fine-tuned on top of mistralai/Mistral-7B-Instruct-v0.2 and mistralai/Mixtral-8x7B-Instruct-v0.1
🗂️The datasets used for fine-tuning have been publicly released i.e. prometheus-eval/Feedback-Collection and prometheus-eval/Preference-Collection
🤝🏻Unified LM evaluator for absolute (a single prompt-completion pair) and relative (two completions for a given prompt) due to model merging
❌No longer needs a mandatory reference / golden answer, but can still be provided optionally
🔝Surpasses the former version of Prometheus, and has a high correlation with human, GPT-4, and Claude 3 Opus scores when evaluating LMs
📝Apache 2.0 license
Long-story short, an amazing job from Kaist AI bridging the gap with LLM evaluators other than proprietary and bigger models!
This week at Argilla, we decided to add a new task to use Prometheus 2 as an LLM evaluator using
😱 Using
Find the generated dataset and the code at distilabel-internal-testing/instruction-dataset-prometheus
prometheus-eval/prometheus-7b-v2.0
prometheus-eval/prometheus-8x7b-v2.0
🌬️Fine-tuned on top of mistralai/Mistral-7B-Instruct-v0.2 and mistralai/Mixtral-8x7B-Instruct-v0.1
🗂️The datasets used for fine-tuning have been publicly released i.e. prometheus-eval/Feedback-Collection and prometheus-eval/Preference-Collection
🤝🏻Unified LM evaluator for absolute (a single prompt-completion pair) and relative (two completions for a given prompt) due to model merging
❌No longer needs a mandatory reference / golden answer, but can still be provided optionally
🔝Surpasses the former version of Prometheus, and has a high correlation with human, GPT-4, and Claude 3 Opus scores when evaluating LMs
📝Apache 2.0 license
Long-story short, an amazing job from Kaist AI bridging the gap with LLM evaluators other than proprietary and bigger models!
This week at Argilla, we decided to add a new task to use Prometheus 2 as an LLM evaluator using
distilabel, so we implemented PrometheusEval.😱 Using
PrometheusEval running their 7B variant with vLLM in a single L40 on top of
HuggingFaceH4/instruction-dataset, we got the 327 existing prompt-completion pairs evaluated and pushed to the Hub in less than 2 minutes!Find the generated dataset and the code at distilabel-internal-testing/instruction-dataset-prometheus
- 1 reply